Every few seasons, a goalie has a career year right before their contract is about to expire. The league takes notice. A general manager, one who is staring down a playoff window and a fanbase desperate for answers, hands over seven or eight figures a year for the next four to six seasons. It feels like the responsible move, locking in that elite-level goalie who is peaking at the right time, and hoping it leads to a Stanley Cup. However, the data says it almost never is.
This article is about why that pattern keeps repeating, why the economics of goaltender contracts are structurally broken across the league, and what a data-based approach may look like when it comes to building the position.
Everything here comes from season-by-season goalie performance records covering the 2016-17 through 2025-26 seasons. This includes 658 qualified seasons across 230 unique goalies. All performance data is sourced from MoneyPuck and NHL.com. All contract data is sourced from PuckPedia. All data is current as of collection on 3/30/2026.
A note before diving in: contract negotiations are messier than any dataset can capture. Goalies and agents push for term and job security. Teams operate under internal structures that we may never know about. A goalie might take less money to stay in a winning situation, or demand more simply because the market lets them. None of that is irrational, it’s just not what this article is about. The goal here is to isolate what the performance data actually says about goaltender value, independent of negotiating tactics, and use that as a baseline for data-based decision-making.
The short version, before we get into the data:
- Never sign based on one peak season.
- Long deals don’t buy you more production; they buy you less flexibility.
- Free agent signings are structurally disadvantaged by team-switch regression.
- Volatility is a real risk factor that should shorten contract length, not just lower AAV.
- Invest in your backup.
The rest of this article is the evidence behind each of these. The data is worth reading.
How the Numbers Work
The core metric here is Goals Saved Above Expected (GSAx): the difference between the number of goals a goalie is expected to allow based on every shot they have faced, and the number that they actually allowed. This model accounts for: location, type, distance, angle, traffic, and game state. Having a positive GSAx means that the goalie performed better than expected (or a replacement-level goalie); a negative GSAx means they performed worse than expected. We will also use GSAx/60, which is how many goals the goalie saves above expected in a standard 60-minute NHL game.
To convert GSAx into something that actually matters to team success, and something we can attach a dollar value to, I use Wins Above Replacement (WAR):
WAR = ( GSAx - Replacement GSAx ) / 6
Where Replacement GSAx = (-0.10 GSAx/60) x ice time in minutes / 60
For this study, replacement level is set at -0.10 GSAx/60, which places it roughly around the bottom 15% of qualifying starting goalies, and the level you can expect a team to be able to access freely from an available backup or AHL callup without spending meaningful cap space. It should be considered the baseline performance before you invest a single extra dollar.
We will use a 6 GSAx = 1 WAR conversion, which is the standard NHL approximation used across skater WAR models. It is seen in public works from Evolving Hockey, whose WAR model is one of the most widely-cited and trusted models in the analytics community.
We will also be using a market rate of ~$3.3M per WAR (around 3.46% of the salary cap). This was derived from the data itself: for every goalie with a known cap hit in 2025-26 and at least two qualifying seasons since 2022, I computed their cap hit divided by their average WAR per year, then took the median across those 50 goalies. I used the median over the mean because there are some largely overpaid contracts (Markstrom at ~$82/WAR, Adin Hill at ~$32M/WAR) that would greatly skew the mean above what a productive goalie costs on the open market.
All analysis is restricted to even-strength situations only, with a minimum threshold of 600 5-on-5 minutes per season to qualify. That filters out short auditions and spot starts without cutting out legitimate backup seasons.
Part One: The Problem
The Collapse Is Instant
The most important thing in the data is also the simplest. Take every goalie who posted a peak season at or above 0.4 GSAx/60, an elite-level season that attracts high-value contract offers. Here is what happened to those goalies in the years that followed, across 50 such cases since 2016:
| Year | Avg GSAx/60 | % Above Average (> 0.1 GSAx/60) | % Below Replacement (< -0.1 GSAx/60) |
|---|---|---|---|
| Peak year | +0.550 | 100% | 0% |
| Year 1 | -0.011 | 34% | 50% |
| Year 2 | -0.013 | 35% | 43% |
| Year 3 | -0.096 | 25% | 64% |
| Year 4 | -0.135 | 20% | 73% |
The average drop from a peak season to Year 1 is 0.561 GSAx/60. This means on average, these goalies go from having a very elite-level season to hardly being above replacement level. It happens immediately, in the first season of what is usually a multi-year contract, where they are paid to replicate that elite performance.
