Hi friends,
Two weeks ago, I came across a viral video about how Manchester City used data science to help find Kevin De Bruyne’s replacement. That led me to write a post breaking down the process.
My second post looked at De Bruyne from the other side: how he used data to strengthen his position in his own contract negotiation.
Now, I’m no social-media expert, but considering I had 650 LinkedIn connections at the time, I think it’s fair to say both posts went a little bonkers. Together, they reached more than 226,000 unique people on the network.
So I decided to double down on the idea and created Data Deals FC.
Every week, I plan to take one player and assess his position in much the same way De Bruyne used data scientists to support the negotiation that helped him secure a 30% pay rise on an €80M contract.
And this is my first attempt: a full contract negotiation assessment of Kai Havertz.
He felt like the perfect first profile. He has started the new season strongly, his role at Arsenal has changed again, and he now has less than two years remaining on his contract.
The report looks at his performance, availability, wage, market value, replacement cost and negotiating position — and then asks the practical question: what should a new deal actually look like?
At the end of the article, you’ll find a link to download the full PDF version.
If you enjoy this kind of analysis, I’d love to hear your thoughts and criticism. And if there’s a player you think I should assess next, drop the name in the comments.
Without further ado, let’s get into it.
Introduction
This report examines the sporting and economic position of Kai Havertz (Arsenal FC, 27) in the context of a potential contract negotiation with his club. It combines a match-level playing-time record, performance data benchmarked against several peer groups, wage and market-value data, and transfer-market evidence, across Arsenal, the Premier League and Europe’s five major leagues.
Case-study assumptions. This is a public demonstration of an analytical framework, not a report on an actual negotiation. We do not know Arsenal’s negotiating position, the player’s wishes, or whether any negotiation is taking place. For the purposes of the analysis, the report assumes that both parties are willing in principle to explore a continuation of the relationship, but may differ on the appropriate economic terms and allocation of risk: the player’s side is assumed to seek to preserve or improve the economic value and security of his existing terms; the club is assumed to seek flexibility, risk management, and terms it believes his role and performance justify. Under those assumptions, the report deliberately asks how the same evidence could reasonably be used by both sides — including, in each section, the points each side would find hardest to defend.
How this differs from a commissioned report. A privately commissioned version of this analysis would begin from the client’s actual objective and would tailor the peer groups, comparable players, compensation scenarios and the selection of analyses to that objective. A player-side renewal report, a club-side sale assessment and a transfer-seeking player’s dossier would not share the same structure. This public report remains bilateral and broader than any single mandate would require, because its purpose is to demonstrate the complete framework.
How to read the report. Not every number carries the same weight. The report therefore separates three things throughout: observed facts (what the data directly records, such as minutes, appearances, contract dates, recorded values and estimated wages); modelled results (estimates produced by a statistical method, such as wage benchmarks, value projections or simulated sporting impact); and interpretation (what those facts and estimates could mean in a negotiation). The annex explains how each modelled result is produced.
Plain-language rule. The technical work sits behind the analysis, not in front of it. In the main report, statistical methods are explained through the practical question they answer; detailed specifications, transformations and assumptions belong in the annex. Where a technical term is unavoidable, it is defined before it is used.
Report structure. The report is organised around five questions, followed by a negotiation framework and contract options:
Current Sporting Position — playing time, availability, role, squad position and team results with/without the player.
Performance & Player Profile — underlying output, role-specific performance, profile characteristics and scarcity, benchmarked against peers.
Compensation Benchmarking — the current wage against squad, league and European comparisons, a statistical wage benchmark, similar-profile wages and the outside market.
Market Value & Contract Position — current value, historical evolution, forward scenarios, and the distinction between market value and the fee a contract allows a club to realise.
Replacement Economics & Club-Specific Value — what Arsenal would need to replace, how a departure could affect Arsenal and a direct rival, and the sporting, financial and profile cost of each path.
Negotiation Framework & Contract Options — what the evidence means for timing, the extend/sell/wait decision, the negotiating range and the structure of a potential agreement.
Each analytical section closes with a Negotiation interpretation: the arguments the evidence would permit each side to make, and the hardest point each side would have to defend. The report demonstrates what the evidence permits; it does not claim to know what either party would actually argue.
Executive Summary
The situation. Kai Havertz, 27, is under contract to June 2028 at an estimated €17.0M a year — Arsenal’s second-highest wage — after two injury-interrupted seasons. In 2025/26 he played only 647 Premier League minutes while the club won the league with new signing Viktor Gyökeres leading the attack: a top-of-market wage attached to very low usage, and a legitimate repricing question. Since the season ended, the evidence has moved: he scored in the Champions League final, scored three goals in four matches at the 2026 World Cup, and has started every Arsenal match of 2026/27 at centre-forward. The summer window closed without a sale. What remains to negotiate is the terms of a continuation, and who carries the risk that the 2025 injury pattern returns.
The strongest evidence for the player’s side. His pooled three-season Arsenal record — 5,675 league minutes, immune to small-sample objections — sits at the 89th percentile of the era’s Premier League attackers for attacking output and the 95th for aerial duels won, a combination almost no other player occupies; among centre-forwards his creation ranks at the 94th percentile. His usage is restored: every 2026/27 match started at centre-forward, ahead of Gyökeres. And he is expensive to replace: every external route to comparable output is estimated to cost at least as much as retention (Section 5), while the internal alternatives narrowed when Martinelli, Trossard and Jesus left in the summer window.
The strongest evidence for the club’s side. The last completed season is a fact: 647 league minutes, a 19% share of scheduled time, and the squad’s highest wage cost per league minute (≈€26,000 — roughly five times Rice or Gyökeres). The statistical wage benchmark values his observed 2025/26 profile at €13.5M against his current €17.0M wage. Behind those numbers sit two extended absences within thirteen months. And the early 2026/27 evidence is genuinely mixed: his usage has returned, but his attacking volume over the opening rounds has not yet been established at the level his rates elsewhere in this report assume.
What the team evidence can and cannot establish. The with/without record is inconclusive: Arsenal’s actual points were better with Havertz starting, but the difference disappears on expected-performance measures. Neither side can build a case on team results; the case rests on the individual evidence.
The principal uncertainty. Availability, not level. Since January 2026 he has played eight months without a new absence — through the title run-in, a Champions League final, a full World Cup and three complete 2026/27 league starts — but a full season at starter load remains undemonstrated, and most of the money rides on it: his projected 2028 market value spans roughly €30M to €72M on that question alone, and the wage benchmark moves from €13.5M at his observed 2025/26 usage to €16.7M — essentially his current wage — at restored usage.
Timing and the contract clock. Waiting is not neutral, and its first instalment has been paid: the last near-full-value sale window (a counterfactual ~€55M in summer 2026) has closed, and the achievable transfer market fee falls with each remaining window — roughly €47M in January 2027 and €38M by next summer on the base path, zero at expiry. The contract clock changes who captures the value: what the club cannot convert into a fee migrates to the player’s side of the table as expiry approaches.
The deal structure. Under the assumptions modelled here, continuation increasingly dominates the remaining exit options: a summer-2027 sale is worth less than keeping him under any view of his fitness, a January sale pays only under deep pessimism that sits awkwardly with the club’s current selection decisions, and running the contract down means paying two seasons of top wages for a player who then leaves for free. The question therefore shifts from whether to continue to on what terms. The evidence supports a structure that separates guaranteed from availability-linked compensation: a guarantee within the €13.5–16.7M benchmark range, with upside that restores or exceeds his current pay when he is available and productive. Such a structure costs the club roughly what a fixed wage costs in expectation while sharply reducing the downside of another lost season — in effect, the contract prices the uncertainty rather than forcing either side to win the argument in advance.
Bottom line. Havertz’s negotiating position has strengthened materially since the end of 2025/26. His role has returned to starter level, his underlying performance record remains strong, Arsenal’s internal replacement options have narrowed, and the club’s highest-value sale window has passed. The unresolved question is no longer primarily his playing level, but whether his availability will hold over a full season. Under the assumptions modelled here, the negotiation is therefore less about whether to continue and more about how much of his compensation should be guaranteed, how much should depend on availability, and for how long.
Negotiation Snapshot
1. Current Sporting Position
Playing time is the most basic indicator of a player’s standing: it reflects, before any performance metric, how his club actually uses him. This section reviews Havertz’s playing time at three levels of detail — his career workload by season, his share of league minutes, and his current position within the Arsenal squad — based on a match-level record of 488 competitive appearances between February 2016 and September 2026.
Career workload
Havertz’s career divides into two phases: seven seasons of established-starter workload, then two injury-interrupted ones. Between 2017/18 and 2023/24 he sustained the workload of an established first-team player, between roughly 3,000 and 5,100 competitive minutes per season, with 2023/24 — his first season at Arsenal — the heaviest of his career at 5,148 minutes across league, European, cup and international football.
Figure 1. Career playing time by season.
The two most recent complete seasons break that pattern. His competitive volume fell to 3,387 minutes in 2024/25 and to 1,299 club minutes in 2025/26 — approximately one quarter of his 2023/24 workload, and the lowest club figure of his senior career. The June 2026 international window then added 459 national-team minutes — a friendly and four World Cup matches — lifting the season’s total competitive volume to 1,758, visible in the national-team block of Figure 1. The 2026/27 bar is the season in progress: 358 competitive minutes by 6 September, all as a starter.
