Kukla's Korner

The Puck Stops Here

David Backes Has A Top Adjusted +/- Rating

As I continue my sabermetrics and hockey posts, I am looking at the number two adjusted +/- last year.  David Backes of the St Louis Blues accomplished this with a +32 rating on a team with a +1.4 baseline. 

Backes had a good season last year.  He led the Blues in scoring with 62 points.  He was the Blues representative in the NHL All Star Game.  I think he should be a Selke Trophy nominee.  Some people take this too far and suggest he is Hart Trophy candidate.

Backes is a good example of why Corsi ratings are better than +/- because of the statistical noise involved in +/- ratings.  At even strength, when Backes was on the ice his opponent`s goaltending had a .883 saves percentage.  This is unsustainably bad.  It is not a repeatable skill of David Backes to have his opponent`s goaltending perform at a level so poor that they would not maintain an NHL job.  This is a fluke.  More goals were scored by the Blues with Backes on the ice than would have been with merely adequate goaltending.

While Backes was on the ice, the Blues had a .920 saves percentage, which is quite good.  With Backes on the ice, St Louis scored at a much higher rate than would be expected but did not allow them at that rate.  Thus counting shots is more meaningful than counting goals.  This makes a Corsi rating more useful than a +/- rating. 

David Backes had a good season last year.  His +/- shows that it was good.  The unsustainably poor goaltending that played against Backes magnifies this and makes it look better than it otherwise should be.  This is a case study that shows Corsi is a better indicator than +/- because it has less statistical noise.

Filed in: | The Puck Stops Here | Permalink
 

Comments

Avatar

I’ll agree that 0.883 is an unsustainably low save percentage but players can drive opposing goalies save percentages down.  For example, over the past 4 seasons when Gaborik has been on the ice 5v5 opposing goalies have a 0.888 save percentage.  Crosby’s 4 year opposing goalies have a 0.891 save percentage.  Daniel Sedin’s is 0.896.  The problem with Corsi is it does not take into account that players can drive shooting and save percentages.  You cannot conclude that corsi is a better stat just because it doesn’t rate Backes so highly.  The list of top corsi players is littered with players you wouldn’t consider worthy of consideration for top line duty (or even second line duty in some cases) on an average team let alone consideration for MVP.

But the question I have when we are handing out awards is, do we hand them out to the most talented player regardless of statistical results on the ice or the player who had the best results on the ice, regardless of talent.  I don’t think Backes is the most talented player, but he did some of the best on ice results this past season.  Conversely, I think Ovechkin is one of the most talented players in the NHL, but he didn’t have the best on ice results this past season.  Should Ovechkin get consideration for MVP?

Or, do we just give the MVP to a player we think is pretty good and put up pretty good numbers - Daniel Sedin (my pick since he performed well in +/- and corsi) or Corey Perry (just because he scored 50g but a poor choice IMO since he wasn’t among the leaders in +/- and had a weak corsi rating).

Posted by HockeyAnalysis on 06/21/11 at 05:04 PM ET

J.J. from Kansas's avatar

If Gaborik could sustain an opposing save percentage of 0.888 over a 232-game span, then why Backes’ considered unsustainably low?

There are three of the better players in the league who have sustained save percentages against that would get an NHL goalie fired over a very good sample size. 

I think it’s irresponsible to say it is not a repeatable skill when it seems to be repeating for good players.

Who has more data on this phenomenon?  I’m afraid there is inadequate statistical analysis here to draw any conclusions about sustainability or nonsustainability of a player’s ability to drive his opponents’ save percentages.

Alex Semin has a 263-game stretch where his opponents’ save percentage has been 0.898

So how much difference are we looking at in David Backes’ season is we change the saves percentage to 0.9?

Posted by J.J. from Kansas on 06/21/11 at 05:19 PM ET

PuckStopsHere's avatar

While it is true that some of the best players can reduce the opposing team`s saves percentage or defensively increase their own team`s saves percentage, there is a lot of statistical noise in the random variations in saves percentages.  On the individual season level, this randomness heavily degrades any signal.  This is a big part of the reason Corsi is more sustained from season to season and better selects top players than +/-

But the question I have when we are handing out awards is, do we hand them out to the most talented player regardless of statistical results on the ice or the player who had the best results on the ice, regardless of talent.  I don’t think Backes is the most talented player, but he did some of the best on ice results this past season.  Conversely, I think Ovechkin is one of the most talented players in the NHL, but he didn’t have the best on ice results this past season.  Should Ovechkin get consideration for MVP?

