Prop LabSpeed, faceoffs, lines and goalies
When a Good Goalie Joins a Bad Team for Goalies, Save % Drops About .005. The Team Explains Part of It, and Luck Plays a Part
We followed every NHL goalie who changed teams since 1980 and rated each team for goalies before the move, which is not the same as rating how good the team is. About 60% of the new team’s goalie rating shows up in the goalie who arrives. Luck plays a part too: a goalie coming off a great or terrible season usually lands closer to their normal level wherever they go.
By Edgehalla ResearchPublished 8 min read
On this page
The question
When Devan Dubnyk left Edmonton, he went from 22.8 points of save % below league average to 21.4 above in Minnesota. When Cam Talbot went the other way, from the Rangers to Edmonton, he dropped from 11.2 above average to 2.0. Stories like these make it sound as if the team decides everything. We wanted to know how much of it the team really decides.
One distinction matters throughout. A bad team for goalies is not the same as a bad team. Edmonton has been a top-12 team in points % for most of the McDavid era and below average for goalies every season. Winnipeg has been mid-table for most of the decade and one of the best teams for goalies. Our ranking of NHL teams for goalies rates every franchise by comparing the same goalies there and elsewhere.
Here we use those ratings to test moves. For every season since 1980-81, we rated each team using only the 10 seasons before it, then watched what happened to goalies who arrived. That way the ratings could not have seen the outcome.
A move counts if the goalie faced 500+ shots for the old team in one season and 500+ for a new franchise the next. That gives 325 moves from 1980-81 to 2025-26. Save % is measured against the league average each season, in points: 1 point is .001.
Good to bad, bad to good: the four moves
We split teams into thirds by their rating for goalies before each season, and goalies into halves by their forecast (their whole record, shrunk toward average, with past teams taken out). Then we looked at two numbers for each kind of move: the change from the goalie’s last season, and the result against the goalie’s own forecast.
What happened to goalies after a move, 1990-91 to 2025-26
Save % points: change from last season vs result against the goalie’s own forecast
Moves to better teams for goalies brought big jumps from last season, but smaller gains against each goalie’s forecast.
- Change from last season
- Against the goalie’s forecast
The first column is the one fans see, and it is large: a weak goalie moving to a good team for goalies gained 11.1 points, about .011 of save %. Luck plays a part in that number. One season of 1,500 shots swings save % by about 7.5 points by chance alone, so a goalie coming off a bad year usually has a better one next, wherever they play. The second column takes that out by comparing each goalie with their whole record. What is left, about 2 to 4 points, is the team.
How much of the new team carries over
The cleaner measure is a slope: for each point of the new team’s rating, how many points show up in the arriving goalie, beyond the goalie’s own forecast?
Share of a team’s goalie rating that shows up in its goalies
Slope of save % (vs the goalie’s forecast) on the team’s rating, fitted before the season
Goalies who arrive pick up about 60% of the new team’s rating, about the same as goalies who were already there.
Arriving goalies get about the same lift or drag as the goalies who were already there. So the team effect is real and it is the team’s, not a story about one goalie fitting in. A slope under 1 means the ratings are noisy and partly out of date by the time the season starts.
Which part of the rating matters? Since 2015, the team’s last season carried over strongly (slope 1.46, 90% range 0.24 to 2.60), while its longer franchise record barely did (0.34, −0.51 to 1.18). A team’s goalie reputation from five or ten years ago is close to useless. Defences, coaches and systems change faster than that.
Good team, or good team for goalies?
Does an arriving goalie do better because the new team is good for goalies, or just because it is a good team? Across the league the two are only loosely linked: from 2016-17 to 2025-26, a team’s goalie rating correlates at 0.22 with its shot share. So we can test both on the same moves, using the new team’s numbers from the season before the move.
| Model | Rating for goalies (slope) | Team quality (points per SD) |
|---|---|---|
| Points % alone | — | +2.5 (1.1 to 3.9) |
| Shot share alone | — | +1.7 (−0.1 to +3.3) |
| Rating for goalies + points % | 0.32 (−0.24 to 0.80) | +2.1 (0.6 to 3.5) |
| Rating for goalies + shot share | 0.56 (0.05 to 1.07) | +1.4 (−0.3 to +3.1) |
Both matter. A good team helps an arriving goalie: a team one standard deviation better in points % the year before added about 2.5 points of save %. But points % already includes how the goalies did, so it overlaps with the goalie rating. Shot share does not, and with shot share in the model, the rating for goalies still carries over at 0.56. With 173 moves we cannot say which of the two matters more. The practical answer is to use both.