Share of league minutes
Expressed as a share of scheduled league minutes, the same trajectory is visible with more precision. Havertz exceeded an informal established-starter threshold (80% of scheduled minutes) four times: twice at Leverkusen (99% in 2018/19, 84% in 2019/20), and in his final Chelsea season and first Arsenal season (82% and 86%). His first two Chelsea seasons, by contrast, were rotation seasons (47% and 57%).
Figure 2. Share of scheduled league minutes played, by season.
The two completed-season values at the right of the line stand apart from the rest of the series. In 2024/25 his share fell to 61%; in 2025/26 it fell to 19% — 647 Premier League minutes across the season. Within this ten-season record, 2025/26 is not a below-average season; it is an outlier. The hollow 2026/27 point is the season in progress, measured only against the three rounds played: he has been on the pitch for essentially all of them. It is an early-season reading, not a completed-season value — but it is the first point since 2023/24 to sit inside the starter band.
Availability
The match-by-match record identifies the immediate cause. Across nine seasons — from his 2016 youth appearances through every club and international window, including the 2022 World Cup and Euro 2024 — his calendar is nearly unbroken. It then shows two extended absences: 102 days without a competitive appearance between February and May 2025, coinciding with his publicly reported hamstring injury, and a second absence of 147 days between August 2025 and January 2026. The record since is a single unbroken run: regular but managed club appearances through the spring (23 club appearances to the end of 2025/26 averaging 50 minutes), then a rising load — 104 minutes in the Champions League final, 395 minutes across four World Cup matches including one decided in extra time, and full-length starts in every 2026/27 match through 6 September. No new absence has appeared in the eight months since his return.
Figure 3. Match-by-match minutes, 2016–2026.
The figure in a nutshell: The gaps are two discrete absences in 2025; around them the calendar is nearly unbroken, including eight absence-free months since January 2026.
Two observations follow. First, the reduced 2024/25 and 2025/26 figures appear availability-driven rather than selection-driven: when available across 2022/23–2023/24, he played at an 82–86% league share, and when available again from August 2026 the club immediately returned him to full starts. Second, the record contains no evidence of chronic fragility before 2025 — across the seven prior seasons he never missed an in-season period of five weeks or more — and none after January 2026. The two absences increasingly resemble discrete events bracketed by normal availability; but eight months is not a season, and the match record registers only matches played, so these gaps measure unavailability without distinguishing its cause.
Role
Havertz’s positional use has migrated across his career: predominantly attacking midfield at Leverkusen and in his early Chelsea seasons, then a progressive conversion to centre-forward that peaked in 2024/25, when 94% of his club minutes came at centre-forward. In 2025/26 that concentration reversed — 44% centre-forward, 35% attacking midfield, 22% central midfield — reflecting both his mid-season return and the arrival of Viktor Gyökeres, who accumulated 2,494 Premier League minutes at centre-forward in 2025/26. The 2026/27 season has so far reversed it again, entirely: every competitive start — the Community Shield and three league matches — has come at centre-forward, with Gyökeres’s 16 league minutes all from the bench. The hybrid-role question that shaped last season’s peer-group choices in Section 2 is, for now, resolved by the club’s own selections: he is being used as the No. 9.
Figure 4. Share of club minutes by starting position role.
Position within the Arsenal squad
The squad-level view summarises his current standing. In 2025/26 Havertz ranked 20th of the 25 players Arsenal used in the Premier League — below every established outfield starter, below the other options in his positions (Gyökeres 2,493 minutes, Eze 2,015, Merino 1,190), and ahead of only a handful of rotation and cover players. Two seasons earlier he had been one of the squad’s most-used outfield players.
Figure 5. Arsenal Premier League minutes hierarchy, 2025/26.
The 2026/27 season, three rounds old, has inverted that hierarchy. Havertz ranks 4th of the 19 players Arsenal has used in the league by minutes (281 of a possible 298, behind only the three ever-present defensive players), with three starts from three, two goals and an assist; Gyökeres has 16 league minutes, all as a substitute. Madueke and the summer signing Tzolis have carried the wide rotation in the reshaped front line. Three rounds establish selection intent, not a season — but the newest evidence now points the same way as the era view below, not against it.
Figure 6. Arsenal Premier League minutes hierarchy, 2026/27 through 6 September (3 rounds).
The pooled era view remains the player’s strongest usage exhibit, and it is now consistent with the current one. Across the 37 players the club used over 2023/24–2025/26 — including those who have since left — Havertz ranks 9th in total Premier League minutes, despite 2025/26 contributing only 647 of his 5,675. Restricting to attackers at the club in all three seasons, he is fourth of the five ever-presents — behind only Saka, Trossard and Martinelli — the latter two since sold — and well clear of Jesus, also sold. Over the full period the club used him as one of its core outfield players; 2025/26 is the deviation from that pattern, and 2026/27 has so far resumed it.
Figure 7. The Arsenal era in total: pooled Premier League minutes of the club’s 25 most-used players, 2023/24–2025/26.
Team results with and without Havertz
With Havertz starting, Arsenal took 2.29 points per game; in the matches he missed entirely, 1.90 — a gap of about four tenths of a point per game (0.39), worth roughly 15 points over a season if taken at face value. It should not be taken at face value. Alongside actual points, the figure and table also report expected points: each match’s chances for and against converted into the points a team creating and conceding those chances would take on average, which strips out finishing and goalkeeping luck. On that fairer yardstick the two groups are level (2.04 with him starting, 2.06 without). Seventeen shorter appearances (under 60 minutes) are excluded from this headline comparison — Arsenal won thirteen of them, mostly matches already under control when he came on, so they say more about when managers make comfortable substitutions than about his effect; they are counted in the alternative groupings reported in the annex.
Figure 8. Arsenal’s Premier League results with and without Havertz, 2023/24–2025/26: raw averages and the adjusted no-Havertz estimate (substitute appearances excluded).
Table 1. Arsenal Premier League results, matches Havertz started vs matches he did not play, 2023/24–2025/26 (17 shorter appearances excluded — see text).
The adjusted comparison below asks whether the raw gap survives context. Controlling for opponent strength (the opponent’s final league points that season), venue and season, the no-Havertz penalty on actual points keeps its size (−0.44, though the confidence band spans −0.99 to +0.11), while expected points (+0.10) and xG difference (+0.08) show no gap at all: Arsenal’s underlying performance was maintained in his absence. None of this is causal; it is what 97 matches can and cannot show. The annex sets out the adjustment step by step.
Table 2. Adjusted difference, matches without Havertz versus matches he started, controlling for opponent strength, venue and season; the 95% confidence band is shown in brackets.
What this means for the negotiation. Neither side can lean hard on with/without records. The player’s side gets the headline (2.29 with, 1.90 without, direction intact after adjustment and under every alternative grouping); the club gets the counter (on expected points, the squad performed identically without him). The honest reading is that team results neither prove his indispensability nor prove the squad outgrew him — which pushes the argument back to the individual evidence of Section 2, where his case is stronger.
Negotiation interpretation – Sporting Position
Evidence available to the player’s side — the established record, with the recent seasons framed as an interruption:
An established starter when available. Havertz played 82–86% of league minutes across 2022/23–2023/24 and carried the heaviest workload of his career in 2023/24. The club’s own selection decisions over two full seasons document his fit-state role.
No pre-2025 fragility. Seven seasons without a five-week in-season absence; the two recent absences are discrete events in an otherwise strong availability record.
Current usage now favours the reintegration reading. The 23 managed post-return appearances of spring 2026 were followed by a Champions League final start, a full World Cup and three full league starts with Gyökeres benched — the team sheets are observable facts, and the most recent ones on record.
Versatility. The 2025/26 role split (centre-forward, attacking midfield, central midfield) documents cover for three positions.
Potential club counterarguments — the most recent evidence, priced as risk:
The current facts. 647 league minutes (2025/26), a 19% share, 20th of 25 players used: on the season just completed, remuneration and role have diverged.
A new availability risk. Two absences totalling roughly 249 days within thirteen months, at ages 25–26, in a profile that previously had none — a discontinuity a club could reasonably price into the guaranteed portion of any offer.
Increased competition for his primary position. Gyökeres accumulated 2,494 league minutes at centre-forward in 2025/26; a return to 80%+ of minutes requires either displacing an established competitor or a hybrid role.
The availability record, not the role, is the live risk. The seasons that establish his durability at starter load remain two years old; eight months of sustained availability since January 2026 — much of it at partial club load — is supportive but is not yet a full season carried at starter minutes.
Hardest point for the player to defend: the 2025/26 usage numbers are facts, and the favourable current evidence is three league matches deep. The reduced-minutes season happened; a repricing argument built on it cannot be answered with an opening month alone.
Hardest point for the club to defend: treating the recent availability problem as a permanent change means placing heavy weight on two injury interruptions after seven seasons without a comparable absence. The club can price that risk; the data cannot yet show that it is chronic.
Both readings can be supported from the same record; the sections that follow test which parts of the player’s level have actually changed and how the market prices the two readings.
2. Performance & Player Profile
This section evaluates Havertz’s underlying attacking performance. Because he has played both as a centre-forward and as an attacking midfielder, his performance is assessed against both positional standards, using two weighted composite indexes in addition to the standard output metrics. All comparisons are per 90 minutes of league play, which separates the question of how well he plays from the question of how much he plays. The two questions receive opposite answers.
Career league record by season, 2016/17–2026/27 (as of 8 September 2026).