This question has been discussed here in the League of Extraordinary Statisticians and I htink has the wrong premise.

David Backes was not the one of the best players in the NHL last year.  That statement doesn`t really need to be backed up if you actually watched hockey last year. Only a poor statistical anaylsis can possibly come up with that conclusion. Alexander Ovechkin wasn`t the best player either.  He may have been the most talented player but he struggled through injury.  That said, he was a far bigger contributor to his team than David Backes was.

I try to pick MVPs by asking the question which player produced the most wins for his team this season.  While I don`t have a complete mathematical theory to answer who that player might be, it does provide a framework about how to think about this problem.  Last season, I think Tim Thomas was that player.  He wasn`t nomimated for the Hart Trophy.  Of the nominees, I think Daniel Sedin should win.

Your problem is that you are trying to interpret Corsi or +/- or similar statistics as a single number that tells you everything you need to know to rank players.  That is false.  That is what makes a bad sabermetrician.  Corsi is one of several numbers that put together can give you a full picture of what happened in a season.  Point totals are another valuable number, but one you seem to overlook in your comment. 

While I have hope that a Corsi related analysis might eventually develop into such a system, I recognize that we do not know how to do this (and that it may not be possible).  In the meantime I hope that it adds to the statistical record and better quantifies the game.  You think that it represents the entire game.

Posted by PuckStopsHere on 06/21/11 at 05:40 PM ET

PuckStopsHere's avatar

How do I know that opponent`s saves percentage with Backes on the ice is unsustainably low?  The only player with a lower opponent`s shooting percentage in the top 20 adjusted +/- ratings is Toni Lydman (who is the only player ranked ahead of Backes).  Do you believe that a defensive defenceman has a bigger effect on saves percentages than anyone else?

In 2009/10, Backes had an oppenent`s saves percentage of .914.  For the record Lydman`s was .923 last year.  It is unsustainable.  Should time permit this summer I will try to quantify the repetitive portion of individual players contributions to opposing team`s saves percentages.

Posted by PuckStopsHere on 06/21/11 at 05:52 PM ET

J.J. from Kansas's avatar

In 2009/10, Backes had an oppenent`s saves percentage of .914.  For the record Lydman`s was .923 last year.  It is unsustainable.  Should time permit this summer I will try to quantify the repetitive portion of individual players contributions to opposing team`s saves percentages.

Posted by PuckStopsHere on 06/21/11 at 03:52 PM ET

I look forward to that.  Also to the explanation of why Backes couldn’t be considered to be entering his prime, considering his age.  Or are Henrik Sedin’s point totals still unsustainably high?  I’d say the opponent’s save percentage is unlikely to be sustained, but it is not impossible, especially considering that lower-than average saves percentages for several players have indicated that they may have a good measure of sustainability.

Posted by J.J. from Kansas on 06/21/11 at 06:36 PM ET

PuckStopsHere's avatar

Last season 9 players had opponents saves percentages better than .890 while playing 50 or more games.  They were Ryan Getzlaf, Bobby Ryan, Toni Lydman, Lubomir Visnovsky, David Backes, Adam McQuaid, Alex Tanguay, Lauri Korpikoski and Brenden Morrow.  The first four are Anaheim players who got the saves percentage while playing together. 

How did those nine do in 2009/10?
Getzlaf .891, Ryan .896, Lydman .923, Visnvsky .918, Backes .914, Tanguay .911, Korpikoski .940, Morrow .922. McQuaid only played 19 games .954

This is somewhat consistent with the idea that oppoent`s saves percentages are random.  The best players in one season are all over the place the year before.  Perhaps a case can be made for Ryan Getzlaf (he did .908 the year before so its not so strong) having some sustained skill, but its nearly lost in the randomness..

I think the numbers David is quoting are different from the ones I am (assuming his are correct).  My numbers are even strength only.  David`s I suspect include power play and penalty kill time.  Opponent`s saves percentages on the power play are significantly worse than at even strength and penalty kill is significantly worse than this.  This is caused by the better quality of shots on the power play and the usually bad quality of shots on the penalty kill.  I suspect the players he is saying have sustained high saves percentages for years have them because they are players who play significant power play time and essentially no penalty kill time.  This will make their opponent`s saves percentage lower than it would be if it only included even strength as my numbers do.

Posted by PuckStopsHere on 06/21/11 at 07:05 PM ET

J.J. from Kansas's avatar

Just curious, but what’s the Corsi differences for the same players over the same span?