Half of these goalies fall below replacement level (< -0.1 GSAx/60) in Year 1 alone. Two-thirds are below average (< 0.1 GSAx/60). Only 8% of them sustained elite-level performance (>0.3 GSAx/60) over the post-peak window.
Juuse Saros is the clearest current example. He peaked at +0.738 GSAx/60 in 2022-23, one of the best single seasons in this dataset. He then dropped to -0.06 GSAx/60 the following season, after which Nashville signed him to a massive extension (8 years, $61.92M). He has since posted -0.290 and -0.284 in the two seasons since signing.
Jack Campbell peaked at +0.39 GSAx/60 in 2020-21, and even after dropping to -0.05 the following year, that one year was enough to earn him a 5-year, $25M contract in Edmonton. The following two seasons, Campbell fell even further to -0.54 and then -0.8, essentially playing himself out of the league and forcing the Oilers to buy him out. Ville Husso, Joonas Korpisalo, Alex Nedeljkovic, the pattern repeats often enough that it stops looking like bad luck and starts looking like a structural problem.
None of these decisions were obviously wrong at the time. Peak performance occurs, and locking in a goalie who just carried your team through a playoff run feels like the right move. But the data is consistent enough that it warrants a harder look at what that peak is actually predicting.
This could just be chalked up to a run of bad luck, but the consistency suggests there is something more structural at work. The most likely explanation is regression to the mean operating on a position with some of the highest single-season variance in all of sports. The average starting goalie produces +0.067 GSAx/60, with a standard deviation of 0.291. A peak season of +0.550 is nearly 1.7 standard deviations above average. Using a normal distribution, this only has a 4.46% chance of occurring. So, expecting the same performance, or better, is near impossible without a consistent body of work.
The Team-Switch Penalty
Most peak-driven goalie signings involve a team change. Many times, when a goalie has an elite season in a contract year, they play themselves off the team, demanding too much to fit within the team’s cap structure. This causes them to be moved, or simply hit free agency, where a new team will gladly overpay for the goalie to produce middling returns. The data tracks 18 cases where an elite goalie (0.40+ GSAx/60) changed teams.
Only 3 of 18 maintained above-average performance (0.10+ GSAx/60) on their new team. The other 15 averaged a -0.647 GSAx/60 drop from where they were previously. The correlation between pre and post-team switching performance is r = -0.574, meaning the better a goalie looked when switching, the larger the expected decline.
The worst examples:
| Goalie | To | Pre-signing | Post-signing | Drop |
|---|---|---|---|---|
| Eric Comrie | Buffalo | +0.829 | -0.380 | -1.209 |
| Scott Darling | Carolina | +0.604 | -0.543 | -1.147 |
| Joonas Korpisalo | Ottawa | +0.609 | -0.217 | -0.826 |
| Ville Husso | Detroit | +0.452 | -0.271 | -0.723 |
| Kevin Lankinen | Vancouver | +0.572 | -0.131 | -0.703 |
The only team change that really held up was Anthony Stolarz, who signed with Toronto in 2024. He dropped just 0.058 GSAx/60 from his historic peak in Florida (0.958 to 0.9). However, he has dropped down dramatically to -0.54 GSAx/60 during the 2025-26 season, right after signing a sizable extension in Toronto. This can be chalked up to an injury-riddled season, but it still supports the previous section of regression after an elite season.
Volatility Is a Risk Nobody Prices In
It should be well known that goaltenders are extremely volatile from season to season, but some are more consistent than others. Standard deviation in GSAx/60 is measurable, but there’s a wide spread.
Most consistent (3+ seasons, positive average):
| Goalie | Avg GSAx/60 | Year-to-Year Std |
|---|---|---|
| Joel Hofer | +0.288 | 0.092 |
| Joey Daccord | +0.245 | 0.151 |
| Corey Crawford | +0.157 | 0.152 |
| Igor Shesterkin | +0.333 | 0.156 |
| Thatcher Demko | +0.172 | 0.184 |
Most volatile (positive average, high variance):
| Goalie | Avg GSAx/60 | Year-to-Year Std |
|---|---|---|
| Anthony Stolarz | +0.155 | 0.623 |
| Eric Comrie | +0.232 | 0.495 |
| Laurent Brossoit | +0.212 | 0.411 |
| Filip Gustavsson | +0.150 | 0.398 |
| Jeremy Swayman | +0.399 | 0.354 |
| Logan Thompson | +0.313 | 0.351 |
Igor Shesterkin’s consistency is a genuinely underrated part of his value, and why he was able to command the highest AAV of any goalie in NHL history at $11.5M. His career average (+0.333 GSAx/60) is elite, but the standard deviation of only 0.156 means that you actually know what you’re getting year after year from him. Paying that much for a goalie is still a tough pill to swallow, but the floor is safer than it looks on paper.
