Jun 302011
 

Anyone who knows anything about hockey, save for Florida GM Dale Tallon it seems, immediately thought ‘bad contract’ when they heard that the Florida Panthers had signed Tomas Kopecky to a 4 year contract at $3M/year.  But how bad is that contract?  Well, lets take a look.

Goal based stats (i.e. any stats that requite goals to calculate which is pretty much everything except things like shots or corsi) are heavily influenced by random events over the short term but over the long term tell a much more accurate picture than shot or corsi based stats.  Personally I consider goal based stats over a 3 or 4 year period to be a fairly reliable indicator of a players talent level so lets see how Kopecky compares over the past 4 seasons.

There are 183 NHL forwards who have played over 3000 minutes of 5v5 ice time over the past 4 seasons and Tomas Kopecky is one of them.  Here is how Kopecky compares to his forward peers over that time.

Statistic Value Rank out of 183
Goals For/20 minutes 0.719 150
Goals Against/20 minutes 0.798 106
Goals For % 0.474 157
Opposition GF/20min 0.737 183
Opposition GA/20min 0.760 71
Opposition Goals For % 0.492 183

To summarize, his performance numbers are bad and he has arguably played against the easiest competition of any forward with 3000 minutes of ice time over the past 4 seasons.  On top of that he has never played any significant time on the penalty kill.  Last season he got increased ice time on the Blackhawks top 2 lines and on the PP which boosted his offensive numbers to a career high of 15g and 42 points (previous high was 10g, 21pts) but that is more a result of who he was playing with and not his own talent level.

Kopecky for the most part has been a third line player who got some top six minutes last season because of the post-Stanley Cup salary cap induced fire sale that left the Blackhawks short of experienced forwards. For the past 2 seasons Kopecky earned $1.2M and I don’t think he deserved a raise from that level.  To pay Kopecky $3M over 4 years is probably a $1.5-2M/yr over payment and probably 1-2 years longer than he deserved.  It might be the worst $3M/yr contract ever signed in the NHL but he may be in good company soon because I suspect Maxime Talbot will get a similar contract and his numbers are equally bad, if not worse (though Talbot has played significantly more on the PK than Kopecky).  I pray that the Leafs do not sign Talbot.

  15 Responses to “How Bad is the Tomas Kopecky Contract?”

  1.  

    Goal based stats (i.e. any stats that requite goals to calculate which is pretty much everything except things like shots or corsi) are heavily influenced by random events over the short term but over the long term tell a much more accurate picture than shot or corsi based stats.

    This is false and you should know it. As two examples that show failure of your methods, you picked Toni Lydman as the second best defenceman in the NHL this season and David Backes as a Hart Trophy candidate. Those conclusions are just stupid. The fact you keep claiming it to be true seems to be out of your own obstinance.

    In this case it doesn’t matter because no statistic will tel you Kopecky is playing well.

    •  

      The team that scores the most goals wins the game, not the team that attempts the most shots. Shot rates have a relatively poor correlation with goal scoring rates and that is a fact. Deal with it. Goal scoring rates do have a level of persistence given enough data. Deal with that too. These are facts.

      As for Lydman and Backes, I never made the claim that they were the most talented players in the NHL, just that they had the best results when they were on the ice this past season. You cannot deny that. You picked Visnovsky as the best defenseman but no one on the planet would claim he is the most talented either.

      If I performed a similar analysis for corsi I’d have concluded the top 5 forwards this past season might have been Elias, Kesler, Steen, Clowe and Justin Williams which is clearly bogus. The MVP Corey Perry turns out to have pretty crappy corsi numbers too.

      •  

        As for Lydman and Backes, I never made the claim that they were the most talented players in the NHL, just that they had the best results when they were on the ice this past season. You cannot deny that. You picked Visnovsky as the best defenseman but no one on the planet would claim he is the most talented either.

        20 voters thought Visnovsky was the best defenceman this year. Not one thought Lydman was. 20 voters thought Visnovsky was second best as well. Not one thought Lydman was. 26 voters thought Visnovsky was the third best defenceman this year. Not one thought Lydman was. 28 voters thought Visnovsky was fourth best. Not one thought Lydman was.

        One stathead who did his stats wrong thought Lydman was the second best defenceman in the league. That was you. It was out of touch with reality. Picking Visnovsky as the best defenceman is not at all comparable. 20 voters agreed. It is a reasonable position to hold. This is an obstinant person trying to throw anything at the wall to try to find an argument that he isn’t wrong.

        If I performed a similar analysis for corsi I’d have concluded the top 5 forwards this past season might have been Elias, Kesler, Steen, Clowe and Justin Williams which is clearly bogus. The MVP Corey Perry turns out to have pretty crappy corsi numbers too.

        If *YOU* did the Corsi analysis I have no doubt that you would get it wrong and pick players like you named. If somebody who wasn’t trying to create a strawman by doing the Corsi analysis as half-assed as possible in an effort to show it doesn’t work did it, they would find much more logical results. I don’t think they would get Corey Perry as the best player last year because he was a poor MVP choice, but they would get a far better group of players than the ones you listed.

