A goalie saves prop is an over/under on how many shots one goaltender stops. Books hung a median number of 25.5 across 2,364 games in our archive. Against 3,516 posted numbers the goalie finished under 53.75% of the time, which is almost exactly what the median price of −115 already charges.
Every figure below comes from our own archive of 18,564 saves quotes joined to official box scores.
What is a goalie saves prop?
An over/under on how many shots one goaltender stops in a game. The book posts a number, you take the side you think the goalie finishes on, and it settles on the save count in the official box score. Saves are shots on goal that did not go in, so the number depends on two things pulling in opposite directions: how many shots the goalie faces, and how many of them beat him.
That second part is what makes the market behave unlike a scoring prop. A goalie in a busy, losing game can pile up saves. A goalie whose team controls the puck all night can win comfortably and finish nowhere near the number. Shots faced, not quality of goaltending, is the main driver.
What numbers do books hang?
Our archive holds 18,564 goalie saves quotes across 2,364 NHL games and 106 goaltenders, from 2023-11-02 to 2026-06-12. The median number is 25.5, and almost everything sits between 21.5 and 29.5. Not every goalie in a game gets one: 1,371 of the games in the archive had a number on a single goalie, and 977 had numbers on both.
| Number | Quotes | Share | Games |
|---|---|---|---|
| 20.5 | 316 | 1.70% | 55 |
| 21.5 | 766 | 4.13% | 142 |
| 22.5 | 1,722 | 9.28% | 316 |
| 23.5 | 2,552 | 13.75% | 443 |
| 24.5 | 3,022 | 16.28% | 535 |
| 25.5 | 3,125 | 16.83% | 543 |
| 26.5 | 2,533 | 13.64% | 454 |
| 27.5 | 2,230 | 12.01% | 401 |
| 28.5 | 1,410 | 7.60% | 269 |
| 29.5 | 560 | 3.02% | 111 |
| 30.5 | 256 | 1.38% | 49 |
Numbers with fewer than 100 quotes are omitted. n=18,564 quotes, 2023-11-02 to 2026-06-12.
Do books agree on the saves number?
Mostly, which is worth knowing because plenty of prop markets do not. Of 2,982 goalie-games in our archive quoted by two or more books, 2,822 carried the same number at every book, 94.63%. The remaining 160 split, and never more than two ways: no goalie-game in the archive had more than 2 different numbers on it at once. When books do split it is by a single save.
The prices are where they compete instead. Across 9,280 quotes where one book priced both sides at one instant, the two implied probabilities summed to a median 106.93 rather than 100, so the median margin held in this market was 6.48%. The median price on each side was −115.
| Bookmaker | Quotes | Games | Captured |
|---|---|---|---|
| betmgm | 5,416 | 1,903 | 2023-11-17 to 2026-06-12 |
| draftkings | 4,442 | 1,582 | 2023-11-10 to 2026-06-12 |
| williamhill_us | 2,588 | 941 | 2023-11-03 to 2025-03-01 |
| betonlineag | 2,560 | 950 | 2024-10-29 to 2026-05-24 |
| bovada | 1,830 | 694 | 2023-11-06 to 2026-06-12 |
| fanduel | 1,380 | 476 | 2023-11-02 to 2026-06-10 |
| fanatics | 314 | 109 | 2025-10-09 to 2025-10-30 |
Books with at least 300 quotes in the archive. Coverage is uneven and some books stop and start, which is why the windows differ.
What did the number actually do?
Because we hold the box scores as well as the prices, each posted number can be compared with the saves the goalie went on to make. Across 3,516 posted numbers, the goalie finished over 1,626 times and under 1,890 times, 46.25% against 53.75%. There were no exact landings, because books hang these on the half.
That under share is not the finding it might look like. A bet at the median price of −115 needs 53.49% to break even, so the whole gap is 0.26 of a percentage point. On a sample of 3,516 the standard error is 0.84 points, which makes that gap about 0.31 of one standard error. It is noise, not a finding, and it is printed here because reporting the rate without the yardstick beside it would be the misleading version.
The averages say the same thing more directly. The mean posted number was 25.35 and the mean save count was 24.51, a gap of 0.84 of a save. The median save count was 25 against a median number of 25.5, so the typical game landed half a save under while the average landed nearly a full save under. That difference between the median and the mean is the tail: games where the goalie was pulled or the night ended early drag the average down without moving the middle.
| Number | Posted | Over | Over share | Mean saves |
|---|---|---|---|---|
| 20.5 | 55 | 26 | 47.27% | 18.53 |
| 21.5 | 144 | 73 | 50.69% | 21.51 |
| 22.5 | 329 | 155 | 47.11% | 21.74 |
| 23.5 | 481 | 225 | 46.78% | 22.65 |
| 24.5 | 568 | 271 | 47.71% | 24.10 |
| 25.5 | 566 | 248 | 43.82% | 24.33 |
| 26.5 | 488 | 213 | 43.65% | 25.22 |
| 27.5 | 436 | 210 | 48.17% | 26.91 |
| 28.5 | 271 | 116 | 42.80% | 27.52 |
| 29.5 | 112 | 58 | 51.79% | 28.37 |
| 30.5 | 49 | 26 | 53.06% | 29.45 |
One observation is one game and one posted number. Numbers with fewer than 40 observations are omitted. n=3,516 across all numbers.
| Season | Posted | Over | Over share | Under share |
|---|---|---|---|---|
| 2023-24 | 1,119 | 510 | 45.58% | 54.42% |
| 2024-25 | 1,148 | 509 | 44.34% | 55.66% |
| 2025-26 | 1,249 | 607 | 48.60% | 51.40% |
Split by season so the overall figure cannot hide a single unusual year. The 2023-24 row starts in November because that is when our capture of this market starts.
