Goalie reliability stats

[TweetyLeaf]

Looking for stats of goalie reliability.

I'm interested in who are the goalies who are the most consistent.
Have fewest bad games.

I would love if it would be tied to save percentage.
So the lower the % in single game, the more it penalizes the stat.
AVG % would be all nhl goalies in single season.
If its lower, goalie get penalized, if its higher goalie stat gets boosted.

Thing is, the lower 1 game sv% goes it would penalize it more heavily.

Something like this or similar has to exist, please tell me where?

Click to read more...

 

[Bear of Bad News]

I do a "Goaltender Variability" calculation on http://hockeygoalies.org that I describe in detail here:

http://hfboards.mandatory.com/showthread.php?t=1392449

I'd start with post #42, since the posts prior to that are not the final version (that assumes that the current version is the final version, of course ).

It's basically a measure of consistency. Currently calculated for the NHL between 1984-85 and 2012-13, although I'm close on 1982-83 and 1983-84.

 

[TweetyLeaf]

BAV and VAR helps me in this case somewhat.

Appreciate the effort with those stats, smart that they are side by side keep them that way

Still, i need this season running data also :/

 

[PenguinMario]

Looking for stats of goalie reliability.

I'm interested in who are the goalies who are the most consistent.
Have fewest bad games.

I would love if it would be tied to save percentage.
So the lower the % in single game, the more it penalizes the stat.
AVG % would be all nhl goalies in single season.
If its lower, goalie get penalized, if its higher goalie stat gets boosted.

Thing is, the lower 1 game sv% goes it would penalize it more heavily.

Something like this or similar has to exist, please tell me where?

This might be time intensive, but you could plug the goalie's game logs into Excel and run the standard deviation function for save%. It's going to be limited by the flaws of save percentage itself, of course, but the results could be suggestive.

 

[Bear of Bad News]

This might be time intensive, but you could plug the goalie's game logs into Excel and run the standard deviation function for save%. It's going to be limited by the flaws of save percentage itself, of course, but the results could be suggestive.

Other than the flaws with save percentage (which I agree with 100%), two other concerns:

(1) This wouldn't take into account the fact that the same goaltender playing at the same level would be expected to do better against some teams than others.

(2) This would weight each game equally, which may be appropriate, although in my method (linked to above) I weight the games by shots faced.

Having said that, I don't have results for seasons in progress, which would be an advantage to doing it this way.

 

[TweetyLeaf]

The reason im searching for these stats is pretty simple (apart this being interesting).

I want to create a shortlist of goaltenders who are great number 2-3 goalies for great fantasy goalie. I have Rask number 1 in my long time cap enforced keeper league that has drafts and farm systems etc.

So, i want to research if there is a goalie to support Rask's great games and with fewest possible chance for bad games, who is not getting premium money for his effort.
That's why i liked the BVA stat.
For the basic goalie stats SV% is the one i heavily trust, because i have been following it quite long and im comfortable in my ability to "read it" and compare it correctly.

In fantasy, its a shame to see a good goalie getting constantly ruined by, not a bad goalie, but unreliable in his performance. One bad game for the week can lose the stats for you.
So im looking for a goalie shortlist who are opposite of "box of chocolate", but as said, are not the "top 10" goalies of this days NHL.

 

[Hammer Time]

Hockey Prospectus' Quality Start stat is similar to what you're looking for. It defines each game as either "good" (which means the goalie either a) gave up 2 or fewer goals or b) had a sv% above league average) or "not good".

A goalie with a 70% QS means that seven out of every 10 games is a "good" game.

 

[Bear of Bad News]

Hockey Prospectus' Quality Start stat is similar to what you're looking for. It defines each game as either "good" (which means the goalie either a) gave up 2 or fewer goals or b) had a sv% above league average) or "not good".

A goalie with a 70% QS means that seven out of every 10 games is a "good" game.

It's a fine metric, although - as I mention in the thread that I link to above - it suffers from the same two problems I describe above; it assumes that all opponents are equally hard to play against, and it weights all games equally (no matter how quickly a goaltender is removed).

 

[wgknestrick]

A team's PK ability and their tendency for penalties can influence their goaltender's overall SV%. This is a source of significant variation to the base SV%. Starting with the least penalized teams is a good start in eliminating the noise/variation.

5v5 SV% average is right around 91% (+-2% for the population), while 4v5 average SV% is around 88%(+-3% for the population). 3% difference is a huge amount of noise to eliminate in SV%.

I also give much more credit to the team in front of the goaltender for their SV% than most people though. Example BOS/PHX/VAN (good) or EDM/NYI/CGY (bad).