Famous moves, checked
| Goalie | Move | Season | Team gap | Year before | After | Forecast |
|---|---|---|---|---|---|---|
| Devan Dubnyk | EDM → MIN | 2014-15 | +9.0 | −22.8 | +21.4 | −1.5 |
| Jaroslav Halak | NYI → BOS | 2018-19 | +9.7 | −4.5 | +12.4 | +2.4 |
| Mackenzie Blackwood | SJS → COL | 2024-25 | +6.2 | −4.6 | +13.2 | −0.2 |
| Joonas Korpisalo | OTT → BOS | 2024-25 | +9.0 | −13.7 | −6.6 | −3.6 |
| Chad Johnson | CGY → BUF | 2017-18 | +7.5 | −3.2 | −21.4 | −0.7 |
| Cam Talbot | NYR → EDM | 2015-16 | −8.8 | +11.2 | +2.0 | +1.9 |
| Jaroslav Halak | MTL → STL | 2010-11 | −8.9 | +13.0 | −2.5 | +0.1 |
| Brian Elliott | STL → CGY | 2016-17 | −7.3 | +15.1 | −3.9 | 0.0 |
| Linus Ullmark | BOS → OTT | 2024-25 | −9.0 | +11.3 | +9.0 | +4.1 |
| Anton Khudobin | BOS → DAL | 2018-19 | −7.1 | +1.0 | +13.0 | −0.8 |
| Dan Vladar | CGY → PHI | 2025-26 | −7.4 | −2.3 | +10.2 | −3.4 |
- Moves up mostly worked. Dubnyk, Halak to Boston and Blackwood to Colorado all landed well above their forecasts. Korpisalo improved by 7 points in Boston and was still below average. Chad Johnson’s move to Buffalo went badly even though Buffalo rated well.
- Moves down mostly hurt, but less than the drop suggests. Talbot, Halak to St. Louis and Elliott each lost 9 to 19 points from a great season, but all three were forecast near average. The great season was the outlier.
- A great goalie survives a bad team for goalies. Ullmark had the best forecast in this list (+4.1). Ottawa rated 9 points worse than Boston, and he still finished 9.0 above average.
- Counter-examples are common. Khudobin and Vladar joined teams rated well below their old ones and finished 12 to 14 points above their forecasts. With one season of shots, luck can swamp a 7-point team gap.
Can you bet on goalie moves?
If the market priced a moved goalie on last season’s save %, it would overrate goalies coming off big years and underrate the rest. We tested this on 2024-25 and 2025-26 closing moneylines and totals, using only numbers known before each game.
| Bet | Bets | Won | Priced | ROI (90% range) |
|---|---|---|---|---|
| Back every goalie in a first season with a new team | 850 | 51.4% | 50.7% | −2.9% (−9.5% to +3.2%) |
| Fade new-team goalies with hot reputations | 419 | — | — | −10.1% (−17.2% to −2.9%) |
| Back new-team goalies with cold reputations | 431 | — | — | −3.0% (−10.8% to +3.8%) |
| Back teams with the best goalie-rating edge | 1,308 | 52.4% | 53.9% | −6.7% (−11.5% to −2.1%) |
Nothing beat the close. Goalies with hot reputations on new teams actually did slightly better than priced, and that flipped between the two seasons. A few goalies drove most of it: Logan Thompson won 73.8% of his Washington starts in 2024-25 against 55.2% priced. On totals, overs in games with a hot-reputation goalie on a new team lost 7.4%.
The limit is sample size. Only 13 to 16 established goalies change teams each season, so two seasons of odds can rule out only a large mispricing. Saves props may be the better test: shots against follow the team, so a goalie’s saves line on a new team should come from the new team’s shots against. That needs a season of saves lines, which we have not tested yet.
How to project a goalie on a new team
- Start from the goalie’s long-run level, not last season. In our forecasting test, last season’s raw save % was worse than guessing league average for goalies who moved. Use several seasons, pulled hard toward average.
- Add about 0.6 of the new team’s current goalie rating. A goalie forecast at +2 who joins a team rated −5 projects at about −1. One who joins a team rated +4 projects at about +4.4.
- Weight the team’s last season over its history. The franchise’s old reputation adds almost nothing.
- Expect a wide range. One season of 1,500 shots swings save % by about 7.5 points from luck alone, more than any team effect.
For fantasy, this matters more for wins and goals-against average than for save %. Both follow the team almost completely: they repeat year to year at about 0.42 for goalies who stay and about 0 for goalies who move. The details are in how much of save % is the goalie.
What we could not settle
- Small groups. Each kind of move has 20 to 35 cases since 1990, so the ranges are wide.
- Benched goalies drop out. We need 500+ shots on the new team, so a goalie who flopped and lost the job early is missing. That probably makes moves look a little better than they were.