2.1 Core attacking output
A useful starting point is expected goals plus expected assisted goals per 90 (xG+xAG). In practical terms, it combines the quality of the shooting chances a player takes with the quality of the chances he creates for teammates, per 90 minutes of league play — the standard one-number summary of attacking output, and the basis for the valuations later in this report. Where a value excludes penalty kicks, the report marks it with the prefix np — npxG is non-penalty expected goals — and each figure’s labels state the basis actually used.
Figure 9. npxG and expected assists (xAG) per 90 by season.
Havertz’s per-90 output has risen, with one dip, across his career: 0.34–0.39 in his first two Leverkusen seasons, 0.59 in his breakout 2019/20, back down to 0.41 in his difficult first Chelsea season, then 0.51–0.58 across his remaining Chelsea and first two Arsenal seasons — including 0.58 in the injury-shortened 2024/25 — and 0.78 in 2025/26, his highest recorded rate and the tenth-highest among all 2,081 qualifying top-5-league players that season. For reference, the median top-5-league forward produced roughly 0.38 per 90 in 2025/26.
The delivered numbers tell the same story, with one instructive exception. A striker or attacking midfielder is ultimately judged on goals and assists, so the next figure places what he actually delivered next to the expected series, pair by pair. Both sides of the scoring pair exclude penalty kicks — his six league penalty goals (three in 2018/19) and their penalty xG are removed — so the gap between the lines reads finishing against chance quality, not penalty duty. The scoring pair ran hot at Leverkusen (0.44 non-penalty goals per 90 against 0.30 npxG in 2018/19), cold at Chelsea (0.21 against 0.38 in 2022/23), and close to expectation at Arsenal (0.41 vs 0.39, then 0.43 vs 0.46). In 2025/26 the two pairs diverged: he scored under his chances (0.28 goals per 90 against 0.50 npxG — two goals from 3.6 expected) while assisting well above them (0.42 assists against 0.28 xAG), netting out just below his combined expected rate. In percentile terms, his delivered non-penalty goals plus assists per 90 ranked at the 89th, 88th and 94th percentile of top-5-league attackers over his last three completed seasons.
Figure 10. Non-penalty goals, assists, npxG and xAG per 90 by season — delivered numbers solid, expected series dotted.
The principal qualification is sample size: the latest point rests on 647 league minutes. Even discounting it entirely, the pattern through 2024/25 stands — his output rate has never declined for two consecutive seasons, and the injury years did not interrupt it.
2.2 Two position-facing indexes
No single metric captures Havertz’s hybrid role. We therefore build two summary scores: one asks how his statistical profile looks against centre-forwards, the other how it looks against attacking midfielders. Before combining the metrics, each is put onto the same scale so that goalscoring, creation, aerial play and progression can be compared fairly. The detailed standardisation and weights are set out in the annex. As a check, the scores were also rebuilt without choosing the weights ourselves; the check shows no fundamental differences — slightly lower percentiles when every dimension counts equally — with the details in the annex.
Striker-facing index — npxG (30%), touches in the box (15%), xAG (15%), shot creation (10%), progressive receptions (10%), aerial win rate (10%), take-ons (10%) — scored against forwards.
Creator-facing index — xAG (20%), key passes (20%), shot creation (15%), npxG (15%), passes to the box (10%), progressive passes (10%), progressive carries (10%) — scored against midfielders.
Each score is calculated relative to players in the same season and position group. A percentile of 85 therefore means that the player ranks above 85% of the relevant comparison group. This makes the result easier to interpret across seasons even when the overall level of a league changes.
Figure 11. The two indexes by season, including 2025/26. Each index is shown only for the seasons in which he actually played that role.
Three findings. First, in every season in which a role was substantially his, Havertz has scored above that role’s positional median on the corresponding index. Second, the striker-facing index has been stable in the 62nd–84th percentile band across 2020/21–2024/25 and reads 97th in the small 2025/26 sample. Third, the creator-facing index traces his role history: from a Leverkusen peak (94th percentile) it settled in the low-to-mid 70s in the seasons when creation remained a substantial part of his role — it is not scored at all in 2022/23 and 2024/25, his most striker-weighted seasons, which is itself the statistical footprint of his conversion into a No. 9 — before rebounding to the 89th in 2025/26, when his role again included midfield minutes.
2.3 The latest season versus the Arsenal norm, metric by metric
Because 2025/26 contains only 647 league minutes, it is useful to compare it with his broader Arsenal record against exactly the same group of Premier League peers. That way, any movement in the chart comes from Havertz’s own numbers rather than from changing the comparison group.
Figure 12. Arsenal era vs 2025/26 against the centre-forward standard (same pool of 51 PL centre-forwards, 1,500+ minutes pooled).
Against the centre-forward standard, the 2025/26 season is not a diminished version of his Arsenal norm — on the attacking metrics it is a sharpened one: npxG per 90 rose from the 74th percentile of the era’s centre-forwards to the 90th, xAG from the 90th to the top of the distribution, box touches from the 88th to the 94th, shot assists from the 94th to the 98th. What eased is the physical-running side: aerial duels won (86th → 70th), progressive runs (66th → 52nd) and pass accuracy (92nd → 72nd) — the components that depend on legs and sustained load, consistent with a player managed back from injury in short appearances.
Figure 13. The same comparison against the attacking-midfield standard (83 PL attacking midfielders).
The attacking-midfield card completes the picture. By midfield standards his box threat is exceptional in both periods (npxG 98th → 100th percentile, box touches 89th–92nd, aerial duels won at or near the top of the pool), and his 2025/26 creation rises sharply (xAG 39th → 98th, shot assists 36th → 81st), while his carrying remains far below the midfield norm throughout (progressive runs 16th → 8th, successful dribbles 8th → 7th). He produces like a striker against midfielders and creates like a midfielder against strikers — and does either without the ball-carrying volume typical of both.
2.4 The Arsenal years against Premier League peers
Single seasons are noisy — his 2025/26 especially. Pooling his three Arsenal seasons (2023/24–2025/26, 5,675 Premier League minutes) and benchmarking against every Premier League attacker with 1,500+ minutes over the same window gives the most stable picture of what Arsenal have actually received. The pooled metrics are penalty-free — npxG-based — so penalty duty does not tilt the comparison.
Figure 14. Pooled 2023/24–2025/26 percentiles versus the era’s 173 Premier League attackers (attackers = centre-forwards, attacking midfielders and wingers).
The pooled record is strong in the scoring-and-aerial dimensions — npxG (91st percentile), box touches (87th), aerial duels won (95th) — while creation measured against the all-attacker pool is mid-pack (xAG 56th; against centre-forwards specifically it ranks at the 94th percentile, as the second scatter below shows) and the carrying-and-passing dimensions are modest (progressive runs 30th, dribbles 14th, through passes 21st). His profile is not that of an all-round attacker; it is a specific and unusual combination, which a few scatters will make visible.
The unusual striker
The scatters below place every attacker of the era on the same canvas, to show where his profile actually sits: the first pairs attacking output with aerial duels won — the region he occupies is close to unshared — and the second, restricted to centre-forwards, pairs output with creation, the precise version of the “creative striker” claim.
Figure 15. Attacking output vs aerial duels won, pooled 2023/24–2025/26.
Across the era’s 173 Premier League attackers, output and aerial presence trade off for most players: the highest-output attackers (Haaland, Salah) win few aerials, and most high-aerial attackers produce less. The combination Havertz offers — 89th-percentile output with 95th-percentile aerial volume — has almost no neighbours: the only attacker at or above him on both dimensions is Šeško, with Rodrigo Muniz and Beto nearby. Dot sizes in the figure are proportional to pooled minutes, so established records weigh more than part-season ones.
Figure 16. Output vs shot assists among Premier League centre-forwards, pooled.
Read against the scatter, the claim sharpens. Among the era’s centre-forwards, only Julián Álvarez pairs clearly more creation with the same output — 2.28 shot assists per 90 against Havertz’s 1.37, at a near-identical 0.61 npxG+xAG per 90, in his Manchester City season before the move to Atlético — and it is no surprise that Arsenal reportedly wanted him back in the Premier League. The only other No. 9 at or above Havertz on both axes is Gabriel Jesus (1.45 shot assists per 90 at 0.60 npxG+xAG), whose era minutes came at Arsenal itself: the nearest precedent for the profile was the club’s own previous centre-forward, sold in the summer window. Against all attackers — wide creators such as Saka or Palmer included — his creation is mid-pack; restricted to centre-forwards it sits at the 94th percentile alongside 78th-percentile output. The creative premium is real, but it is a premium for a No. 9, not an absolute one. Dot sizes are again proportional to pooled minutes.
Clean touch & Shooting diet
The aerial and creation marks above were chosen; a final check let the data choose instead. Scanning the per-90 rates and ratios of the pooled era — 173 Premier League attackers — surfaces two further marks the market rarely combines with output, each shown below against npxG+xAG per 90.
Clean touch. On the pooled record he completes 79% of his passes — the 71st percentile of the era pool, and above the median of the high-output group, where security usually collapses. For a centre-forward who plays with his back to goal, above-median security at top-eighth output is the linking-forward claim in one number — real, though the more modest of the two fingerprints.
Figure 17. Pass accuracy vs attacking output per 90, pooled 2023/24–2025/26.
Shot diet. He shoots about once every 37 minutes (71st percentile for volume), and each shot is on average close to a one-in-five chance — 0.18 npxG per shot, better than 91% of the pool — and he converts at roughly that expected rate. Quality with volume, and no finishing-luck inflation: his output level is earned from positions, not from a hot streak, which is one reason the projections in Section 4 can take his rate seriously.