 

[PenguinMario]

A team's PK ability and their tendency for penalties can influence their goaltender's overall SV%. This is a source of significant variation to the base SV%. Starting with the least penalized teams is a good start in eliminating the noise/variation.

5v5 SV% average is right around 91% (+-2% for the population), while 4v5 average SV% is around 88%(+-3% for the population). 3% difference is a huge amount of noise to eliminate in SV%.

I also give much more credit to the team in front of the goaltender for their SV% than most people though. Example BOS/PHX/VAN (good) or EDM/NYI/CGY (bad).

I'm going to have to look it up, but I think that team defense has a much greater effect on shots on goal than save percentage, where it was small but nonzero.

That's even strength only, of course, as you say - special teams is another matter entirely.

 

[Bear of Bad News]

One question is which statistics are relevant for goaltenders in the league that you're developing this for (and how are they weighted)?

 

[TweetyLeaf]

One question is which statistics are relevant for goaltenders in the league that you're developing this for (and how are they weighted)?

If the question was directed to me these are the stats:
W
GAA
SV
SV%
SHO

But player cap hit is extremely important in this thing.

I think, at the moment, my best bet is to look for these: SV%, QS%, BAV and VA

Comparing those together and getting a familiarity with them so i know what is good and what is bad is the way to find the Goalie.
I did find one interesting one, and i send PM to you. (Dont want to release it here, because there might be ppl reading this who play my league).

 

[Bear of Bad News]

I liked the one that you sent to me - I think he's a good fit.

The original question was directed at you, but more specifically at anyone developing fantasy hockey applications. The trick is to compare "value above replacement level", divided by "cost above replacement level".

Both of those are tricky to estimate, but it's a significant step just to get to the point above.

 

[TweetyLeaf]

Sadly im not good enough with excel to do this on my own.

But here's an idea i think could be developed to show real goalie reliability.

This is based purely on SV% because for me, that's the best basic stat indicator of goalie performance. Even tho it's not a good indicator to reliability by itself.

Every game a goalie plays on or over 91% i consider a okay/good game.

So every game goalie plays that % or over it he gets awarded +1 point.

Every time goalie plays 90-90.9% he gets 0 points.

and this goes on like this:

89 - 89.9 = -1
88 - 88.9 = -2
87 - 87.9 = -3
86 - 86.9 = -4
85 - 85.9 = -5
84 - 84.9 = -6
83 - 83.9 = -7
82 - 82.9 = -8
81 - 81.9 = -9
80 - 80.9 = -10

All below 80 are considered -10 because i don't think one very bad game should be penalized more than -10, it would be simply too hard for the goalie to make up for that game and it doesn't serve the purpose of the stat.

Again, all save% over 91 just award 1 point, so for one 80.9% or under game a goalie needs to play 10 over 91% games to make up for it.

The only thing that's bugging me with this is that the difference between 85% game and 84.9% game is a bit too much. Then again, i don't see reason enough to do this on 0.5% intervals because it complicates it a bit too much.

So, i would love if someone would find the time to help me creating this excel sheet so i could just insert 1 goalies game log sv% into it and it would give me the stat.

 

[hatterson]

Sadly im not good enough with excel to do this on my own.

But here's an idea i think could be developed to show real goalie reliability.

This is based purely on SV% because for me, that's the best basic stat indicator of goalie performance. Even tho it's not a good indicator to reliability by itself.

Every game a goalie plays on or over 91% i consider a okay/good game.

So every game goalie plays that % or over it he gets awarded +1 point.

Every time goalie plays 90-90.9% he gets 0 points.

and this goes on like this:

89 - 89.9 = -1
88 - 88.9 = -2
87 - 87.9 = -3
86 - 86.9 = -4
85 - 85.9 = -5
84 - 84.9 = -6
83 - 83.9 = -7
82 - 82.9 = -8
81 - 81.9 = -9
80 - 80.9 = -10

All below 80 are considered -10 because i don't think one very bad game should be penalized more than -10, it would be simply too hard for the goalie to make up for that game and it doesn't serve the purpose of the stat.

Again, all save% over 91 just award 1 point, so for one 80.9% or under game a goalie needs to play 10 over 91% games to make up for it.

The only thing that's bugging me with this is that the difference between 85% game and 84.9% game is a bit too much. Then again, i don't see reason enough to do this on 0.5% intervals because it complicates it a bit too much.

So, i would love if someone would find the time to help me creating this excel sheet so i could just insert 1 goalies game log sv% into it and it would give me the stat.

If you can give me the format of the game logs to are using (NHL.com, ESPN.com, elsewhere, etc.) I could easily show you how to do the calculations.

