# An Introduction To Adjusted Save Percentage

As Paul Campbell pointed out in a recent article here on *InGoal*, it is extremely difficult to quantify a goaltender’s statistical value to a hockey team. Difficult, but not impossible.

Many analysts suffer because they try to understand goaltenders with outdated statistics, or they simply have a lack of knowledge about how the position is played.

The truth is, traditional statistical evaluations of goaltenders have been extremely flawed since the day the game began. Goals-against-average and wins were the first major categories in the early years. In reality, goaltenders have very little control over those numbers. It’s a bitter pill to swallow for many people that grew up idolizing those traditional stats, but they offer little-to-no insight about an individual goaltender’s performance.

Eventually, sometime in the 80s, save percentage rose to the forefront and became the most common goaltending statistic. Analysts were finally using a stat that actually reflected a goaltender’s performence. Unfortunately, research has proven that it is also flawed. With traditional “unadjusted” save percentage, every shot is weighted equally. A dump-in from the neutral zone that reaches the goal is worth the same amount as a one-timer from the slot. Wouldn’t it be great to fix that?

Welcome to the world of adjusted save percentage.

### Why is Adjusted Save Percentage better?

Adjusted save percentage is a step up from traditional save percentage because it takes shot location into account. Since goalies have almost no control over where the shots they face come from, they are no longer unfairly punished because they play on a team that gives up more shots closer to the net. Alternatively, a goaltender that faces more shots from a distance does not get rewarded.

While shot location does *not* tell the entire story, shots from each location on the ice do come with some level of predictability. War-on-Ice has defined three different zones, each with different levels of expected shooting percentage.

In the blue or “high-danger” zones, each shot has a 10%+ chance of going into the net.

In the red or “medium-danger” zones, each shot has a 3-10% chance of going into the net.

In the yellow or “low-danger” zones, each shot has a 3% or below chance of going into the net.

It is a simplistic, but effective proxy for shot quality, in the same way that “Corsi” and “Fenwick” are proxies for puck possession. *How* *effective* it is as a proxy is still up for debate. Regardless, it is still much more effective than looking at traditional statistics.

Adjusted save percentage is expressed as a 5 on 5 statistic, so penalty killing and powerplay chances are excluded. Save percentage generally drops between 4-5% while on the penalty kill, across the board. It would be unfair to penalize goaltenders that play for undisciplined teams – so it is ruled out.

There is also the myth that shot quantity has an effect on save percentage. Traditional save percentage can be fooled if a goalie is on a team that gives up a lot of shots, but they are mostly from the outside in “low-danger” zones. That was proven in a previous article, and adjusted save percentage fixes that problem by taking shot location into account.

It can also be combined with the “Goals Saved Above Average” statistic that has been discussed here on *InGoal* before. Stephen Burtch of Sportsnet and Nick Mercadante of Blueshirt Banter have both recently applied it to the goaltending numbers from this past season.

##### With adjusted save percentage, this goal…

##### …would have a much larger negative effect on Corey Crawford’s adjusted save percentage than his traditional save percentage because the shot came from a low-danger area.

##### And this save…

##### …would have a much larger positive impact on Kari Lehtonen’s adjusted save percentage than his traditional save percentage because the shot came from directly in front of the net.

### Does this fix all of the issues with save percentage?

No, as Stephen Valiquette’s “royal road” and Chris Boyle’s “shot quality project” will tell you. While theoretically it is better to give up a shot from the blue line than in the middle of the slot – there are many other factors that go into the difficulty of a save for a goaltender.

All shots were not created equal, but we can at least account for *some* of the difference with adjusted save percentage. Its’ effectiveness is limited, but this is the only way to do that based on the data that is publicly collected at the moment.

Even with the advancements in the adjusted save percentage stat, goaltenders are still prone to have one fantastic year – then fall off the map the very next season. Predicting goaltenders from year-to-year is still a nightmare. When attempting to gauge their value in one season, adjusted save percentage is the way to go.

It is possible to use adjusted save percentage to judge goaltenders from different eras, but it requires an extra step. You have to equalize it compared to what the league average save percentage was that particular year, then multiply it by 100 – similar to baseball’s ERA+ statistic. Broad Street Hockey and Canucks Army both show examples of how that can be useful.

### What does a good adjusted save percentage look like?

Great question! During 5 on 5 play, the league average save percentage was .923 last season – so that’s the number that goaltenders should be shooting for. (Thanks to @DTMAboutHeart for calculating those numbers for this article)

Some goalies, like Anton Khudobin in 2014-2015, see an increase in their adjusted save percentage compared to traditional. He went from a .903 unadjusted save percentage, to .910 adjusted – which shows that he was likely a victim of his team giving up many high-danger opportunities.

Other goalies, like Darcy Kuemper in 2014-2015, see their save percentage decrease after it is adjusted. His unadjusted save percentage was .907, and it went down to .898 after being adjusted. That means, if anything, he likely benefited from his team’s defensive play.

(NOTE: This section has been edited to fix an error in the data from War-on-Ice, which has since been corrected.)

### Where can I find adjusted save percentage stats?

War-on-Ice collects all of the data, and it you can find it on their website here.

In conclusion, there is still a lot of work to do when it comes to statistically evaluating goaltenders – but adjusted save percentage is the best we have right now because it at least accounts for shot location. The next step will be electronically tracking players to collect more specific data about each shot that is taken during an NHL game. For now, this is it.

With the information available at our fingertips today, the only thing lazier than writing goaltenders off as “unquantifiable” would be resorting to traditional statistics like goals-against-average and wins.