Figure 18. Average chance quality per shot vs shot volume, pooled 2023/24–2025/26.
Taken together, the two fingerprints raise the practical bar for replacement further: a buyer must match not just output plus aerials, but output plus aerials plus pass security at a high-quality shot diet — a combination the similarity search already showed has no close market neighbour.
Finishing: goals against expectation
Every valuation in this report prices Havertz on expected output — xG and xAG. That choice is only defensible if his expected numbers actually turn into real ones over time, so this check compares the goals he scored with the chances he took, season by season. The ledger is strikingly clean: over eight full league seasons he has scored 67 non-penalty league goals from about 68 npxG — a career finishing balance of −0.8 — and the Arsenal era specifically closes at −0.1, as neutral as finishing gets (46th percentile of the era’s forwards, in a pool where the typical spread is ±3 goals). Inside that flat career line sit two stories that each side will recognise: hot finishing at Leverkusen (+6.9 goals over expectation across his two breakout seasons) and a finishing trough at Chelsea (−7.0, of which −4.8 in 2022/23 alone) — the seasons that built and then damaged his reputation were, in meaningful part, finishing variance around an underlying level that changed far less.
Figure 19. Non-penalty goals vs npxG by league season, 2017/18–2025/26.
Both sides get something from this figure, which is why it belongs in the report. The player’s side gets validation: none of his output carries finishing-luck inflation, the rates every model in Sections 4 to 6 relies on are earned from chance quality rather than a hot streak, and the Chelsea narrative — the trough his critics remember — mean-reverted completely once the context changed. The club’s side gets a genuine, subtler point: he is a neutral finisher, not an elite one. He adds no conversion premium over the chances the system creates for him, which makes his goal output more dependent on the supply line than a clinical finisher’s would be — a fair consideration when pricing how his numbers would travel to a different system, and one the similarity analysis of Section 3 already hints at.
The same expected numbers can also be split into their two components — the shots he takes himself and the chances he creates for teammates — to show where the output comes from. On the pooled era he is scorer-first: 91st percentile on scoring threat (npxG per 90), 56th on creation (xAG per 90), and the only attackers at or above him on both components are Salah, Núñez and Jota. Read together with Figure 16, this is the creator-who-finishes profile seen from the other side: among all attackers he is an elite chance-taker with above-median creation; among fellow No. 9s his creation is elite outright.
Figure 20. Scoring threat vs creating threat: npxG per 90 vs xAG per 90, pooled era.
Are the apparent weaknesses player weaknesses or role effects?
At the start of the section, we spotted four ‘weak’ metrics. Let’s see if they were really weaknesses or role effects. The test is simple: if a low percentile is a role effect, it should rise when the pool narrows to players asked to do the same job; if it is a genuine individual weakness, it should stay low in every pool. The four metrics split into those two cases.
Figure 21. The low-ranked carrying and passing metrics, re-benchmarked against role-appropriate pools.
Two of the four are clearly role effects, one is partly, and one is real. Against the centre-forward pool — the players who share his primary job — progressive runs rise from the 30th percentile (all 173 attackers) to the 66th and passes to the box from the 31st to the 80th: comfortably above the centre-forward median. Through passes rise from the 21st to the 47th — roughly the centre-forward median, no more (and the metric pairs least cleanly across providers, so it is the least precise of the four). Scored against attacking midfielders — his other principal role — all four metrics sit in single digits to the teens, which is exactly the signature of a No. 9’s action profile measured against players whose job is delivery. The headline pool is dominated by wide players and creative midfielders who accumulate carrying and delivering actions by role; centre-forwards who play back-to-goal and finish moves simply do not.
One weakness survives every pool. Successful dribbles stay low everywhere: 14th percentile against all attackers, 31st against centre-forwards, single digits against attacking midfielders, and the mid-20s among Arsenal’s own attackers. Take-on ability is a genuine limitation of his game, not an artefact of the comparison group.
Negotiation implication. The club can legitimately press on dribbling; it cannot legitimately price the carrying and passing percentiles as weaknesses, because they largely reflect the role Arsenal itself assigns him. Conversely, the player’s side should concede the take-on point rather than contest it — conceding the one genuine weakness strengthens the credibility of rejecting the three that are not.
The profile case, in sum. Taken together, the profile chapter leaves each side a clear brief. The strengths are rare in combination rather than extreme in isolation: 91st-percentile npxG alongside 95th-percentile aerial volume (Figure 15 shows how thinly that corner of the market is populated), creation at the 94th percentile of centre-forwards with only Álvarez clearly above him at the same output (Figure 16), and the clean-touch and shot-selection fingerprints of a link-forward (Figures 17–18). The weakness list that survives every pool has one entry: take-ons. For the player’s side, the argument is scarcity — Section 5 prices what buying this combination would cost. For the club, the argument is concentration — the case rests on profile and availability rather than headline goal totals, which is why the wage discussion in Section 3 turns on usage rather than on quality.
2.5 Role-conditional output and current form
Splitting his club output by the role he started in shows both profiles are productive: at Arsenal, 0.54 xG+xAG per 90 as a centre-forward and 0.60 in his (smaller) attacking-midfield sample. On output alone, neither role choice is forced.
Figure 22. xG and xAG per 90 by starting role, per club spell.
Figure 23. Rolling 5-match xG+xAG per 90, all competitive matches since 2017/18; dotted lines mark club changes.
Finally, the rolling form curve — deliberately short at five matches, so streaks are visible — places his current level in career context, and now runs through the World Cup and the opening 2026/27 matches. The window restarts at every season boundary, so no reading mixes matches from two different seasons; the breaks in the line are season changes and absences. The curve is naturally volatile at this length, but the post-return segment oscillates around and above his career median, the final weeks of 2025/26 reach some of the highest readings of his career, and the newest segment — World Cup and early 2026/27 — holds the same level. On a rate basis, the returning player is not measurably below his career standard.
The opening weeks of 2026/27: an early snapshot
On 281 league minutes of the current 2026/27 season, the profile is recognisably his: shot assists (86th percentile of the 84 qualifying top-five centre-forwards), aerial duels won (83rd) and creation (71st on xAG) lead; attacking volume trails — 11th percentile on xG+xAG per 90, with his two goals scored well ahead of the 0.5 expected goals underneath them. Three rounds into a season, that mix — full minutes, elite service and aerial presence, thin shot volume — reads as a No. 9 whose team is early in its attacking rhythm, and it cuts both ways at the table: usage evidence for the player, volume caution for the club. The completed-season benchmarks of Sections 2.1–2.4 remain the report’s performance evidence.
Figure 24. Early 2026/27 snapshot vs top-five-league centre-forwards (281 minutes, 3 matches).
What this means for the negotiation. The role split and the form curve close the performance case on its two live questions: his output survives either role — which protects his value under a manager’s tactical choices — and his current level, through the run-in, the World Cup and the opening weeks, sits at or above his career norm. Neither finding proves durability; both price it more expensively for the club to dispute.
Negotiation interpretation – Performance & Player Profile
Evidence available to the player’s side:
Output per 90 is at a career high, not in decline. The npxG+xAG trajectory has risen through nine consecutive seasons, including the injury years; the 2025/26 rate is the tenth-highest among 2,081 qualifying top-5-league players.
The pooled Arsenal record withstands the small-sample objection. Over three seasons and 5,675 minutes he sits at the 89th percentile of the era’s Premier League attackers on output, 87th on box presence, 95th on aerials won.
A profile with few substitutes. The output-plus-aerial region of the market is sparsely populated (Figure 15); among centre-forwards his creation is near the top of the distribution (Figure 16). Replacing the combination would require either two players or a substantial fee (Section 5).
The form curve is inconsistent with diminishment. Since returning, his rolling 5-match output oscillates at or above his career median — through the run-in, the World Cup and the opening 2026/27 matches.
Potential club counterarguments:
Per-90 rates on 647 minutes carry wide uncertainty. Substitute appearances can inflate per-90 output; a quarter-season rate is indicative, not established.
The pooled record describes a specialist. The same three-season card shows 14th-percentile dribbling, 30th-percentile carrying and mid-pack creation against the full attacker pool; and in his last full fit season (2023/24) his output ranked 64th among Premier League forwards, behind Arsenal’s most-used attackers.
The creator premium is conditional on the position. His creation is exceptional relative to centre-forwards specifically; Arsenal’s options at centre-forward reduce the leverage of that comparison.
Volume is part of performance. Totals, not rates, decide matches; on totals (Section 1), his last two seasons rank among the squad’s lowest. And the early 2026/27 snapshot shows restored minutes with thin attacking volume so far (11th percentile on xG+xAG per 90 through three rounds) — starter usage and restored output are separate claims.
Hardest point for the player to defend: the strongest recent rate evidence comes from precisely the minutes profile (short, substitute-heavy) that the club’s usage argument discounts.
Hardest point for the club to defend: nine seasons of rising output, including through both injuries, leave no data-based route to the claim that his underlying level has declined.
The performance evidence, unlike the playing-time evidence, favours the player’s side: it is consistent with an intact — possibly improved — underlying level. The club’s strongest counters concern proof and context, not level. What the market pays for this combination is the subject of Section 3.