Posted by J.J. from Kansas on 06/21/11 at 07:09 PM ET

Avatar

This is a big part of the reason Corsi is more sustained from season to season and better selects top players than +/-

That’s utter nonsense.  Was Peter Regin anywhere close to being a top player in the NHL last season?

David Backes was not the one of the best players in the NHL last year.  That statement doesn`t really need to be backed up if you actually watched hockey last year.

The ‘if only you watched the game you would know’ argument is such a cop out.  If watching 50, 80, 100, 200 games told you all you needed to know about the game of hockey, why are you going through this whole exercise of statistical analysis?  If adjusted +/- is such a crappy stat, why are you even discussing it here?

Your problem is that you are trying to interpret Corsi or +/- or similar statistics as a single number that tells you everything you need to know to rank players.  That is false.  That is what makes a bad sabermetrician.  Corsi is one of several numbers that put together can give you a full picture of what happened in a season.

For the most part that is a fair statement, but you can’t pick and choose.  You can’t say “David Backes had the top adjusted +/- but that doesn’t make sense so in this case corsi is the better statistic” while simultaneously saying ‘Peter Regin had a very good corsi rating in 2009-10 which makes no sense so <insert another stat of your choice here> is the better statistic in this case.’  Either adjusted +/- has some value, or it doesn’t.  You can’t arbitrarily apply it when you think it makes sense and when you think it doesn’t.  That’ll only introduce bias into the evaluation which is the worst thing we can do.

Point totals are another valuable number, but one you seem to overlook in your comment.

Point totals really only tell you half of the game, the offensive half, and they predominantly influenced by ice time, and in particular PP ice time.  Point totals really are a crappy way of evaluating players.  They can give you an indication of who the best offensive players are in the league, but don’t give you a very refined player evaluation. 

Personally, I don’t think you can truly evaluate a player from one season of data regardless of what statistic, or statistics you use.  My preference is to use 2 or better yet 3 years of data and focus on goal stats (goal for/against rates, adjusted for quality of teammates and opposition).

Posted by HockeyAnalysis on 06/21/11 at 07:09 PM ET

PuckStopsHere's avatar

David

You are setting up a strawman here.  Again you are trying to represent Corsi as a single number which is a be all and end all of hockey analysis.  In order to argue against it you must take the 17th highest raw Corsi and argue that this player isn’t as good as that number shows.  It shows that you have to look as deep in the rankings to get a player to use as your argument.

You are right that Peter Regin was nowhere near the 17th best player in hockey last year.  He played against weak competition with high quality linemates in 2009/10.  Adjusting for those factors, Regin’s Corsi is much more reasonable.

Posted by PuckStopsHere on 06/21/11 at 07:23 PM ET

Avatar

Tanguay’s 4 year opponent 5v5 save percentage is 0.893 and Getzlaf’s is 0.897.  Morrow is .901.  Backes is .907.  Visnovsky is .908.  Lydman is .912.  Korpikoski is .923.

The 4 year league wide average save percentage 0.921 so all of those players, save for Korpikoski, forced opposing goalies into worse than average save percentages while they were on the ice.

For Backes, he has forced opposing goalies to have worse than average save percentages in three of the last 4 seasons and the other year he was about average.

2007-08:  .913
2008-09:  .903
2009-10:  .921
2010-11:  .893

And for Getzlaf:

2007-08: .893
2008-09: .895
2009-10: .909
2010-11: .893

The biggest anomaly for Getzlaf was 2009-10 when he still was substantially better than average.

Posted by HockeyAnalysis on 06/21/11 at 07:36 PM ET

PuckStopsHere's avatar

There is some talent involved in reducing the saves percentage of your opponents but it is largely lost in randomness.  All the players with top adjusted +/- ratings had unsustainably low opponent’s saves percentages this year.  This can be correct using Corsi.

I think your argument that the players who had low opponent’s saves percentages while they were on the ice having low saves percentages over a four year period comes down to this.  Given that a player has one of the lowest opponent’s saves percentages in the league this season, it is quite likely that they have one over four years because we know one of the four years is low.  It is a lot like saying that if we roll dice and get a low value in one roll odds are the total value of four rolls will also be low - although not as low as the first roll.

Posted by PuckStopsHere on 06/21/11 at 07:44 PM ET

J.J. from Kansas's avatar

That’s a good argument that doesn’t take into consideration that the numbers posted show that Getzlaf has all four years with a below-average saves percentage against him.