In regards to the metric itself, I have a few issues:

1.) I think it too harshly punishes poor performances and doesn't reward great ones enough. A performance of 21 saves on 23 shots is worth +1 but a performance of 50 saves on 51 shots is also only with +1. In addition a goalie who has 10 straight performances of 50 saves on 51 shots and then has one bad game where he lets in 3 goals on 10 shots and is pulled would still clearly be the best goalie in the game, but in this metric would only be even.

On a 30 shot game the difference between letting in 0 goals and 2 goals is nothing. The difference between 0-2 and 3 is 1 point and the difference between letting in 3 and letting in 4 is 4 points.

2.) It does not differentiate between PP saves and ES saves. It is harder to make PP saves by a statistically significant margin, especially 5 on 3 saves. Did a goalie really play inconsistent if his team does something like the Nucks did the other night/week and earns a 7 minutes 5 on 3.

3.) It doesn't take the circumstances of the game into account. Carey Price's game against the Sens a few days back was a great game. He's the only reason the Habs made it to OT, really the only reason they didn't get blown out something like 8-4, but because he let in 4 goals on 44 shots he would receive a 0 for the game.

 

[TweetyLeaf]

If you can give me the format of the game logs to are using (NHL.com, ESPN.com, elsewhere, etc.) I could easily show you how to do the calculations.

In regards to the metric itself, I have a few issues:

I would like to use NHL.com game log data. Like this one:
http://www.nhl.com/ice/player.htm?id=8471695&season=20132014&view=log

I would just copy the stat table to excel and clear the other data. So the sv% would be in 0.910 for example.

1.) I think it too harshly punishes poor performances and doesn't reward great ones enough. A performance of 21 saves on 23 shots is worth +1 but a performance of 50 saves on 51 shots is also only with +1. In addition a goalie who has 10 straight performances of 50 saves on 51 shots and then has one bad game where he lets in 3 goals on 10 shots and is pulled would still clearly be the best goalie in the game, but in this metric would only be even.

Yes, and the metric is there to show who are the goalies who play "best bad games".
It's purpose is not to measure how good goalie is in itself but more of who are the goalies who do not fail badly and play consistently okay, even when going trough bad stretches.

For good goalie measurements theres lots of other stats that need to be looked at. Like shutdowns and SV% and GAA and wins.

2.) It does not differentiate between PP saves and ES saves. It is harder to make PP saves by a statistically significant margin, especially 5 on 3 saves. Did a goalie really play inconsistent if his team does something like the Nucks did the other night/week and earns a 7 minutes 5 on 3.

This is something that i should think about, but for now, i think it would be good to do one test run without to keep it simpler.
Maybe even create different stat that takes those into account. Would be interesting how much different those are between different goalies.

3.) It doesn't take the circumstances of the game into account. Carey Price's game against the Sens a few days back was a great game. He's the only reason the Habs made it to OT, really the only reason they didn't get blown out something like 8-4, but because he let in 4 goals on 44 shots he would receive a 0 for the game.

Yes, but with this stat, getting 0 is not bad. So Price played like expected (was good).

I don't know, but would guess, that no goalie will end up into + with this stat after full season. But thats not the point. The point is how far into minus you go.
So don't think this like +/- stat

Appreciate any help i can get and really liked your thoughts on this!

 

[TweetyLeaf]

I've been thinking this a bit more.

Maybe more accurate system would be to really count those SV%'s between the league average and that 80%

There has to be a limit where the minus points stop adding up, otherwise one game where goalie gets pulled with 60% SV is gonna hurt too much.
However, i'm having hard time thinking appropriate % that would actually relate to something and is not something i throw out of my hat.
But for now, i settle to that 80% mark.

So i was thinking about taking the league average that is currently 91.4%

That would be the upper cap.
The lower cap would be 80%

So everytime a goalie plays for example 0.850 SV% The resulting Stat would be 0.914-0.850= 0.064

If goalie's SV% is higher or same than 0.914, for example 0.950 the resulting stat would be
0.914-0.950 = 0

Because the stat is not supposed to account good games, they are just "successfull" games, and result in not negative stat (in other words, 0).

Other "limit" would be that 0.800 limit, if goalie plays worse than that, it only counts as 0.800 performance. For example if goalie plays 0.750 game:

0.914-0.750 = 0.114 (Because the limit, you cant get worse stat than 0.114)

Can someone please help me to do excel/open office calc spreadsheet with this ?
I only need the "code", i can copypaste and stuff the stats in there.

 

[TweetyLeaf]

Problem solved, i have my stats now! Thanks for all who helped.

 

[TheDebaser]

Well the statistical way to do this would be to employ standard deviation.

 

[Bear of Bad News]

Well the statistical way to do this would be to employ standard deviation.

That's exactly what I do - see above.