3. Compensation Benchmarking
Wages are the negotiation’s third anchor, and the bridge between the sporting record (Sections 1–2) and market value (Section 4). Unlike minutes or output, a wage does not adjust to events — it is whatever the last contract fixed. This section places Havertz’s estimated €17M annual wage in 2026/27 against his squad, his league and Europe; against a statistical wage benchmark; against the wages of statistically similar profiles; and against his realistic outside market. All wage figures are Capology estimates of base salary: approximate for any individual, consistent across players.
The squad ladder
The ladder below shows the contracts in force today, 2026/27: Saka’s renewal (€18.3M) makes him the squad’s top earner, with Havertz second at €17.0M — the same contract as the €16.8M 2025/26 estimate, simply re-estimated; the decision models of Sections 5 and 6 price him at this current €17.0M — ahead of Saliba (€15.2M) and Rice and Ødegaard (€14.6M each). The season behind this report carried the tension in one line: in 2025/26 Havertz was the squad’s highest earner, and that wage was attached to 647 league minutes — the highest estimated wage cost per Premier League minute in the squad, approximately €26,000, roughly five times Rice or Gyökeres.
Figure 25. Arsenal wage ladder, 2026/27: the contracts in force today.
Six altitudes
The pattern of Section 1 inverts. His minutes rank near the bottom of every pool; his wage ranks at or near the top of every pool on the contracts in force in 2026/27 — 95th percentile within the Arsenal squad (second only to Saka), 100th among its attackers, 98th in the Premier League, 99th among Premier League attackers and 99th across all waged top-5-league players.
Figure 26. Wage percentile at six scopes, 2026/27.
The market map
Placing wage against output puts both negotiating claims on one canvas. At €16.8M his wage neighbours are Musiala, Vlahović, Raphinha, Olise. Read with his 2025/26 rate (triangle), he sits among the players whose output matches top-of-market pay. Read with his pooled Arsenal-era output (circle), he is paid with the very best and produces above the median but below the frontier — a wage set in 2023, before the injury record, that anticipates the triangle.
Figure 27. Wage vs attacking output per 90, top-5-league attackers, 2025/26 — with both Havertz readings.
The statistical wage benchmark
A statistical wage model provides the benchmark. It asks a practical question: what are players with similar observable characteristics typically paid? The model uses 1,436 top-five-league outfield players with 500+ minutes in 2025/26 and accounts for attacking output, playing time, age, club spending power, league and position. It does not claim to produce a ‘fair’ salary. It provides a market benchmark from observable characteristics, while leaving room for the things contracts also price but public data cannot fully observe — scarcity, status, commercial value, timing and negotiating leverage. Table 3 translates the main effects into plain language; the full statistical specification and complete regression output — every coefficient with its standard error, t-statistic and confidence interval — are reported in the annex (Table 10), not here, to keep this section readable.
log(wage) = a + 0.745 · (xG+xAG per 90) + 0.407 · (share of league minutes) + 0.945 · log(club wage bill) + 0.454 · age − 0.007 · age² + league + position
Figure 28. Actual vs model-expected wage, 2025/26 cross-section.
Table 3. The regression output in plain terms (R² = 0.71, n = 1,436 players, 2025/26).
How to read this table: the model predicts wages on a logarithmic scale, so the coefficients can be interpreted approximately as percentage effects for small changes. For example, a 0.10 increase in xG+xAG per 90 is associated with roughly 8% higher wages, holding the other factors constant. The final column applies each relationship to Havertz’s own characteristics. The model combines all of these inputs simultaneously to produce his estimated wage benchmark. At Havertz’s observed 2025/26 inputs, the model estimates a wage of €13.5M. Under the report’s restored-usage scenario, the estimate rises to €16.7M. Full regression results and scenario inputs are reported in Annex Table 10.
At his observed 2025/26 inputs, the model estimates €13.5M — around €3.3M below his 25/26 salary. That gap is meaningful but not exceptional: roughly one quarter of players in the sample are paid an even larger premium over what the model predicts. More importantly, the estimate changes sharply when we change the playing-time assumption.
The age curve
Havertz has earned at or above the 90th percentile of his age group since his early twenties, and his current wage sits well above the 90th-percentile line. Two facts matter for the next contract. First, a renewal would cover ages 27–30, years in which attacker wages have not yet declined in this data: the model’s own age curve peaks around 32, and in the raw panel the top decile of wages keeps rising into the early thirties — a survivor effect, since only the best older players keep such contracts, but that is exactly the cohort he would be negotiating to stay in. Second, the cohort’s 90th-percentile wage (≈€8–10M) sits far below his current level, so the age curve alone will not carry his wage forward; his own record must. The panel pairs wages with the completed 2025/26 season — the wage-vintage rule of the annex; his 2026/27 estimate, €17.0M, is the same contract re-estimated.
Figure 29. His wage path against the attacker wage-age curve.
The wage-market comparison across leagues
Mapping his Premier League wage percentile (99th among PL attackers on the 2026/27 contracts) onto the other leagues’ current wage distributions gives a first, deliberately narrow, view of the outside market — his outside option, meaning what he could plausibly earn elsewhere if no agreement is reached, which is what gives any wage position its floor. This is a wage-market comparison, not a full outside-option analysis:
Table 4. Wage at Havertz’s Premier League percentile, by league (2026/27).
At the same percentile, La Liga’s top end (€20.9M) prices above the Premier League (€17.2M) and Ligue 1 is broadly comparable (€16.7M); Serie A (€11.3M) and the Bundesliga (€9.3M) pay materially less. The outside market at his level is a handful of contracts at two or three clubs, not a market.
Statistically similar profiles
A sharper comparison asks which players actually resemble Havertz statistically. We put the selected metrics onto a common scale and compare each player’s overall statistical fingerprint with Havertz’s pooled Arsenal profile. The closer the two fingerprints, the more similar the players are in the dimensions used in this report. We run the exercise twice — once giving more weight to striker characteristics and once to creator characteristics — because ‘similar’ depends on what part of his role matters for the decision. As a check, we also reran the search without imposing those weights. Half of the top ten names overlap and the wage picture remains similarly modest, so the conclusion does not depend on one particular weighting choice. The technical details are in the annex.
Table 5. Most similar profiles, striker-facing weights.
Table 6. Most similar profiles, creator-facing weights (top rows).
Two readings. The club’s: statistically similar output profiles earn a median of €3.8M on 2026/27 estimates, and even the most relevant comparators — Ekitike (€12.2M) and Jonathan David (€3.7M on his current Juventus estimate) — earn €4.8M to €13.3M less than Havertz. The player’s: the similarity metric matches output shape, not league, age profile, status or big-club experience — the list spans a 37-year-old at Marseille and a 23-year-old at Hoffenheim — and his own 2025/26 rate profile is substantially stronger than the pooled-era vector used for matching. The tables bound the argument; they do not settle it. What they establish more firmly is a Section 5 point: profiles resembling his are rare at the top of the market — no similar player combines his output level with his aerial volume at a top-six-calibre club.
Potential outside market
A full outside-option analysis — which clubs need this profile, could afford the fee and wage, and play a compatible style — requires club-level need modelling beyond this report’s scope. What the present data does support: on wage capacity, only a small set of clubs outside England (three to five, concentrated in Spain and Germany’s top end) carry wage structures where a €12–17M attacker wage would fit; on style, Arsenal’s usage of aerial attacking play is among the highest of those clubs. Both facts narrow, rather than widen, his realistic outside market — a wage-relevant consideration for both sides. This is an assessment of market capacity, not of any club’s actual interest.
Negotiation interpretation – Compensation Benchmarking
Evidence available to the player’s side:
The wage premium is meaningful but not unusual, and conditional on usage it disappears. At his actual 2025/26 usage, his wage sits about 24% above the statistical benchmark. That premium matters, but it is not exceptional among highly paid players: roughly a quarter of the players in the sample are paid an even larger premium over what the model predicts for them. Under restored starter-level usage at his current rate, the gap almost disappears (€16.7M benchmark against a €17M 26/27 wage). The technical version of this comparison — residuals and their distribution — is in the annex.
His current rate supports his current pay. On 2025/26 output per 90 he sits among players earning comparable wages on the market map.
The timing suits a renewal. Attacker wages peak at ages 27–30, the window of the next contract; his decade-long record of 90th-percentile-of-age-group earnings reflects the sustained judgment of three clubs.
The similarity tables understate him. They match his pooled profile, which includes the injury seasons, and ignore league, status and the 2025/26 rate.
Potential club counterarguments:
The highest wage cost per minute in the squad. €26,000 per league minute in 2025/26, roughly five times the club’s most-used players — an observed fact not subject to sampling objections.
The €13.5M benchmark reflects what was actually delivered. The premium priced a durable starter; two availability failures are grounds to reprice it, and the restored-usage benchmark is conditional on a durability the completed record does not yet show.
Statistically similar profiles earn €4–12M. Even granting the caveats, no comparator in the tables earns within €4M of his current wage.
The outside market has narrowed but not closed. On 2026/27 contracts, La Liga’s top end prices his percentile above his wage (€20.9M) and Ligue 1 about level (€16.7M); Bundesliga and Serie A price it far lower (€9–11M). At the 99th percentile these are a handful of contracts at three or four clubs, not a market — the wage-capacity screen still identifies few realistic alternative payers.
Hardest point for the player to defend: every external benchmark in this section — the model at observed inputs, the similar-profile wages, the cross-league mapping — sits below his current wage; his case relies on internal benchmarks (his own rate, his own history) plus restored usage. The restored-usage assumption is no longer hypothetical — he is starting every week — but the wage-relevant version of the claim is a season of it, and that remains unproven.