The volatility data reframes how to think about goalies like Jeremy Swayman, who is a legitimate elite goaltender. Swayman is averaging +0.399 GSAx/60 over five seasons, which puts him in the top handful of goalies in the league. But his standard deviation of 0.354 (skewed greatly due to the 2024-25 season) means any given season can range from absolutely dominant to replacement-level. On an eight-year deal at $8.25M AAV, that range matters because getting a replacement-level season from him means the cap hit produces almost no return on investment. Shesterkin’s consistency makes it far more likely he provides value, even if it costs more.
Part Two: What to Do Instead
The Case for Internal Promotion
The market acts like promoting a backup to a starting role is a desperate move. Very rarely do you see a goalie go from playing 20-30 games immediately to over 40-50. However, the data suggests that there is a lot of untapped potential in capitalizing on backups who might command less money and term. Across 27 tracked promotions, cases where a goalie went from playing under 40% of team games (backup) to over 55% (starter), the outcomes were:
- 56% performed above average (>0.10 GSAx/60) in their first full starting season
- 26% were elite (>0.30 GSAx/60)
- Average GSAx/60 as backup: +0.036
- Average GSAx/60 after promotion: +0.138
The best promotion cases:
| Goalie | Age | Backup GSAx/60 | Starter GSAx/60 |
|---|---|---|---|
| Antti Raanta | 28 | +0.296 | +0.496 |
| Dan Vladar | 28 | +0.214 | +0.452 |
| Dustin Wolf | 24 | -0.709 | +0.425 |
| Ilya Sorokin | 26 | -0.108 | +0.400 |
Dustin Wolf’s promotion is particularly worth sitting with. His backup season looked terrible on paper, a -0.709 GSAx/60 is near the worst in the dataset. Calgary was always going to be patient with him, given his pedigree as a top goaltending prospect on a struggling Flames team. But the underlying case for doing so was there, regardless of context within the organization: his backup sample was small, low-leverage, and not representative of his talent. His starting season was elite, increasing his GSAx/60 by +1.134. The numbers backed up what Calgary believed, and they were able to get an elite season out of a goalie still on his rookie contract.
Joey Daccord’s promotion in Seattle follows the same logic. -0.161 GSAx/60 as a backup, +0.316 as a starter the next year. The market had priced him like a permanent backup, but Seattle was able to find their answer within the organization after Phillip Grubauer’s struggles.
However, promotion failures do exist. Carter Hutton went from +0.358 GSAx/60 as a backup to -0.276 in his first starting season.
Kevin Lankinen’s case is worth noting separately. His drop from +0.572 to -0.131 in Vancouver reflects both a team change and an unplanned promotion due to Thatcher Demko’s injury, making it one of the harder cases to pin on a single cause. Promotion doesn’t always work, but a 56% success rate on a $750K replacement competing against a 17% success rate on a $8M free agent signing is not a close comparison, even accounting for the fact that a promotable backup isn’t always available.
The Backup Quality Effect on Starters
There is a measurable relationship between how good your backup is and how your starter performs. Teams with elite backups (>0.2 GSAx/60) see their starter average +0.116 GSAx/60. Teams with poor backups (< -0.1 GSAx/60) see their starters average +0.045 GSAx/60.
That is a 0.071 GSAx/60 difference in starter performance based purely on who is playing behind them. Over a full season of roughly 3,000 even-strength minutes, that is about 3.55 goals saved (around half a win) attributable to having a quality backup rather than a warm body on the bench in a baseball cap. Not to mention that the backup themself is producing more WAR as well.
The logic makes sense. A starter who plays 68 games and never gets meaningful rest accumulates fatigue over the course of the season. A quality backup can provide load management for arguably the most important player on the roster and pick up meaningful wins in the process.
The implication for roster construction: a $3M investment in an elite backup is competing for real value with the dollars spent upgrading from a $6M starter to an $8M one. The market doesn’t price them equally, but it should.
The Market Right Now
Using average WAR per season from 2022-23 to 2025-26 against current cap hits, the goalie market looks like this.