Let’s take an unscientific, statistically corrupt poll- Who cares?

Yawn. People who get paid to evaluate goaltenders, and literally any fan who has ever wanted to learn about the statistical value that a goaltender brings to the table. As a goaltender and a stat head, finding that value is important to me.

How exactly do you calculate it? What’s the formula. If I missed it I apologize.

I don’t have the formula off hand, but if you contact War-on-Ice, they will give it to you. All of the data they collect is released to the public.

Hey Sleepy- As a Quality Engineer, I get to do stats all day long. This garbage is stuffing ten kilos of crap in a five kilo bag. You may enjoy it but I think you should poll your readers and see who really has any interest in this. You also need to ask goalies if they care about the ASP. If I am wrong on this this, I will happily fold my tent and stop pointing out the stats folks have no clothes. Are you up for asking the masses what they want to read about?

I work as an RMO and I thought this was a well-researched article of interest. If you don’t like it you could just disregard it.

I don’t have to disregard it. I believe it is still okay in North America to have an opposing viewpoint. In my opinion, advanced stats as applied to hockey are an answer to a question nobody asked. There is no value in applying statistical analysis to a process as random and out-of-control as ice hockey. Carey Price is Carey Price no matter what his ASP number is. I’ll assume you also buy in to the Head Trajectory mania (which no one can seem to explain in 25 words or less).

You complain that these are not topics our readers want to see…then say “I don’t have to disregard it”….so you want to see it then? Confused.

If you don’t want to read it, you certainly don’t have to. Before you slam it as garbage please write a well-reasoned response. “I do stats all day long” and “hockey is random” and “Dubnyk will regress to the mean” are not points in a reasoned argument based on any evidence. At least looking at these stats is a solid attempt to find evidence. And to suggest advanced stats have no place in hockey is a bit strange, IMHO. Until teams can do a more predictable job of scouting and developing goaltenders, we need better ways to evaluate their performance. Otherwise oyu are buying into the “goaltending is voodoo” nonscence.

We will agree to disagree. The worn out cliche ” If you don’t like it don’t read it” is as lazy a response as is possible to give.You should try some argument in place of cliche. My oft stated opinion that hockey is a series of random events is pretty solid. A bounce off a skate or the glass can completely change the outcome of a game or a series. No one has responded to my statement that Carey Price would still be Price no matter what kind of analysis is applied. If you understood basic statistics would would also understand “Dubnyk regressing to the mean” is inevitable. He played over his talent level last season, and “Head Trajectory” not withstanding, he will have an average (for him) or below -average (for him) season soon. One last point, your pages are filled with articles about young goalies destined to be the next big thing. I’ll make an assumption that these goalies were touted based on statistics. Many of those goalies will never see a minute of NHL playing time.

Great article, will share.

Very good article. Seems this and other advancements will become more common with better tracking tech for players and puck. In a cap era, teams can’t afford overpays based on the eye test. As a Jets fan, Pavelec jumps to mind. Wonder how his .920 bounceback season looks through the adjusted save percentage lens.

Paul if you don’t care about it then why bother reading it. This is such a crucial field in evaluating goaltending. Especially when it comes to scouting, when coaches have to look at the masses’ stats and decide who is worth going to watch in person. The best goalie in the league could easily be on the worst team in the league. To think this is pointless crap is ignorant.

I think this was a great article. As a beer league goalie who started really digging into goaltending a bit late by most people’s standard, understanding the professional game more methodically is awesome. I’d like to see more stuff like this, and even take this a bit further by giving a how to guide on reading the various graphs and tabular data. I don’t understand all this argument about uselessness of stats, baseball statisticians have been doing it forever and it has proven results (see the money ball craze going on and how it’s paying dividends in the MLB). Nice job, Mr. Balloch, thanks for bringing this info to the masses!

Statistics in, and by themselves, are just numbers. They must always be placed in a certain context to be of value. If you do not regularly follow a goaltender ( and a specific team) the value of statistical analysis is limited since the data must always be placed in a certain context. Without context, you make incorrect assumptions based on individual biases. If you follow a goalie (and a specific team) on a regular basis you have the insight to place the data in the correct context. This will allow the collectors, and reviewers of the data, to make proper assumption about tendencies, trends and areas of concern. That is always the way to give numbers a meaning. Statistics are important.

My son just had a game where he faced 0 shots against. How do you calculate the save % if that happens? 0% would affect his stats. 100% doesn’t seem accurate either. I understand I’m crazy to keep stats on my son and his counterpart goalie. For me it’s fun and a tool for the coaches. Anyways, if anyone could help me with this one I’d appreciate it.

There are two different responses here. The technical answer is that if there are no shots, there is no save percentage. You can’t save a shot that doesn’t occur. The other response is that for young players save percentage is beside the point. The idea is to have fun and to learn skills appropriate to their age and skill level. You could probably just let it go.

Mike:

While Clare is right about the save percentage for that game being undefined (not 0), I believe the remainder of her response is incorrect. I’ve coached young goalies (Novice to Jr C, boys and now girls) for many years and we always use stats as a way to understand how we are progressing as a goal tender. We spend a tremendous amount of time on skills and tactics, but at then end of the day our goal is to stop the puck. When used properly stats can help your goalie feel a sense of accomplishment, they can help the coach and goalie understand where to focus their efforts and provide valuable feedback to both.

As for how the undefined save percentage works into season stats…You calculate the season SV% by dividing the total number of saves (from all games) by the total number of shots (from all games), not by averaging all games sv%. Make sense?