Hardest point for the club to defend: its own €17M valuation is only three years old, was set on the same player, and the benchmark reproduces it the moment starter usage is assumed — so the club’s case reduces to the availability question, on which the sample is two events.
Taken together, Sections 2 and 3 point to a simpler disagreement than the wage tables initially suggest. The central question is not whether Havertz has the level to earn near his current wage; the model can reproduce that salary when starter-level usage is assumed. The disagreement is whether Arsenal should guarantee that level of pay before the availability question has been resolved. Section 4 asks how the same uncertainty affects his transfer value, and Section 6 shows how contract structure can share the risk.
The summer window changed the wage arithmetic itself. The departures of Martinelli, Trossard and Jesus removed three senior attacking salaries and cut Arsenal’s estimated bill from about €219M to €179M, with only Tzolis (€5.5M) arriving in their place; Saka’s renewal made him the squad’s top earner. The statistical wage benchmark and every wage-performance pairing remain estimated on the completed 2025/26 season, the last for which wages can be matched to a full season of output. The freed headroom weakens any affordability-based counterargument to the ranges assembled here.
The compensation range
Assembling the anchors: the statistical benchmark spans €13.5M (observed 2025/26 usage) to €16.7M (restored usage at his current rate — the state his 2026/27 usage so far matches); statistically similar profiles earn €1.9–5.2M on 2026/27 estimates (the interquartile range — the middle half of the group), with the most relevant comparators at €3.7–12.2M; Premier League centre-forwards excluding him earn €7.6–9.4M at the 75th–90th percentile; his current wage is €17.0M. No single anchor is correct. Together they suggest that guaranteed compensation between roughly €13M and €17M can be defended from the evidence, with the position inside that range determined by the availability assumption — which is precisely what contract structure, rather than a fixed number, can price (Section 6).
Figure 30. Compensation anchors: five reference points for guaranteed annual compensation.
4. Market Value & Contract Position
Market value brings many of the earlier questions together: performance, age, availability and market demand. But it is important to separate two ideas. The player’s underlying market value can remain stable or even rise; the transfer fee Arsenal can actually collect can still fall simply because the contract is running down. This section models those two quantities separately.
Value history and projections
Havertz’s recorded market value has fallen by more than a third across his two transfers — €81M when Chelsea signed him at 21, €55M when Arsenal did at 24 — and the latest Transfermarkt reading (June 2026) stands at €55M — the anchor used for the projections. The recorded anchors are Transfermarkt valuations at the two transfers plus that June 2026 update. For the projection, the age component is deliberately modest: among comparable established attackers, recorded values drift by roughly −2% per year between ages 27 and 31. The large movements therefore come from the availability scenarios rather than from age itself.
Figure 31. Market value 2019–2026 — the full Transfermarkt valuation history — and three projection paths to age 31.
The light-grey background line is his complete Transfermarkt valuation record, fetched from the source: the €90M Leverkusen peak in 2019, the long €70M Chelsea plateau, a 2024 recovery to €75M at Arsenal, and the injury discount down to €50M by March 2026 — with the June 2026 update ticking back up to €55M. The projections anchor at €55M, the June 2026 reading; the three scenario paths then separate what age alone would do from what availability decides:
Base — the age curve only. Nothing changes: his 2025/26 pattern neither repeats nor fully reverses, and value simply drifts with age from ~€55M to ~€53M by 2028. Implication for the club: his value roughly holds, but — as the next sub-section shows — the fee the club can charge for him does not. Waiting costs the club money even in the neutral scenario, which is a mild incentive to either extend or sell rather than defer.
Bull — two healthy starter seasons. Availability is restored and the market re-prices him as a proven starter again: modelled as a +25% re-rating after the first healthy season and +5% after the second, applied on top of the age drift. Value rises toward €69–72M. Implication: this is the path on which holding the contract pays — had Arsenal sold in the now-closed summer-2026 window at ~€55M, it would have locked in today’s price before any of this upside; under this path the upside accrues to whoever holds the registration.
Bear — the availability pattern repeats. Another materially interrupted season: modelled as a −35% re-rating, roughly the discount markets apply to attackers of his age with repeated absences; value roughly halves toward €30–36M. Implication: this was the scenario that argued for selling before the window closed. With that exit gone, its live implication is contractual: shifting part of the wage into availability-linked terms, so the club is not paying a full wage through a second lost season.
By 2028 the difference between the healthy-starter and repeated-absence scenarios is around €42M. Put simply, availability is doing almost all of the work in the forecast. The age effect is modest; the major question is whether the market again sees Havertz as a reliable starter. The scenario multipliers are assumptions rather than facts, so the important message is the sensitivity of value to that question, not the exact euro at either endpoint.
The contract clock
The remaining contract changes what Arsenal can realistically charge. A player with two years left gives the selling club much more bargaining power than the same player with one year left, even if his football level is unchanged. We estimate that effect from recent transfers by comparing fees with recorded market values at different points in the contract.
Figure 32. The resale clock from September 2026: the windows still available, with the closed summer-2026 window as a counterfactual benchmark.
Applied to the projections, from where the parties actually stand in September 2026: the near-full-value window has already closed — a summer-2026 sale would have realised an estimated €51–55M under every performance scenario, and that number now serves only as the benchmark of what unresolved waiting has cost. The windows that remain are worth less and depend on the season: January 2027 (about 1.5 years left) is estimated at €39–54M across the bear-to-bull paths — a clock-only estimate that ignores mid-season effects; summer 2027 (one year left) at €25–49M — the CIES one-year ratio (×0.71) applied to the diverged values; and at expiry in June 2028 he leaves without a fee. The player’s market value need not fall at all over this horizon — under the bull path it rises — while the club’s potential fee falls with every window that passes. That wedge, between value the player retains and fee the club can capture, migrates to the player’s side of the table as expiry approaches, in the form of signing bonuses and wages from a future employer. It is the report’s most time-sensitive quantity for both parties: the same clock that erodes the club’s fee erodes the player’s contractual security.
Negotiation interpretation – Market Value & Contract Position
Evidence available to the club’s side: each further year of deferral still costs an estimated €9–17M of potential fee under the base path; January 2027 is now the highest-value exit that exists, and it exists only briefly. Evidence available to the player’s side: the highest-value window closed without a sale while he started every opening match — facts that sit more comfortably with continuation than with an exit strategy; the value projections reward restored availability and starter usage; and at expiry the entire remaining value accrues to the player, which disciplines how far the club can press the repricing argument.
Hardest point for the player to defend: the bear path is not symmetric — a second lost season roughly halves his value and coincides with the club’s last meaningful sale window, leaving him negotiating from an expiring contract at a reduced valuation.
Hardest point for the club to defend: it has already declined its best sale window. Having kept him through summer 2026 while starting him every week, arguing a deep repricing from availability risk means asking the player to accept a discount the club’s own recent decisions did not apply.
5. Replacement Economics & Club-Specific Value
The final analytical section asks the question a club would ultimately care about: if Havertz leaves, what does Arsenal actually lose — and what would it cost to replace it? We look at the answer in sporting results, financial cost and the specific profile he provides.
In points: the Premier League title margins
The first analysis converts his output into league points. To do so we need an exchange rate between goal contributions and points, and we estimate it from the league tables themselves: across 780 top-five-league team-seasons (every club in the big five leagues, 2017/18–2024/25), one additional goal of goal difference is associated with 0.632 additional points in the final table — a simple regression of season points on season goal difference, which explains about 91% of the variation. These are team-seasons, not players: the rate describes what one more goal scored, or one fewer conceded, has historically been worth to a team’s total; by approximation we apply it to expected contributions (xG+xAG) as well.
On that basis, the value of restoring him is modelled at about 2 expected points: a healthy Havertz playing ~2,800 league minutes at his Arsenal-era rate, against the 647 he actually played, adds 2,153 extra minutes (2,800 − 647) ÷ 90, times the rate gap between him and the depth players who would otherwise play those minutes — his 0.53 xG+xAG per 90 (2023/24 through 2025/26) against the 0.40 per 90 that Arsenal’s non-starter attacking minutes actually produced (the “depth rate”) — times 0.632.
Selling him without replacement, from a 2025/26 baseline of 647 minutes, costs under one expected point — that is all a sale at last season’s usage would actually have forfeited, because his remaining potential minutes were already being covered by squad depth. The two calculations partition one sum: the ~2 points prices restoring him from 2025/26 usage to starter usage, the ~0.6 prices the baseline minutes a sale would remove, and together a healthy starter Havertz is worth about 2.6 expected points over depth-level output. Which figure applies depends on the counterfactual: selling the player as used in 2025/26, or keeping the player the fitness case says he can again be.
What two points buy is best read from history: the title has been decided by two points or fewer in three of the last nine seasons (Arsenal’s own 2025/26 margin was seven), so a contribution of his modelled size has been the difference between champion and runner-up in a third of recent seasons. Sensitivity: varying his assumed rate (0.45–0.58) and the depth rate (0.36–0.45) moves the modelled contribution from about zero to 3.3 points across the grid.
Figure 33. What two points buy: Premier League title margins 2017/18–2025/26, against his modelled contribution.