Where teams are getting value:
| Goalie | Cap Hit | Avg WAR/yr | $/WAR |
|---|---|---|---|
| Jakub Dobes | $965K | 1.84 | $523K/WAR |
| Casey DeSmith | $1.0M | 1.96 | $510K/WAR |
| Dustin Wolf | $850K | 0.71 | $1.2M/WAR |
| Filip Gustavsson | $3.75M | 2.04 | $1.8M/WAR |
| Logan Thompson | $5.85M | 2.97 | $2.0M/WAR |
| Connor Hellebuyck | $8.5M | 4.39 | $1.9M/WAR |
Dobes and DeSmith sit at the top of the value table for completely different reasons. Dobes is still on his entry-level contract, delivering very strong numbers over the course of his first two seasons, making the very little money that Montreal is spending on him worth its weight in gold. DeSmith is a veteran backup/fringe starter who keeps posting legitimate starting-caliber numbers at only $1M. It appears Dobes will be paid a lot more appropriately once his ELC ends (the market value of $3.3M/WAR would put him around $6M AAV for his next deal). However, DeSmith has never made over $1.8 AAV, despite his track record. He is providing elite backup numbers over his two seasons in Dallas, allowing the Stars to pick up a large amount of points even when resting starter Jake Oettinger.
Hellebuyck and Sorokin have averaged the most WAR over the last 4 years (4.39, 4.10, respectively), so even at their $8.5M and $8.25M AAV, the Jets and Islanders are getting quite the deals. Logan Thompson has shown that last season was not a fluke, replicating his dominant performance since coming over to Washington from Vegas. This rewarded the confidence bestowed upon him by Washington’s management when they gave him a six-year contract at $5.86M AAV.
Where teams are paying for what they used to have:
| Goalie | Cap Hit | Avg WAR/yr | $/WAR |
|---|---|---|---|
| Jacob Markstrom | $6.0M | 0.07 | $81.7M/WAR |
| Sergei Bobrovsky | $10.0M | 1.07 | $9.3M/WAR |
| Juuse Saros | $7.74M | 1.16 | $6.7M/WAR |
| Jordan Binnington | $6.0M | 0.94 | $6.4M/WAR |
Markstrom’s situation is really tough to see when looking at the numbers. Among all starters this season, he is averaging the lowest WAR at 0.07 WAR/Year, while collecting $6M a year (with a 2-year extension kicking in next season at the same cap hit). The return on that investment has been almost nonexistent. Bobrovsky is having a rough season on the tail end of his contract, drastically lowering his average WAR, so although the optics may not look good on paper, the Panthers are not going to complain with the two cups he has given them.
Saros is a difficult case as well, we mentioned him earlier. He peaked at +0.738 GSAx/60 in 2022-23 after seemingly improving every year, but he has not been above average in any season since. The contract was signed after he dropped all the way down to -0.06 GSAx/60 in 2023-24.
The pattern across all of these goalies: teams signed them at or near their career peaks, on long deals that would only be worth it if they kept up the play that post-peak data says very few goalies sustain.
The Buy/Sell/Hold Framework
Based on current performance trajectory, age, and year-to-year stability, we can look at current goaltenders and create guidelines on how to handle goaltenders once they’re up for their next contract:
BUY - Age ≤ 27 AND above-average GSAx/60. Young and producing, lock them up cheap:
- Jeremy Swayman, 27: Already signed before the 2024-25 season (eight-years, $8.25M AAV), Swayman has a three-year average of +0.417 GSAx/60, and is entering the phase of his career where sustained elite performance can be more predictive. His 2024-25 season, in which he performed below replacement level, adds a bit of risk, but there is no doubt that his ceiling and consistency throughout other seasons warrant a long commitment.
- Jakub Dobes, 24: We talked about him before. +0.395 GSAx/60 over his first two seasons, and getting better. Get him signed on the cheaper now while you can.
- Joel Hofer, 25: Lowest year-to-year GSAx/60 standard deviation in the entire dataset (0.092), +0.288 average, reliable in a way most younger goalies are not, even more impressive on a struggling Blues team.
- Spencer Knight, 25: +0.276 average GSAx/60, low variance, ascending trend. He has performed well as the starter in Chicago. The Blackhawks extended him before the 2025-26 season at three-years, $5.83 AAV, and it looks like their gamble is going to pay off before Knight can command more money.