- Teams choose who they sign. A team that knows its system suits a certain style of goalie will pick one, and that is hard to separate from the team effect.
- Strong goalies might be helped less. Our data cannot tell whether a good team for goalies lifts a weak goalie more than a strong one. The interaction estimate was 29 with a 90% range of −76 to +146.
Show the work
- Tests written down
- Not in advance (exploratory)
- Results computed
Goalies who faced 500+ shots for one franchise and 500+ for a new one the next season, 1980-81 to 2025-26 (NHL stats API game logs). Team ratings and goalie forecasts come from a crossed random-effects model fitted only on the 10 prior seasons. Betting checks use 2024-25 and 2025-26 near-close odds. None of these tests was pre-registered.
| Test | Result | Verdict |
|---|---|---|
| A new team’s goalie rating carries to an arriving goalie | slope 0.58 (90% 0.28 to 0.90), n 325, 1980-2026 | Real |
| Arrivals get the same team effect as goalies already there | movers 0.58 vs stayers 0.64 (0.52 to 0.76) | Supported |
| Last season’s team rating matters more than franchise history | 1.46 vs 0.34, 2015-26 movers (n 107) | Supported, wide |
| Bad → good team for goalies beats the goalie’s forecast | +1.9 pts (90% −1.6 to +5.4), n 35 | Small, not proven |
| Strong goalie, good → bad team for goalies: change from last season | −5.1 pts (90% −9.0 to −1.2), n 20 | Real |
| Overall team quality helps an arriving goalie beyond the goalie rating | shot share +1.4 pts per SD (90% −0.3 to +3.1) with rating 0.56 (0.05 to 1.07), n 173 | Both matter, wide |
| Good teams for goalies lift weak goalies more than strong ones | interaction 29 (90% −76 to +146) | Not detected |
| Closing lines misprice goalies on new teams | 51.4% won vs 50.7% priced, 850 starts; back-all −2.9% | Priced |
| Fade new-team goalies with hot reputations | −10.1% ROI (90% −17.2% to −2.9%), 419 bets | Loses |
| Back teams with the best goalie-rating edge | −6.7% ROI (90% −11.5% to −2.1%), 1,308 bets | Priced |
Frequently asked questions
What happens when a good goalie goes to a bad team for goalies?
On average, save % falls about 5 points (.005) from the season before. Luck plays a part: a strong season is usually followed by a more normal one. Against the goalie’s own forecast, the drop is about 1 point (20 strong goalies moving from a top-third to a bottom-third team for goalies since 1990-91).
Does a bad goalie get better on a good team for goalies?
Usually, by a lot from the year before: weak goalies moving from a bottom-third to a top-third team for goalies gained 11.1 points of save %. About 4 of those points beat their forecast; luck explains much of the rest, because a bad season rarely repeats.
Does save percentage carry over when a goalie changes teams?
Only a little. Last season’s save % carries over at a slope of about 0.07 for goalies who move, against 0.40 for goalies who stay. The new team’s goalie rating carries over at about 0.6.
How should I project a goalie who signed with a new team?
Take the goalie’s long-run level pulled toward league average, then add about 0.6 of the new team’s current goalie rating. Weight the team’s last season over its history, and ignore last season’s raw save %.
Can you bet on goalies who change teams?
Not on the evidence. Over 850 starts in 2024-25 and 2025-26, goalies in their first season with a new team won 51.4% against a 50.7% closing price, and every rule we tried lost money after the vig.
More in Speed, faceoffs, lines and goalies
- Verdict: Promising
Do Faster NHL Teams Win More? What NHL EDGE Speed Data Says
- Verdict: Myth
Does Playoff Hockey Reward Hitters Over Speed? We Tested Five Postseasons
- Verdict: Mixed
NHL Faceoff Matchups: Ignore the Head-to-Head, Watch the Hands
- Verdict: Mixed
NHL DFS Stacking Works. “Chemistry” Is Mostly Noise.
- Verdict: Mixed
How Much of a Goalie’s Save Percentage Is the Goalie?
- Verdict: Mixed
The Best and Worst NHL Teams for Goalies: Boston Helps Most, Philadelphia Hurts Most
Sources and method
- NHL stats (NHL.com)
- NHL API reference (community documentation)
- The Odds API (historical NHL moneylines, puck lines and totals)
- Empirical Bayes shrinkage
- Responsible Gambling Council (Canada)
- ConnexOntario (1-866-531-2600)
- National Council on Problem Gambling (1-800-GAMBLER)
Research and analysis, not betting advice. Past results do not guarantee future returns; bet only what you can afford to lose.