I wouldn’t say that the talent is lost in randomness here.  If a player has a below-average saves percentage in only one of four years, it stands to reason that true randomness would assign him an above-average percentage in one of the following three years and about average for the next two years, which would only correct his number to the mean over what becomes a fairly large sample size.

There are a handful of players who have either been incredibly lucky of a four-year period or should be considered players with the kind of talent in or around them to positively influence the outcome.  Since the entirety of hockey’s history tells stories of players who are better than their contemporaries at making goalies and defenses miss pucks, I feel confident that it is not luck which is driving this outcome.

This is one of my problems with Corsi - it does not try to account for these things and simply calls them all anomalies.  Burying those things involves talking over what the numbers are telling you.  That is a very fitting descriptor of the concept of statistical “noise”.

Also, four dice rolls are a very small sample size - 260 NHL games is not.

Posted by J.J. from Kansas on 06/21/11 at 07:58 PM ET

Avatar

There is some talent involved in reducing the saves percentage of your opponents but it is largely lost in randomness.  All the players with top adjusted +/- ratings had unsustainably low opponent’s saves percentages this year.  This can be correct using Corsi.

But the problem with Corsi is it assumes players can’t influence shooting/save percentages, which they can.  Backes’ opposing goalie save percentage was unsustainably low last season, but it is just as wrong to assume that his true on ice shooting percentage is league average 0.921 but that is what corsi assumes.  As I said, both methods have their flaws and it is my belief both are probably fairly equally flawed over the course of a single season with goal metrics having an edge when dealing with more than a years worth of data and corsi having an edge with less than a years worth of data.  For true player evaluation, lets look at goal metrics over 3 years.

Posted by HockeyAnalysis on 06/21/11 at 08:08 PM ET

PuckStopsHere's avatar

But the problem with Corsi is it assumes players can’t influence shooting/save percentages, which they can

That is false.  This is your assumption to argue with a strawman version of Corsi analysis.  To be meaningful, +/- or Corsi must be adjusted for multiple factors.  One of which would be the effect of a player’s shooting percentage or ability to reduce his teams saves percentage.  This can be calculated to some degree. 

This is much easier to do after the fact with Corsi because it is quite random.  Players routinely have .03 and .04 jumps from year to year in their opponent’s or team’s saves percentage while they are on the ice.  That is huge.  That is the difference between an outstanding saves percentage and a bad one.  It adds a lot of noise to +/- ratings.  This noise is not present in a Corsi rating.  Instead of making an adjustment to a noise filled number (+/-) we can make a similar adjustment to a more noise free number (Corsi).  Thus is it a better starting point to get a meaningful number and not be lost in statistical noise.

Posted by PuckStopsHere on 06/21/11 at 09:30 PM ET

Avatar

That is false.  This is your assumption to argue with a strawman version of Corsi analysis.  To be meaningful, +/- or Corsi must be adjusted for multiple factors.  One of which would be the effect of a player’s shooting percentage or ability to reduce his teams saves percentage.  This can be calculated to some degree.

Not, according to you, with a single season of data you can’t as you’ll come up against the same sample size issues as with goal analysis (for example, +/- analysis).

Posted by HockeyAnalysis on 06/21/11 at 09:36 PM ET

PuckStopsHere's avatar

The significant source of noise we are discussing is unsustainable randomness in saves and shooting percentages.  In order to best avoid this, you use the numbers that are not affected (Corsi) instead of the ones that are significantly affected (+/-) as a starting point.

Posted by PuckStopsHere on 06/21/11 at 09:39 PM ET

statelouis26's avatar

So who’s this Backes guy? lol.  I think if backes didn’t take so many bad penalties his cause would be helped.  Just because a guy has a great +/- doesn’t mean he’s a good two-way forward.  It simply could mean he’s on the ice for a lot of goals.  Do you know if he backchecks at all?  What are his takeaway numbers.  I think the perfect example of how this stat is manipulated is when you look at players like Green, Wideman, Crosby, and Stamkos have had great +/- records before.  (I don’t mention Ovechkin b/c he is a winger and typically they are not asked to backcheck compared to the center).  Look at the past Selke winners (Datsyuk) or guys that are nominated often (Kesler).  They are known by their effort.

Posted by statelouis26 from Detroit, MI on 06/21/11 at 09:45 PM ET

Avatar

The significant source of noise we are discussing is unsustainable randomness in saves and shooting percentages.  In order to best avoid this, you use the numbers that are not affected (Corsi) instead of the ones that are significantly affected (+/-) as a starting point.

