Fancy Stat Summer School - Quality Starts

Goaltending is an area of the NHL game that is ripe for microanalysis, but has only delivered a couple of relatively simple stats that are used by the hockey analytics community. Quality Starts try to award goalies who consistently give their teams a chance to win.

What is a Quality Start?

A Quality Start (QS) is a goaltending statistic that is awarded to a goaltender who gave his team a reasonable chance to win a game. This is quantified by the goaltender's save percentage within an individual game itself and comparing it to the league standards for an average SV%, and the established 'replacement level' SV% if the goalie faces a low amount of shots faced in a given game (usually less than 20). If a goalie outperforms the league average save percentage, he is awarded a quality start. Additionally, a goalie is awarded a quality start if he allows 2 goals or less while keeping his save percentage above the expected performance of a replacement level goaltender.

As the name implies, it is only awarded to goalies who start the game, so a goaltender coming on in relief of either a pulled goalie or injured goalie cannot, by definition, be awarded a Quality Start.

What Is Replacement Level SV%?

Obviously, to fully understand Quality Starts and its applications, one has to understand how we come up with these different standards. Average SV% is quite simple: it is the total amount of saves divided by the shots faced by goaltenders across the league through the entire season. According to Rob Vollman, author of Hockey Abstract and editor of Hockey Prospectus, the average save percentage was 91.3% from 2005-06 to 2009-10, and has increased to 91.7% since 2010-11.

Replacement Level SV% is defined by taking the cumulative save percentage of all goaltenders who are readily available to every team, either as an AHL call-up or an infrequently utilized backup. I'm not entirely sure the definition Vollman and Hockey Prospectus use, but I calculate it using the NHL's 'definition', which is every goalie that is not eligible to win either the William M. Jennings (For best Team GAA) or Roger Crozier (for top SV%) Awards. The list of goalies eligible for those awards is accessible by simply clicking on the 'Save % Leaders' in the statistics section of

The NHL essentially offers the view that appearing in 24 or fewer NHL games in an 82 game season is not worthy of being compared to other NHL goalies (it was 14 games for the shortened 2012-13 season). So, if one calculates the save percentage of all goalies with less than 25 appearances, you can come up with the league's Replacement Level SV%. It has typically worked out to be 88.5% according to Vollman.

One can use the same methodologies used for the NHL on different leagues that track SV% and game to game shot statistics across the world, coming up with standards that make sense for the AHL, KHL, SHL, or Canadian major junior leagues. We'll continue our education by just using the NHL as the example to keep things simple.

Practical Example

Let's judge two goalies from last year's playoffs who played in a 7 game series against each other by quality starts. The two goalies used are Braden Holtby of the Capitals and Henrik Lundqvist of the Rangers.

Holtby vs Rangers
Game Dec. GA SA Sv SV% QS?
1 W 1 36 35 97.2 Y
2 W 0 24 24 100.0 Y
3 L 4 30 26 86.7 N
4 L 4 34 30 88.2 N
5 W 1 25 24 96.0 Y
6 L 1 29 28 96.6 Y
7 L 5 27 22 81.5 N
Lundqvist vs Capitals
Game Dec. GA SA Sv SV% QS?
1 L 3 30 27 90.0 N
2 L 1 38 37 97.4 Y
3 W 3 31 28 90.3 N
4 W 3 30 27 90.0 N
5 L 2 35 33 94.3 Y
6 W 0 27 27 100.0 Y
7 W 0 35 35 100.0 Y

As you can see, both goalies finished with four quality starts in the seven game series, meaning that both goalies gave their teams equal opportunities to win the series. The difference in the series was that the Capitals were unable to win any games in which Holtby didn't give their team an above average chance to win the game. Such a win, where the goalie gets credited with a win even though he was below average, is called a 'cheap win'. Conversely, a loss in which a goalie loses a game in which he gave a quality start, is called a 'tough loss'.

Lundqvist had two 'cheap wins' in this series (Games 3 & 4), where he was fortunate that neither goaltender provided a quality start for their team. However, Lundqvist had two 'tough losses' (Games 2 & 5) to Holtby's one (Game 6). Two of those tough losses in this series were the result of the other goaltender pitching a 1-0 shutout win, while the third such game ended in a 2-1 Capitals victory. In those three games, quality starts were awarded to each goalie.

Finally, Holtby's bad Game 7 performance in which he allowed five goals is considered a 'blow-up' (BU) or a really bad start (RBS), since he was unable to stop even 85% of the shots he faced.

What do Quality Starts tell us?

Quality Starts tell us how often a goalie gives his team a chance to win. According to Vollman, a quality start results in a victory for their team 77.5% of the time. When one breaks it down to the other qualifiers such as cheap wins and tough losses, teams only win 32.5% of games in which their goalie does not produce a quality start, and quality starts only result in losses 22.5% of the time. Much more so than simply measuring wins and losses, quality starts tell us how effective a goalie's play has been on a consistent basis.

In the NHL, a goalie who posts a QS in over 60% of their starts is considered elite, while a QS% of under 40% is considered very poor.

What are the limitations of Quality Starts?

Quality Starts measures a series of small samples and is still subject to a lot of things that could be out of a goalie's control. If a team gives up two five on three situations in a game, or has to kill a major penalty, that will more than likely raise the probability of the goalie allowing a goal than a start in which the goalie only has to face one to three power plays on minor penalties. The effect of facing other high scoring situations such as breakaways or odd man rushes is also not considered. While shot quality has yet to be properly quantified over long samples, there is no doubt that in small samples are greatly affected by the quality of opportunities.


I'd say the major flaw in this statistic is that it is not tracked extensively, even amongst the typical hockey analytics databases. It was invented at Hockey Prospectus borrowing from a similar statistic used in baseball to evaluate pitchers and as of right now, really only exists on people's own personal spreadsheets, periodic articles by its adherents such as this piece by Vollman, and a table in Vollman's recently released book Hockey Abstract (available for purchase here). I hope to participate in making such data more readily available to people in the future.

Another criticism is that the statistic doesn't really add anything that save percentage doesn't already calculate over the long term. Inevitably, the relationship between overall save percentage and QS% (the percentage of games a player records a quality start) are predictably very closely linked.

However, QS% is definitely a better measure when comparing goalies historically, as the NHL has had a varying level of save percentages throughout it's history. This can be illustrated simply by comparing Patrick Roy's career SV% of 91.0% to that of Michael Leighton at 90.1%. The difference between the goalies was definitely a lot more than 9 goals per 1000 shots, but a simple look at their stat lines wouldn't lend itself to that being obvious.

It also will hopefully be seen as a better measure when looking at goalies playing in different leagues than simply looking at a goalie's save percentage figure, as that can also be greatly influenced by the league the goalie plays in.

Top of comments section | Top of article | Homepage