        •  

          20 voters thought Visnovsky was the best defenceman this year. Not one thought Lydman was. 20 voters thought Visnovsky was second best as well. Not one thought Lydman was. 26 voters thought Visnovsky was the third best defenceman this year. Not one thought Lydman was. 28 voters thought Visnovsky was fourth best. Not one thought Lydman was.

          Those voters didn’t vote for Visnovsky because he had a great corsi. They voted for him because he led all defensemen in points and was second in goals. They voted for him because he put up great offensive numbers. Of course, offense is only half the game but voters don’t always take the other half into consideration. You can’t use people voting for Visnovsky as a defense for using Corsi and not goal based analysis.

          •  

            Voters voted for Visnovsky for any number of reasons. Some may have done it because of Corsi and some didn’t. The important point here is that players with good Corsi ratings are good players. Good players can be identified by any number of different statstical analyses. Good players for example score points and get good Corsi’s.

            Your complete failure as a statistician is to believe your numbers in place of common sense. You believe that Toni Lydman is the second best defenceman in the NHL because of your numbers and nothing else. Worse, you expect that is what everyone else does. So when a player like Visnovsky shows that he is a good player by multiple statistical indicators, you throw out Corsi since the other analyses showed it. Therefore you don’t need Corsi to show he is a good player, so for some stupid reason you discard his Corsi entirely.

            A good statistician would see a player who has nothing exceptional about him except for the statistical indicator of choice that the statisitician uses (Toni Lydman and David Backes are two examples in your case) would view them with suspician that perhaps their statistical indicator is invalid in their case. You view your statistical indicator as reality and further argue that another statistical indicator cannot be right if it is not treated in that way by others – if a player has a high Corsi and a high point total you discard the Corsi and claim it is a selection entirely on point total. The more logical reason is that players with good Corsis also usually score points.

  2.  

    You are sitting here arguing that goal based analysis works better than Corsi based anaylsis when your own goal based analysis made ridiculous picks. You picked Toni Lydman as the second best defenceman in the league last year and you picked David Backes as an MVP candidate. The only reason you picked them is because of statistical flukes in the saves percentages with them on the ice. You made your conclusions based on statstical noise and that is why they are wrong. At the same time here you are arguing that your method isn’t wrong and overcomes the statistical noise. These are two examples that clearly show that claim is wrong.

    •  

      The only reason you picked them is because of statistical flukes in the saves percentages with them on the ice.

      Visnovsky had an 11.8% shooting percentage which is ridiculously high for a defenseman (compared to 3% for Lydman) and his teams shooting percentage while on the ice was 10.53, only marginally less than Lydman’s 10.91. If you are going to argue that Lydman’s stats are statistical flukes, I’ll argue the same for Visnovsky. Their on-ice corsi ratings were almost identical.

      You made your conclusions based on statstical noise and that is why they are wrong.

      I could argue the same for your Visnovsky pick. Also, using Corsi you are basing your conclusions on a stat that is only loosely correlated with scoring/stopping goals.

      •  

        Visnovsky had an 11.8% shooting percentage which is ridiculously high for a defenseman (compared to 3% for Lydman) and his teams shooting percentage while on the ice was 10.53, only marginally less than Lydman’s 10.91. If you are going to argue that Lydman’s stats are statistical flukes, I’ll argue the same for Visnovsky. Their on-ice corsi ratings were almost identical.

        Davod you know better than arguing this. We have already argued well beyond this point. Here are the top adjusted +/- ratings and the team shooting percentages for and against when they are on the ice.

        Lydman despite a pathetic shooting percentage of 3% had his team shoot 11.7% with him on the ice. That isn’t Lydman’s skill. It isn’t even repeatable. The year before in Buffalo his team had a very poor shooting percentage. The team’s saves percentage was .927 with Lydman on the ice. As for Visnovsky, his numbers are also rather good (after all he and Lydman were on the ice at the same time often in 5 on 5 situations). Visnovsky has an 11.2% team shooting percentage and a .913 team saves percentage. All players with top +/- ratings shjow similar patterns. It is necessary to have a higher team shooting percentage than saves percentage. All players with high +/- ratings have this.

        Shooting percentages (saves percentages) arte largely unrepeatable. The biggest repeatable part is the situation in which a player is used. Players used in offensive situations tend to have higher shooting percentages than those used defensively. That is because defensive players do not tend to take high quality shots as they are not trying to get into situations to get them. Similar arguments can be used on team saves percentages as well.

        I think a large portion of your “signal” where you are using shooting percentage to predict goals is that the situation a player plays in dictates his shooting percentage to a large degree. Players tend to play in the same situations from year to year, so that portion is repeatable. You will see that is you bother to read my blog this summer.

        As I said above, there are multiple indicators that show Visnovsky had an exceptional season. Corsi correlates well with the others. No indicator besides +/- ones show Lydman has an exceptional season. That makes +/- look questionable and it makes Corsi look stronger.

  3.  

    Therefore you don’t need Corsi to show he is a good player, so for some stupid reason you discard his Corsi entirely.