The rival effect: losing the player versus strengthening a competitor
The second analysis asks how often a contribution of this size decides a title — and what changes with the destination of a sale. The tool is a deliberately simple two-team race. Arsenal start each simulated season from their reconstructed 2025/26 total (85 points) and Manchester City — the nearest challenger — from theirs (78), both taken from the 2025/26 final standings; both totals then receive independent random noise of about ±4.4 points — the typical amount by which a team’s actual points deviate from what its goals would predict, capturing finishing luck, timing and fine margins. This is repeated 20,000 times, and the title probability is the share of simulated seasons in which Arsenal finish ahead. Keeping a healthy Havertz shifts Arsenal’s starting total by +2 and wins about 92% of the simulated races; the status quo wins about 87%. Three simplifications make those levels generous — only one challenger can win, both teams start from last season’s totals as if squads never change, and nothing else varies — which is why the report treats the differences between scenarios, and the historical title margins of Figure 33, as the message rather than the 92% itself. The annex sets out the same construction in full.
A sale has one sporting price if the player leaves the league and a different one if he strengthens the direct challenger, because the head-to-head points gap moves twice — Arsenal loses his contribution and the rival gains it. In expected points, an outside sale moves the gap by about 0.6 (small because, from the 2025/26 baseline, a sale removes only the 647 minutes he actually played — his remaining potential minutes were already being covered by squad depth); a rival sale moves it by 1.9 to 3.1 points, depending on whether 50%, 75% or 100% of his Arsenal output transfers to the new system — a swing larger than the title margin in three of the last nine seasons. Read through the illustrative two-team simulation, the same swing is title probability falling from 92% (keep, healthy) to 85% (sell abroad) and to 73–79% (sell to the rival); the levels are illustrative, the differences between them are the message.
Figure 34. The rival effect: simulated 2025/26 title probability when the same sale goes to different destinations.
The rival here is generic, calibrated to the nearest challenger’s baseline points; no actual buying club is implied, and no transfer interest from any club is evidenced. Output also does not transfer one-for-one across tactical systems, which is why the transfer-efficiency sensitivity is shown rather than a single number. The levels of this two-team simulation are illustrative; the differences between bars carry the message.
Negotiation implication. For the club, the destination of any sale is a price term: a domestic-rival sale carries a sporting cost the fee does not compensate, which argues for contractual control over destination (or a materially higher price for it). For the player’s side, the same logic is leverage — the plausible existence of a domestic buyer raises the cost of the club’s walk-away option.
From sporting impact to financial impact
The estimates above are sporting quantities — points and title probability; this step converts them into money, by asking what each sale scenario’s shift in title probability is worth in expected competition payments under three assumptions about the value of a title. Table 7 reports the conversion, scenario by scenario.
Table 7. Sporting scenarios translated into expected competition payments (EUR M, vs keeping a healthy starter).
Two honest findings, and the second matters more than the first. First, the channel: Arsenal’s Champions League qualification probability is essentially unchanged in every scenario — the club is far above the qualification line in all of them — so the financial value of the expected-points swing runs almost entirely through title probability, not through qualification. Second, the magnitude: priced only through observable competition payments, the expected financial impact is small. The low case uses the Premier League merit ladder alone (≈ €3.6M separates first from second under the 2024/25 framework); the base adds the associated broadcast-appearance uplift; the high case adds a €10M allowance for title-linked bonuses that are common in sponsor and kit contracts but not publicly observable — an assumption, labelled as such, not an estimate.
Even in the worst case the expected prize-money cost of a rival sale is under €3M — an order of magnitude below the fee and wage numbers this report trades in. The conclusion is not that the sporting swing is worthless; it is that its value to the club is mostly strategic (title odds, squad standing, the cost of re-acquiring contention) rather than bankable, and the negotiation should treat it that way rather than dressing it up as revenue.
In money
Replacing the full profile is expensive for the reason Sections 2 and 3 established: the profile has no close market neighbour. Buying the closest market equivalent (Šeško moved for ~€76.5M in 2025; an Isak-tier profile is priced at €120M+) is estimated at ~€39M a year all-in — more than double his current €17M. A mid-tier replacement (~€45M fee, €8M wage) costs about €17M a year — approximately the same annual outlay as keeping him, for part of the profile. The internal option — Gyökeres as the sole senior No. 9, with Merino as emergency cover — is cash-free but, after the summer departures, leaves centre-forward cover one injury deep and concentrates exactly the risk the club is pricing in him.
Figure 35. The replacement menu: annual cost of each option.
Actual Arsenal replacement paths
The replacement menu above priced generic options; this table prices Arsenal’s concrete courses of action from September 2026 — keep him, or sell in 2027 and cover internally, mid-market or like-for-like — on one line each: transfer cash, annual squad cost, expected output and expected points. The summer-2026 internal-redistribution path that earlier drafts priced no longer exists: Martinelli, Trossard and Jesus, who carried 1,600 of the 2,800 minutes it redistributed, were sold in the summer window, and only Tzolis, a wide player, arrived. It is retained as a clearly labelled counterfactual row — the road not taken — and nothing in the live menu builds on departed players.
Table 8. Four Arsenal-specific 2026/27 paths (EUR M; points vs keeping him at restored usage).
The minute allocation behind the live internal path (B) uses only the current squad and sums to the 2,800 league minutes he would otherwise play: 1,900 to Gyökeres as the full-time No. 9 (pooled rate 0.49 xG+xAG per 90), 500 to Merino as emergency centre-forward cover (0.48), and 400 to Madueke and depth options (0.30 and the squad depth rate 0.40). Expected contributions use each player’s own pooled 2023/24–2025/26 Premier League rate against Havertz’s 0.53; annual cost uses fees amortised over four years; sale fees for the 2027 paths are the state-averaged clock values of Section 4. No freed minute is assumed to go to one player, and no incoming player is assumed beyond the stated fee and wage band. Full calculation detail for every path is in the annex.
The table’s shape is the finding, and the summer window sharpened it. The only path that unambiguously saved money — selling in 2026 with a full internal bench behind him — is the counterfactual row: it required both the closed window and the departed players. Of the live options, internal cover after a 2027 sale (B) banks a diminished, state-dependent fee and hands the role to a single No. 9 with thin cover behind him; the mid-market path (C) is roughly cost-neutral annually and still surrenders about a point; only the similar-profile path (D) holds the sporting line, at €7.6M a year more than his current contract plus a negative fee flow. A club that wants the level held cannot get there cheaper than his current deal — and every sale path now passes through the weaker 2027 fee windows.
In profile
What would depart is a combination rather than a quantity: a player who clears both the centre-forward and the creator bar (Figures 12–13), wins aerials at close to a target-man’s rate, and has played three positions at this level. The cost of losing such a player is concentrated in the moments squads cannot plan for — injury pile-ups, tactical Plan B situations late in a season — rather than in the August depth chart.
Negotiation interpretation – Replacement Economics & Club-Specific Value
Evidence available to the player’s side:
In the scenarios modelled here, external replacement costs at least as much as retention. The mid-market path costs roughly the same annually while losing about a point; the closest-profile path holds the sporting line but costs more. The conclusion is conditional on the stated fee, wage and output assumptions.
The value projections reward restored availability and starter usage. Two healthy starter seasons are modelled to add roughly €17M of asset value relative to today’s €55M base, while another materially interrupted season moves the value sharply the other way.
The expiry endgame limits the club’s leverage. At zero fee, the player captures the asset value himself — a well-documented pattern for players in comparable positions.
Potential club counterarguments:
The recovered-fee argument is weakened but not gone: the last near-full-value exit closed in summer 2026, but January 2027 remains an estimated €39–54M sale if the club concludes the availability risk is unacceptable — and every window after it is worth less.
Measured at his 2025/26 usage, his departure would have been near sporting-neutral — under one expected point from that season’s baseline, a season Arsenal won the league with limited contribution from him (647 league minutes). That argument weakens with every week of restored starter usage in 2026/27.
The internal option is already paid for. The recovered fee funds needs elsewhere in the squad.
If extending, the projections argue for contingent structure: the club should pay for the bull path only when it materialises — through minutes- and performance-linked compensation — rather than guarantee it in advance.
Hardest point for the player to defend: the club can reach sporting adequacy at centre-forward without him at zero cash cost; his irreplaceability argument rests on profile breadth and Plan-B value, which is real but hard to price.
Hardest point for the club to defend: the replacement arithmetic — every external route to comparable output costs the same or more than the wage the club would be seeking to reduce.
6. Negotiation Framework & Contract Options
This section synthesises Sections 1–5 into a practical negotiation framework, built around the options actually available from September 2026: the summer 2026 window is closed, so “sell now at near-full value” is no longer on the menu. It sets out what the evidence means for timing, the live extend/exit/run-down decision, the negotiating range and the structure of a potential agreement.
What the evidence strongly supports
His underlying per-90 performance level is intact and at or near a career high (Sections 2.1, 2.5) — and the finding holds in every peer pool tested and in the pooled three-season record, not just in the small 2025/26 sample.
His 2025/26 usage was an extreme outlier against his own ten-season record, availability-driven rather than form-driven — and he has now played eight months without a new absence (Section 1).
The current evidence on role is unusually clear: he has started every competitive 2026/27 match at centre-forward, with Gyökeres used from the bench (Section 1) — team-sheet facts, though only an opening month of them.
His profile (output + aerials + creation-for-a-No.9) is rare among Premier League attackers, and external replacement is estimated to cost at least as much per year as his current wage (Sections 2.4, 5).
His wage tops every relevant pool while his 2025/26 minutes sat near the bottom of each — and the benchmark gap closes almost entirely under restored usage (Section 3).
The fee Arsenal can realise decays with the contract clock, and the first step of that decay has already happened: near-full value in the closed 2026 window (counterfactual ~€51–55M), roughly €39–54M at January 2027, €25–49M in summer 2027 depending on the season path, zero at expiry (Section 4).