How close is the number to the result?
Not very, and that is the most useful thing on this page. Only 23.21% of posted numbers finished within two saves of the result, and 46.70% landed more than five away. A saves number is the centre of a wide distribution rather than a close prediction of one night's work, which is what you would expect from a statistic driven by how many shots the other team happens to take.
| Distance from the number | Observations | Share |
|---|---|---|
| within 2 | 816 | 23.21% |
| 2 to 5 | 1,058 | 30.09% |
| 5 to 10 | 1,121 | 31.88% |
| more than 10 | 521 | 14.82% |
Absolute difference between the save count and the posted number. n=3,516.
What happens if the goalie does not play?
The prop is normally voided and the stake returned, because the market has nothing to settle against. This is not a hypothetical: 17 of the 3,533 numbers in our archive belong to goalies who were given a number and then recorded no saves that day. Confirm the rule your book applies, since a goalie who enters in relief and faces one shot is a different case from one who never leaves the bench, and books word that differently.
The same question decides whether these props are worth watching at all before a starter is confirmed. A number hung before the starting goalie is announced is priced across more than one possible goaltender.
Does overtime count for goalie saves props?
Overtime saves count, because they are shots on goal that were stopped. Shootout attempts do not, since a shootout is a skills competition rather than part of play and its attempts are not recorded as shots on goal. That distinction matters more often in hockey than people expect: see our NHL overtime rules page for how far it reaches across the rest of the menu.
How this was measured
- Source. odds.lines, sport_key icehockey_nhl, market_key player_total_saves: 18,564 quotes across 2,364 games, 8 bookmakers and 106 goaltenders, 2023-11-02 to 2026-06-12. Save counts come from stats.player_game_stats_icehockey_nhl, built from the free official NHL API.
- Opening prices only. Every row in this market in our archive is an opening capture. Nothing here is a closing line and no closing-line claim is made from it.
- One observation. One game and one posted number. A goalie quoted at the same number by five books counts once, not five times, so the rates are not weighted by how many books happened to be captured. Where books split, each distinct number is its own observation.
- How the two sides are joined. The odds feed names goalies in full and the stats feed abbreviates the first name, and no shared game id exists between them, so the join is built on first initial, surname and the game date, with accents folded. A goaltender plays at most one game a day, which is what makes that key unique. Any name that resolves to more than one player, or any date and name pair with more than one row, is dropped: 6 of 31,772 pairs. The join matched 3,516 of 3,533 posted numbers, 99.52%.
- The unmatched remainder. 17 posted numbers, 0.48%, belong to goalies with no save count that day, which is what a goalie who did not take the ice looks like in a box score. They are excluded rather than scored as a loss, because a voided prop is not a losing one.
- Break-even and error. The 53.49% figure is the implied probability of the median price in this market, −115, computed in closed form. The 0.84-point standard error is the binomial standard error of the measured rate at n=3,516.
- What this cannot tell you. These are base rates over a whole archive. They say nothing about any particular goaltender on any particular night, and this site publishes no selections. Past frequencies are measurements of what happened rather than forecasts of what will.
Next steps
Frequently asked questions
How does a goalie saves prop work?
The book posts a number of saves and you take the over or the under. It settles on the save count in the official box score, which counts shots on goal that the goalie stopped. Overtime saves count. Shootout attempts do not, because they are not recorded as shots on goal.
What is a typical NHL goalie saves line?
25.5 is the median in our archive of 18,564 quotes. The most common numbers are 25.5 and 26.5, and the range runs from about 20.5 up to 30.5. Numbers outside that band are rare.
Does the over or the under hit more often on goalie saves?
The under, but not by enough to mean anything. Across 3,516 posted numbers in our archive the goalie finished under 53.75% of the time, against the 53.49% a bet at the median price of −115 needs just to break even. The 0.26-point difference is about 0.31 of one standard error on this sample, which is noise.
What happens to a saves prop if the goalie does not start?
It is normally voided and the stake returned, because there is nothing to settle against. 17 of the 3,533 numbers in our archive belong to goalies who recorded no saves that day. Books word the relief-appearance case differently, so check the rule on your account.
Do overtime saves count towards a goalie saves prop?
Yes. Overtime is part of play, so a save made in overtime is a shot on goal that was stopped and it counts. A shootout is a skills competition and its attempts are not shots on goal, so nothing that happens in one changes the save count.
Do sportsbooks post the same saves number?
Usually. Of 2,982 goalie-games quoted by two or more books in our archive, 94.63% carried the same number everywhere, and no goalie-game ever carried more than 2 different numbers. Books compete on the price attached to the number more than on the number itself.
Related: NHL overtime betting rules · NHL key numbers and final scores · NHL puck line explained · NHL betting guide
Every figure on this page is a historical measurement of what already happened, produced by a committed, re-runnable script over our own odds archive. Past frequencies are not forecasts and are not betting advice: Nebula Insights publishes no picks, no selections and no recommended wagers. 18+ only.