NHL Goalie Saves Props: Why Shot Volume Beats Save Percentage
The saves prop has a goaltender's name on it and is mostly about the team in front of him. Over 7,840 starts, how many saves a goalie makes tracks the shots he faces at 0.98 and his own save percentage at 0.57 — and that single fact reorganises how the market should be read.
Published September 2026 · 8 min read
1. A Saves Prop Is a Shot-Volume Market
A save requires a shot. That sounds trivial and it is the entire market. Across 7,840 starts in the 2023-24, 2024-25 and 2025-26 seasons, the correlation between a goalie's saves and the shots he faced is 0.98. The correlation between his saves and his save percentage that night is 0.57.
Save percentage clusters tightly — most starts land near .900 — so it cannot generate much spread in the save count. Shots faced range from the high teens to the high thirties, so it generates nearly all of it. When you take an over on saves you are betting the opponent will shoot a lot, and hoping the goalie is merely ordinary while they do it.
2. How Sportsbooks Set the Number
Books build a saves line in two steps: project the shots the goalie will face, then apply an expected save percentage near .900 and round to the nearest half save. The result almost always lands between 24.5 and 29.5 — the same band our model prices, because it is where the distribution lives. The league average start in our archive is 27.9 shots faced and 25.1 saves.
Since the second step barely varies, the shot projection is the line. Anything that moves expected shots — a heavy forechecking opponent, a team missing two defencemen, a game expected to be chased from behind — moves the number more than anything about the goaltender.
3. What Each Line Actually Hits
Over the same 7,840 starts:
| Line | Cleared in |
|---|---|
| Over 24.5 saves | 52.6% of starts |
| Over 27.5 saves | 36.1% of starts |
| Over 29.5 saves | 26.6% of starts |
Those are league-wide rates, and like any base rate they describe a population you are not betting into. The split that matters is by the team in front of the goalie.
4. The Good Goalie Problem
Strong teams suppress shots. Strong teams also tend to employ the goalies whose names a casual bettor recognises. The two facts pull against each other, and the saves market is where it shows:
| Goalie's team | Starts | Saves per start | Over 27.5 |
|---|---|---|---|
| Goalies on teams allowing under 27 shots | 2,531 | 22.9 | 23.9% |
| Goalies on teams allowing 30 or more | 944 | 28.1 | 52.9% |
The same 27.5 line is a coin flip for one group and a 24% shot for the other, and nothing in that gap is about goaltending. A goalie on a team that concedes 30 shots a night is the structural over; the starter behind a shot-suppressing defence is the structural under, however good he is.
The mirror image of this is the shots on goal prop, where the same opponent shot volume sets a skater's number.
5. How Our Model Prices a Saves Line
The saves projection is an XGBoost regression on 16 features, trained on 4,939 starts and tested on 2,118 across three seasons. The inputs mirror the argument above:
- Volume: shots against over the last 5 and last 10, and the opponent's shots and goals per game.
- The goalie: saves over the last 5 and last 10, the spread of the last 10, save percentage over the last 10 and on the season, goals against, and the season save total and its spread.
- Context: home or away, rest days, a back-to-back flag, and how many starts the goalie has made.
Probabilities for each line come from a normal distribution around the projection whose width is the larger of the goalie's recent spread and 0.6 of his season spread. Held-out error is 5.4 saves (MAE), with cross-validated error of 5.45 ± 0.13 and an R² of 0.05.
That R² deserves saying plainly rather than burying: the model explains about five percent of the game-to-game variance in save totals. It is not a weak implementation of a strong idea — it is an honest measurement of a noisy market. Whether a goalie faces 22 shots or 36 depends on how one hockey game happens to go, and no amount of history fixes that. Anyone selling certainty on this market is selling something the data does not contain.
6. What We Will Not Claim
Our goalie saves tracker has 11 graded picks. We will not print a hit rate off eleven results, and the same holds for our other NHL prop markets until they have a season behind them. The NHL numbers we do publish and grade in public are the team markets: 64.4% on the puck line over 261 picks and 60.3% on the moneyline over 58.
What the saves model is genuinely for is the shape of the distribution — where a book's 27.5 sits inside a goalie's real range on a given night, given who he is facing. That is context for a price, not a claim to beat one.
7. Frequently Asked Questions
How are NHL goalie saves props set?
The book projects how many shots the goalie will face, multiplies by an expected save percentage close to .900, and rounds to the nearest half save. Most lines land between 24.5 and 29.5. Because save percentage barely moves between goalies, the shot projection does almost all the work.
Is a better goalie a better saves bet?
Usually the opposite. Elite goalies tend to play for teams that concede fewer shots, and you cannot make a save you never face. Over 7,840 starts, goalies on teams allowing under 27 shots averaged 22.9 saves, while goalies on teams allowing 30 or more averaged 28.1.
What correlates with a goalie’s save total?
Shots against, almost perfectly. Across 7,840 starts the correlation between saves and shots faced is 0.98, while the correlation between saves and save percentage is 0.57. A saves prop is a shot-volume market wearing a goaltender’s name.
How often does the over 27.5 saves hit?
In 36.1% of all starts over three seasons. That splits hard by team: goalies facing 30 or more shots a night cleared 27.5 in 52.9% of starts, against 23.9% for goalies on teams that suppress shots. The line moves less than the team behind it.
Do back-to-backs matter for goalie saves props?
They matter mainly because they decide who starts. A backup drawing the second night is often a different projection from the starter, and teams playing their own second night tend to concede more shots. Our model carries rest days, a back-to-back flag and the start count as separate inputs.
See tonight's goalie saves projections and the shot volume behind them
Projected saves with over/under probabilities at every line from 24.5 to 29.5, built from the opponent's shot volume rather than from reputation.
Start your free 3-day trialThe week ahead, once a week
Last week's graded record — wins and losses — the plays for the week ahead, and a subscriber discount code. One email, no daily noise, unsubscribe in a click.