Good offensive players don’t score goals at a high rate predominantly because they take shots at a high rate they score goals at a high rate predominantly because they produce good shooting percentages.  I’ll say it again, given enough data using goal scoring rates as the basis for evaluating players is better than using corsi as the basis.

Corsi and whatever variants of it you want to use is a flawed player evaluation tool because it doesn’t take into account a players ability to drive/suppress shooting percentages.  Goal scoring rates (or adjusted +/-) is a better player evaluation tool because it takes into account shot generation and shooting percentage but suffers from sample size issues.  If we eliminate the sample size issues by using a sufficiently large dataset, then goal scoring rates (or correctly adjusted +/-, etc) are the better player evaluation tool.  So what is sufficiently large?  Two or more years and goals based evaluation is better than corsi based, 1 year probably about the same - both somewhat flawed, less than 1 year stick to corsi analysis, but conclusions have to be taken with a grain of salt because corsi is flawed too.

Posted by HockeyAnalysis on 06/21/11 at 10:22 PM ET

PuckStopsHere's avatar

Two or more years and goals based evaluation is better than corsi based, 1 year probably about the same - both somewhat flawed, less than 1 year stick to corsi analysis, but conclusions have to be taken with a grain of salt because corsi is flawed too.

These statements are false.  We showed that in yesterday’s comments, yet you state it again.  Corsi is more repeatable, selects a better group of players and better differentiates them.

Either +/- or Corsi has to be corrected for other things to produce meaningful results, but I would much rather start with the more repeatable stat that selects a better group of players with better differentiation between them.  You wouldn’t.  I suspect the reason for this is not logical but rather because you have been in enough arguments on this topic with a large enough set of people that you are unwilling to admit that you have been wrong.

Posted by PuckStopsHere on 06/21/11 at 10:27 PM ET

Avatar

Corsi is more repeatable, selects a better group of players and better differentiates them.

Corsi is more repeatable, but PIMS are repeatable too but that doesn’t make PIMS a good player evaluation stat.

As for selecting a better group of players, you looked at a list of 20 players, eyeballed it and came to that conclusion.  That is hardly proof.  Truth is, over the long term, corsi doesn’t match shooting percentage at predicting goal scoring ability.  See the second table in http://hockeyanalysis.com/2011/05/30/goal-rates-better-than-corsifenwick-in-player-evaluation/  The list of offensive players identified by fenwick shooting percentage is a much better list of offensive players than the list identified by fenwick rate.

Posted by HockeyAnalysis on 06/21/11 at 10:35 PM ET

PuckStopsHere's avatar

David

You have gone way off topic now.  the biggest problem with your analysis is that it required four years of data, which is too long an amount of time tp be useful in most cases.

At any rate, you are merely showing that scoring goals is related to having a good shooting percentage.  The problem is having a good shooting percentage is not very repeatable. 

It is like showing that winning is blackjack is related to having your cards add up to 21.  While true, it doesn’t say anything.  There is little reproducable skill in getting your cards to add up to 21.  The repeatable skill is elsewhere in the game.  Corsi better quantifies the other skills (+/- does too) but Corsi is more repeatable,

Comparing Corsi to penalty minutes is disingenous.  You know it is more related to winning hockey than penalty minutes,  If you require me to show you that, I don’t think you have ever understood a hockey game.

Posted by PuckStopsHere on 06/21/11 at 10:50 PM ET

Avatar

You have gone way off topic now.  the biggest problem with your analysis is that it required four years of data, which is too long an amount of time tp be useful in most cases.

No, I don’t need 4 years of data for goal rate to be a better evaluator of talent in terms of contribution to winning than corsi.  That’s the point.  In about 1 year, and certainly with 2 years of data goal metrics have enough sample size to perform better than corsi.

Corsi is more sustainable yes.  If I wanted to predict next years corsi, using this years corsi is what we should use.  But, if I want to predict next years goal scoring rate, corsi isn’t that great of a predictor.  For example:

FF20 = Fenwick For per 20 minutes (fenwick = shots + missed shots) while on ice 5v5.
GF20 = Goals For per 20 minutes while on ice 5v5.

The correlation of GF20 to year+1 GF20 is r^2=0.203
The correlation of FF20 to year+1 GF20 is r^2=0.204

That is for players with min. 500 minutes in each of the last 4 seasons.

