seventieslord
Student Of The Game
I dislike getting into long posts about numbers because I always feel like I’m not doing the best job of getting the point across compared to guys like Iain Fyffe, overpass, Czech Your Math and Contrarian Goaltender can talk the numbers better than I can. But since I think I have an idea that hasn’t been done yet, I’m going to try to do my best to slog through this.
OK, let’s start with CORSI. It’s a simple metric that calculates a player’s +/- in shots directed at the net. However, it is very dependent on Quality Of Competition (QoC) – A player who faces the opposition’s best players (by CORSI) will have a harder time posting a better CORSI himself, so it’s important to consider QoC when looking at his CORSI and comparing to someone else.
With that said, I think the majority of the stat community now agrees that sv% is the best way to judge a goalie’s individual performance – or, stated more explicitly, “sv% after adjusting for factors such as shot over/undercounting, shot quality, special teams, and othersâ€.
One metric that is often popular is the “comparison to backups†method because it eliminates factors such as special teams, shot quality, coaching tactics and shot over/undercounting. Goalies on the same team in the same season should have these factors somewhat wash out, meaning we can compare them more directly. And it’s generally said that most backups are of a rather uniform vanilla quality.
However, that last part, as much as we would like it to be true, isn’t. Sometimes there are great backups that should be starters on another team, some are more of a 1B goalie, some are average backups, some are sub-replacement level scrubs. Over decades, this should all wash out, but maybe it doesn’t.
I’ve seen posts that attempt to take the next step by saying “goalie A outperformed his backups by 10 points over this time, while goalie B only outperformed them his 2 points. However, goalie B had better backups because of how they performed relative to other goalies when on other teamsâ€
So, my thought was, is it possible to take this all the way? Can there be a stat where a goalie’s “sv% +/-“ versus different goalies can be turned into an adjusted sv% that more or less takes QoC into consideration?
To give a really simple example, Andre Racicot’s career line would look like this:
1991: 13 minutes, 15 sv% points below Patrick Roy, 29 points ahead of JC Bergeron, 24 points ahead of Chabot*
1992: 436 minutes, 19 points below Roy, 8 ahead of Melanson*
1993: 1433 minutes, 13 points below Roy*
1994: 500 minutes, 68 points below Roy, 27 points below Kuntar, 10 points below Tugnutt*
also, I know the “proper†way to compare save percentages is percentage wise, and not just “points above or below†– I use that for illustration and as shorthand only.
On their own, those numbers don’t mean a lot. But after this is done for all goalies since the start of sv%, the system running it would them account for the fact that Patrick Roy always outperformed his backups significantly and spit out a number representing how Racicot performed relative to normalized competition.
I think this is the same in principle to QoC for CORSI. We don’t subjectively define who the best players are, we define it by CORSI. So a CORSI metric that considers QoC would “fix†a player (call him player C) with a very high CORSI earned by playing 3rd rate competition, by seeing that these 3rd rate players always had poor CORSIs against everyone, and that this player C just happened to face 3rd rate competition a disproportionately high percentage of the time, and “fix†his score to “predict†what his CORSI would look like against average competition. It could do the same for a top checker who always matches up against powerful scoring lines.
Couldn’t a goaltending stat like that be done somewhat easily? We can’t say how good Chris Osgood’s sv% was just by looking at it, and comparing to the league average still leaves the door wide open for Detroit factors to influence it, so we compare to other goalies on his teams. But he played with Vernon, and with Howard, and they’re good, so that’s not really fair, right? So look at how Vernon did compared to other goalies on other teams. But were they good or bad? How did they do on other teams? And so on and so forth.
Unless there’s some obvious piece I’m missing, I think that this could be put together relatively easily by someone with an understanding of scripts in excel and a save% spreadsheet. I can provide the latter to anyone willing to give it a go.
I have invited the smarter stats guys to this thread in hopes that they will provide their take on this question. I’m really sorry if this doesn’t make a lot of sense, fellas.
OK, let’s start with CORSI. It’s a simple metric that calculates a player’s +/- in shots directed at the net. However, it is very dependent on Quality Of Competition (QoC) – A player who faces the opposition’s best players (by CORSI) will have a harder time posting a better CORSI himself, so it’s important to consider QoC when looking at his CORSI and comparing to someone else.
With that said, I think the majority of the stat community now agrees that sv% is the best way to judge a goalie’s individual performance – or, stated more explicitly, “sv% after adjusting for factors such as shot over/undercounting, shot quality, special teams, and othersâ€.
One metric that is often popular is the “comparison to backups†method because it eliminates factors such as special teams, shot quality, coaching tactics and shot over/undercounting. Goalies on the same team in the same season should have these factors somewhat wash out, meaning we can compare them more directly. And it’s generally said that most backups are of a rather uniform vanilla quality.
However, that last part, as much as we would like it to be true, isn’t. Sometimes there are great backups that should be starters on another team, some are more of a 1B goalie, some are average backups, some are sub-replacement level scrubs. Over decades, this should all wash out, but maybe it doesn’t.
I’ve seen posts that attempt to take the next step by saying “goalie A outperformed his backups by 10 points over this time, while goalie B only outperformed them his 2 points. However, goalie B had better backups because of how they performed relative to other goalies when on other teamsâ€
So, my thought was, is it possible to take this all the way? Can there be a stat where a goalie’s “sv% +/-“ versus different goalies can be turned into an adjusted sv% that more or less takes QoC into consideration?
To give a really simple example, Andre Racicot’s career line would look like this:
1991: 13 minutes, 15 sv% points below Patrick Roy, 29 points ahead of JC Bergeron, 24 points ahead of Chabot*
1992: 436 minutes, 19 points below Roy, 8 ahead of Melanson*
1993: 1433 minutes, 13 points below Roy*
1994: 500 minutes, 68 points below Roy, 27 points below Kuntar, 10 points below Tugnutt*
also, I know the “proper†way to compare save percentages is percentage wise, and not just “points above or below†– I use that for illustration and as shorthand only.
On their own, those numbers don’t mean a lot. But after this is done for all goalies since the start of sv%, the system running it would them account for the fact that Patrick Roy always outperformed his backups significantly and spit out a number representing how Racicot performed relative to normalized competition.
I think this is the same in principle to QoC for CORSI. We don’t subjectively define who the best players are, we define it by CORSI. So a CORSI metric that considers QoC would “fix†a player (call him player C) with a very high CORSI earned by playing 3rd rate competition, by seeing that these 3rd rate players always had poor CORSIs against everyone, and that this player C just happened to face 3rd rate competition a disproportionately high percentage of the time, and “fix†his score to “predict†what his CORSI would look like against average competition. It could do the same for a top checker who always matches up against powerful scoring lines.
Couldn’t a goaltending stat like that be done somewhat easily? We can’t say how good Chris Osgood’s sv% was just by looking at it, and comparing to the league average still leaves the door wide open for Detroit factors to influence it, so we compare to other goalies on his teams. But he played with Vernon, and with Howard, and they’re good, so that’s not really fair, right? So look at how Vernon did compared to other goalies on other teams. But were they good or bad? How did they do on other teams? And so on and so forth.
Unless there’s some obvious piece I’m missing, I think that this could be put together relatively easily by someone with an understanding of scripts in excel and a save% spreadsheet. I can provide the latter to anyone willing to give it a go.
I have invited the smarter stats guys to this thread in hopes that they will provide their take on this question. I’m really sorry if this doesn’t make a lot of sense, fellas.