What the evidence does not establish
Whether his availability is durably restored: eight strong months against two recent extended absences supports neither “resolved” nor “chronic”; only a completed season at starter load can.
That current usage will survive a full season: the team sheets settle the hierarchy question for now, but three league rounds cannot show durability at starter load — which is precisely the input the wage and value models turn on.
That his output volume returns with his minutes: the elite 2025/26 rates came from short appearances, and the early 2026/27 sample pairs full minutes with thin attacking volume (11th percentile xG+xAG per 90 through three rounds) — consistent with a team early in its rhythm, but unproven either way.
Whether any club actually wants to sign him. The outside-market analysis in Section 3 identifies clubs that could afford a player on his wage — it says nothing about whether any of them wants him. Actual interest cannot be observed in this data, so the outside option is a boundary, not a fact.
A single “correct” wage: the defensible anchors span roughly €13–17M guaranteed.
Timing: what changes if nothing is agreed
Waiting is not neutral, and its first instalment has been paid: the near-full-value 2026 window closed unresolved. Each further window moves the contract closer to expiry and shrinks the realisable fee (Section 4); for Havertz, another interrupted season would weaken both his renewal case and his outside market, while a sustained season strengthens them. A ‘prove-it’ half-season to January is therefore a deliberate risk-taking strategy for both sides, not the absence of a decision.
The economics of the decision, in brief
Sections 1 to 5 each priced one piece of the situation; put on a single economic basis to June 2028 — wages paid, fees received, replacement costs and the value of the contract still held — they point one way. Selling in summer 2027 is worth less than continuing under any view of his fitness: the fee ladder of Section 4 shows why, since by then the achievable fee in a bad scenario (≈€25M) is below what the club would still hold by simply keeping him under contract, while in a good scenario the comparison is not close. A January 2027 sale pays only if the club privately believes his availability is very likely to fail again — a belief its own team sheets contradict. And running the contract down is the worst position of all: two seasons of top wages for a player who then leaves free.
That ranking is robust to the assumptions that matter: moving the fee and asset estimates a fifth in either direction changes the numbers, not the order. The one genuine decision the arithmetic leaves open is not whether to continue but on what terms — the guarantee, the availability-linked share, and the length. Those are the subject of the rest of this section.
Illustrative contract structures
The recurring disagreement in this report is not primarily about how good Havertz is when he plays. It is about how often Arsenal can expect him to be available. A contract can price that uncertainty directly instead of forcing either side to win the argument in advance.
A — High guarantee (~€17M fixed): the club bears availability risk; defensible if the club accepts the restoration case.
B — Risk-sharing (~€11M guaranteed + up to ~€7.5M by minutes/performance): expected compensation ranges from ~€13.2M at 25% of minutes to ~€18.5M at 90% — the player exceeds current compensation when available and productive; the club’s cost tracks what it receives. This structure prices the availability question directly instead of resolving it by assertion.
C — Short extension at a higher rate (~€18M × 2 years): the player trades security for rate; the club limits duration risk and preserves a 2028 decision point.
D — Longer security at a lower rate (~€13.5M × 4 years): the player trades rate for duration; the club lowers the annual guarantee and extends fee-relevant contract cover.
Figure 36. What a risk-sharing contract would pay at each availability outcome, next to a fixed wage.
The figure in a nutshell: A risk-sharing deal pays his current wage or better when he plays, and materially less when he does not — the contract prices the dispute.
Table 9. Annual pay under each structure, by the share of league minutes he ends up playing (EUR M).
The table turns a disagreement about fitness into a financial choice. If the player is confident he will play most of the season, a structure with meaningful availability upside can pay him more than a fixed deal. If the club is worried about repeat absences, the same structure limits what it pays in a low-availability season. Structure B therefore appears attractive to both sides for different reasons: neither side has to concede its view of the medical risk before signing.
Stress-testing these structures across many possible availability outcomes adds magnitudes but not direction: a structure whose pay moves with minutes costs the club about the same as the fixed wage in expectation — the two sides are arguing over well under €1M a year in expected money — while removing most of the scenario in which the club pays top wages through a lost season, and still paying the player his current wage or better in roughly two of three fit-enough seasons.
A disclaimer applies to every wage figure in this report: Capology numbers are public estimates of base salary. Real contracts typically carry appearance, goal and other performance bonuses that are not publicly observable, so the structures and simulations here price the guaranteed-versus-linked question on estimated base wages only — the true split between fixed and variable pay in his current deal is unknown, and a commissioned report would work from the actual contract.
Potential negotiating zone
The analysis does not produce one objectively correct salary. It produces a negotiating range and shows what moves the player within it. The evidence can support guaranteed compensation of roughly €13M–€17M; full-availability compensation can reach or exceed the current €17.0M through variable components; and contract length can trade annual pay against security. The unresolved availability risk is therefore better handled through structure than hidden inside a single fixed wage — and with the sale options fading, structure is also where the club’s remaining leverage actually lives.
Where this leaves the two parties
Strip the machinery away and the position is unusually symmetrical: both sides have more to gain from an early agreement than from testing each other for another season. The player is 27. A contract signed now covers his peak years and is, on any realistic reading of the attacker wage curve, likely the largest of his career — top-decile attacker wages hold into the early thirties only for those who keep starting, and almost nobody re-signs at his level after 30. His hand is as strong as it has been since 2024: career-high output, restored starter usage, a World Cup and a Champions League final on the record — and yet his outside market is thin, because only a handful of clubs in Europe pay centre-forwards what he already earns, and the players statistically closest to him earn a fraction of it. Waiting exposes him to the one thing that would collapse his position: another absence on an unimproved deal.
Arsenal’s interest points the same way. The profile is genuinely difficult to buy — every external route to comparable output costs the same or more, and the closest statistical equivalents play for clubs that do not sell cheaply — while the squad’s own alternatives at centre-forward thinned to one senior option in the summer window. And the club’s remaining leverage decays mechanically: each window that passes converts transfer value into the player’s future signing terms. A club that intends to keep its centre-forward, and is starting him every week, maximises nothing by negotiating against its own shrinking clock. The evidence therefore points both parties toward agreeing terms this season, with the structure — a guarantee near the benchmark range, upside that restores current pay when he plays — carrying the one question the record cannot yet answer.
What evidence would change this assessment
A completed healthy season at starter minutes (the opening months point that way; a full season would validate the restoration case, move the wage benchmark to ~€16.7M and the value path toward the bull scenario).
A third extended absence (validates the risk case; bear path — and with the 2026 window gone, the club’s remaining exit values are the diminished January and summer 2027 fees).
A material change in role (sustained return to attacking midfield would shift the relevant peer groups, benchmarks and comparables).
Credible external interest at a full-value fee (would re-price the outside option this report can only bound).
Updated market-value or wage comparables as new windows close.
Base-case outlook: what is most likely from here?
The analysis above prices several possible outcomes; it does not predict Arsenal’s or Havertz’s actual intentions. But taken together, the evidence points to one path as more plausible than the others.
The base case is a new contract during 2026/27 rather than a sale. Arsenal allowed the last near-full-value sale window to close, have started Havertz at centre-forward in every competitive match of the new season, and now have fewer internal attacking alternatives after the summer departures. Meanwhile, waiting until summer 2027 would leave the club negotiating with only one year remaining on his contract and a substantially weaker transfer position. None of those facts guarantees a renewal, but together they make continuation more consistent with the observable evidence than an exit.
The more difficult question is the price. Havertz already earns an estimated €17.0M a year, second only to Saka’s €18.3M at Arsenal. The statistical benchmark reproduces approximately that wage once starter usage is restored, while his current role, age and shrinking contract length strengthen his bargaining position. At the same time, two recent extended absences make a large increase in fully guaranteed compensation difficult for the club to justify.
The most plausible compromise is therefore headline compensation around the top of Arsenal’s existing wage structure, but with some of the upside contingent on availability and performance. A deal that keeps guaranteed compensation close to Havertz’s current level while allowing total annual pay to reach roughly €18–19M when he plays regularly would place a healthy season around Saka-level compensation without requiring Arsenal to guarantee that level through another disrupted one. That is also close to the risk-sharing structure modelled above, which pays €17.6M at 75% availability and €18.5M at 90%.
Timing matters almost as much as price. If Havertz remains available and retains his starting role through the first half of 2026/27, his case strengthens while Arsenal’s leverage weakens: the club would approach summer 2027 with only one year left on the contract. On the evidence available today, that makes an agreement during the current season more plausible than allowing the decision to drift into the final year.
The principal reason this forecast could change is the same one that runs through the rest of the report: availability. Another extended absence would reopen the club’s case for repricing or exit. A sustained season at starter load would push the evidence further toward a renewal at the upper end of the negotiating range.
Thank you for reading until the end. ❤️
You can grab a PDF version of the report below (in which you’ll also find the technical annex).
About the Author
Martin Angelov is a competition economist with a decade of experience in data analytics and consulting. His work combines applied economics, quantitative analysis and evidence-based advisory.
About DataDealsFC.com
Data Deals FC provides data-driven player contract assessments designed to support contract negotiation processes. The reports combine performance analysis, peer and wage benchmarking, market-value assessment, replacement economics and scenario analysis, tailored to the objectives of the player, representative or club commissioning the work.
For enquiries about tailor-made reports: martin.angelov@datadealsfc.com & DataDealsFC.com