If we correlate 2 years with the following 2 years we get:
2007-09 GF20 vs 2009-11 GF20:  r^2 = 0.389
2007-09 FF20 vs 2009-11 GF20:  r^2 = 0.248

So, as we move beyond a single season of data, GF20 becomes a better predictor of future goal scoring.

So, GF20 is a reproducible skill.  That’s the point.  And the more data we get, the more trust we should have in GF20 being a reliable stat.

Now, if you want to predict future corsi, then sure, go ahead and use corsi as it will be the best predictor, but corsi isn’t the primary skill in scoring goals.

Posted by HockeyAnalysis on 06/22/11 at 12:12 AM ET

PuckStopsHere's avatar

Once again, you are stuck on goals.  Goals and possession stats (+/-, Corsi or whatever) are complimentary pieces of information.  We want to get a well rounded view of what is going on in hockey and possession nmber are valuable as are goals. 

My claim all along has been that as far as possession numbers go Corsi is better than +/-.  And i guess now (since we have gone sufficiently far off topic) possession numbers provide information that goal numbers do not.

Posted by PuckStopsHere on 06/22/11 at 12:23 AM ET

Avatar

I am stuck on goals because I believe that goals are what matters in hockey.  You win games by out scoring your opponent, not by out corsi-ing them so naturally my focus is on goals.  Corsi is just a component of scoring goals.  Think of corsi as scoring opportunities and shooting/save % as conversion rates which results in this equation:

Goal Rate = Corsi Rate * Corsi Shooting %

So when I do my goal rate analysis, I am factoring in corsi rate as well so I am not ignoring corsi, I am just using it to the extent it is useful.

You are right when you say Corsi has less noise than Shooting %.  That is because corsi encompasses approximately 10x the events which smooths out the noise.  But, and this is important, given enough data the noise in shooting % dissipates as well to the point where shooting % becomes useful and at that point we should (or must if our goal is to correctly evaluate players) use it.  I believe that point is in around the 1 year mark but it will vary depending on the players ice time and opportunity/goal distribution.

So, with that said, getting back to your original post when you say “This makes a Corsi rating more useful than a +/- rating” I say, no, not always.

Posted by HockeyAnalysis on 06/22/11 at 02:07 AM ET

PuckStopsHere's avatar

We disagree on this point But, and this is important, given enough data the noise in shooting % dissipates as well to the point where shooting % becomes useful and at that point we should (or must if our goal is to correctly evaluate players) use it.  I believe that point is in around the 1 year mark but it will vary depending on the players ice time and opportunity/goal distribution.

It is quite obvious that this point is not on the one year level.  The fact that players with top Corsis tend to have more top Corsis in future seasons more often than top +/- rating players do.  The reason is huge random variations in saves percentages from year to year. 

We have already discussed this point.

Posted by PuckStopsHere on 06/22/11 at 02:46 AM ET

Avatar

The fact that players with top Corsis tend to have more top Corsis in future seasons more often than top +/- rating players do.

Again, yes, corsi is more persistent on a year over year level and if I wanted to predict future corsi, present corsi is the best tool.  I agree with that.

But, I am not trying to predict future corsi, I am trying to evaluate players and I am trying to evaluate players at producing/stopping goals since producing/stopping goals is what ultimately matters in hockey, not producing/stopping shot attempts.

At one year, GF20 predicts next seasons GF20 as well as FF20 predicts next years FF20 (see a few comments ago).  At 2 years GF20 is the clear leader.  Conclusion:  At 1 year GF20 and FF20 have the same reliability at predicting next years GF20 and thus, with 1 year of data GF20 and FF20 are equally good (or equally poor actually) at identifying talent.  With more than 1 year of data GF20 is the far better predictor of future GF20 and thus with 20 or more years of data is the far better evaluator of ability to drive goal scoring.

We have already discussed this point.

Yes, we have, and you have chosen to ignore what I have said.

Posted by HockeyAnalysis on 06/22/11 at 09:29 AM ET

Add a Comment

Please limit embedded image or media size to 575 pixels wide.

Add your own avatar by joining Kukla's Korner, or logging in and uploading one in your member control panel.

Captchas bug you? Join KK or log in and you won't have to bother.

Smileys

Notify me of follow-up comments?

Feed

Most Recent Blog Posts

About The Puck Stops Here

imageThe Puck Stops Here was founded during the 2004/05 lockout as a place to rant about hockey. The original site contains over 1000 posts, some of which were also published on FoxSports.com.

Who am I? A diehard hockey fan.

Why am I blogging? I want to.

Why are you reading it? ???

Email: y2kfhl@hotmail.com