EXTEND - Age 28-31 AND elite GSAx/60. Proven prime, worth a multi-year commitment:
- Logan Thompson, 29: Averaging 2.97 WAR over the last four seasons is fifth best in the league over that span. He has never dropped below replacement level and has been elite since being traded to Washington. Well worth his previously mentioned extension.
- Ilya Sorokin, 30: Elite numbers throughout his career. Averaging the second-most WAR over the last four seasons (4.1), worth an implied value of $12.74M AAV. He has shown no signs of regression, potentially on his way to his first Vezina this season.
HOLD - Above-average numbers but doesn’t fit BUY or EXTEND criteria (e.g., age 28-31 but not quite elite):
- Mackenzie Blackwood, 29: Blackwood’s raw three-year numbers undersell where he actually is right now. His 2025-26 GSAx/60 of +0.245 is above average, and his trajectory has been consistently upward. If he posts another above-average season, the label likely flips to EXTEND. For now, you want him on your roster; you just don’t want to overpay for a track record he hasn’t fully established yet.
- Karel Vejmelka, 29: Vejmelka’s numbers say above average on a weighted basis, but his latest season was a step back (+0.103 GSAx/60 after +0.367 the prior year), and his volatility is among the higher marks in the dataset (0.357 std). At $4.75M AAV, he gives you league-average goaltending at a price that doesn’t hurt you, which in today’s market is more valuable than it sounds.
SHORT-TERM ONLY - Age 32+ with above-average OR elite numbers. Not a sell, but don’t commit long-term:
- Casey DeSmith, 34: Mentioned before, DeSmith has put up starter-level numbers throughout his career and has been arguably the best backup in the league over the last two seasons. He is worth an extension, but no need to commit to more than 2 years, especially since he has never commanded more than $2M AAV.
- Anthony Stolarz, 32: The most volatile goalie in the dataset (GSAx/60 std 0.623) just crashed from a historic +0.9 GSAx/60 to -0.54 in one season. The ceiling is there, but not worth anything more than a one or two-year deal. Toronto signed him to a four-year, $3.75M extension before this season that kicks in next season. While a long contract is not advised, it may be worth it at that cap hit.
SELL - The data does not support continued investment at current cap hit:
- Juuse Saros, 31: Three straight below-average seasons, high volatility, $7.74M AAV for seven more seasons. His peak was well worth the money, but Nashville may be having trouble competing if Saros doesn’t rebound soon.
- Charlie Lindgren, 32: Put up above-average numbers for Washington in 2023-24 (+0.161 GSAx/60), but has been around replacement level otherwise, especially since Logan Thompson joined the team. He only has two more seasons at $3M, but you could likely get someone who performs just as well for around $1M and improve team depth elsewhere.
Five Rules
All of it reduces to this:
1. Never sign based on one peak season. The average drop from peak to Year 1 is -0.561 GSAx/60, and it happens immediately. Require two consecutive elite seasons before committing to long-term money.
2. Long deals don’t buy you more production; they buy you less flexibility. A 1-year and a 5-year deal signed after a peak produces almost identical WAR per season. The difference is that the shorter deal preserves flexibility when, not if, the regression arrives.
3. Free agent signings are structurally disadvantaged by team-switch regression. A goalie who looked elite in one market will revert on average by -0.647 GSAx/60 after changing teams. The best returns in this dataset come from internal promotions and early extensions, not open market competition in July.
4. Volatility is a real risk factor that should shorten contract length, not just lower AAV. A high-mean, high-variance goalie on a seven-year deal carries enormous floor risk.
5. Invest in your backup. Teams with elite backups see their starters perform measurably better. A $3M investment in an elite backup is not just insurance; it is one of the more efficient uses of goalie cap dollars available.
If you want to dig into the underlying numbers yourself, the Goaltender Database has every goalie season in this dataset - filterable by season, situation, team, age, and workload. Sort by GSAx/60, SV%, or any other column to build your own picture.
All data sourced from MoneyPuck, covering the 2016-17 through 2025-26 NHL seasons. Performance analysis uses even-strength situations only with a minimum qualifying threshold of 600 five-on-five minutes per season. Replacement level is set at -0.10 GSAx/60. Goals-to-wins conversion uses the standard 6:1 approximation, consistent with Evolving Hockey’s WAR methodology. Contract figures are approximate, sourced from PuckPedia. All WAR figures are the author’s own calculations derived from MoneyPuck shot and goalie performance data. All data is current as of collection on 3/30/2026.