    I discard corsi not because it doesn’t tell me what I want but because it isn’t highly correlated with scoring goals, which to me is the most important thing in the NHL. I have said this 1000 times but you don’t seem to get it. When it comes to scoring goals, shooting percentage matters more than getting shots. Look at a list of top corsi for rate players and look at a list of top shooting percentage players over multiple seasons and it will become clear that shooting percentages do a far better job at identifying offensive talent.

    Also remember that Lydman isn’t a one year wonder. If you analyzed his performance he has probably been a top 20 defenseman (when both offense and defense are considered) over the past 4 years while playing against very tough competition. He is an extremely underrated defenseman who is had an extremely good season. This is what statistical analysis is all about. We don’t look for statistics to confirm what we think we already know, we do statistical analysis to tell us things we didn’t know. How good Lydman is is something most people don’t know. How good Backes is becoming is also something people don’t necessarily know.

    Visnovsky has an 11.2% team shooting percentage and a .913 team saves percentage. All players with top ± ratings shjow similar patterns. It is necessary to have a higher team shooting percentage than saves percentage. All players with high ± ratings have this.

    Yes, it is necessary to have a higher team shooting percentage than save percentage. Why? Because shooting percentages drive goal scoring more than shot rate. The best way to score more goals than your opponent is to be a better shooter than your opponent. Look at the list of top forwards in on-ice shooting percentage over the past 4 seasons (http://stats.hockeyanalysis.com/ratings.php?disp=1&db=200711&sit=5v5&pos=forwards&minutes=3000&type=goal&sort=ShPct&sortdir=DESC) and you’ll see a list of the most talented offensive players. If you look at a list of the top corsi for rates (http://stats.hockeyanalysis.com/ratings.php?disp=1&db=200711&sit=5v5&pos=forwards&minutes=3000&type=corsi&sort=F20&sortdir=DESC) and you find a list that includes a ton of Red Wing players, several Capital players and several Leafs including Grabovski, Ponikarovsky, Blake, Stajan, none of whom would be considered top offensive players. That is why I don’t use corsi. It’s crappy at identifying top players.

  4.  

    I discard corsi not because it doesn’t tell me what I want but because it isn’t highly correlated with scoring goals, which to me is the most important thing in the NHL. I have said this 1000 times but you don’t seem to get it. When it comes to scoring goals, shooting percentage matters more than getting shots. Look at a list of top corsi for rate players and look at a list of top shooting percentage players over multiple seasons and it will become clear that shooting percentages do a far better job at identifying offensive talent.

    A large part of their “repeatable” shooting percentage comes from playing in offensive situations where they get good shots. That is the repeatable part. They are used in offensive situations. It is a bigger repeatable part than their actual talent that goes into a high shooting percentage. In short you are looking at noisy data and the c.”correlation” you think you have found is due to the fact that scoring is related to having a high shooting percentage. Anyone with no goals has a shooting percentage of zero. Anyone with lots of goals must have a relatively good shooting percentage because nobody gets thousands of shots a year. You found a correlation between two variables that are by definition linked. That is a sign of a poor statistician.

    •  

      Anyone with no goals has a shooting percentage of zero. Anyone with lots of goals must have a relatively good shooting percentage because nobody gets thousands of shots a year.

      Use 4 years of data and use on ice shooting percentage and use players with >2500 minutes of ice time. Under those conditions no one is close to having zero goal scored while they are on the ice. Also, I am not correlating with goals scored, I am correlating with goal scoring rates (i.e. goals per 20 minutes ice time). Players who take the most shots do score the most goals, but mostly because they are the players who are on the ice the most. But when we look at the correlation between shots per 20 minutes and goals per 20 minutes as well as shooting percentage and goals per 20 minutes, the shooting percentage vs goals per 20 minutes has a much higher correlation. There is no link like you describe between shooting percentage and goals per 20 minutes.

      We have a very simple equation.

      Goals per 20 minutes = Shots per 20 minutes * Shooting percentage

      Now the question is, is goals per 20 minutes more driven shots per 20 minutes or shooting percentage. The answer is shooting percentage.

  5.  

    Nothing you have stated breaks the trap that goals and shooting percentage are linked by definition.

    What you have said is that you now need four years of data, which shows why your method is a failure. If we have a player he has to play regularly in the NHL for four years before you claim you can say anything meaningful sbout him. If you only use one year of data you get stupid results. You claim Toni Lydman is the second best defenceman in hockey and David Backes is a Hart Trophy candidate. In reality all you are doing is reporting on statistical noise.

  6.  

    Nothing you have stated breaks the trap that goals and shooting percentage are linked by definition.

    There are two components to scoring goals. Shots and shooting percentage. One is no more or less linked ‘by definition’ to goals than the other.

    What you have said is that you now need four years of data, which shows why your method is a failure.

    Actually, that is not what I said. I just used 4 years as an example. You don’t need 4 years of data for goal based stats to be better than just shot or corsi based stats. 2 years is definitely enough, and with just 1 year of data the value of goal based analysis is about as good as corsi based analysis, especially for players who get a lot of ice time. Both are still fairly significantly flawed, but they are close to equally flawed at 1 year of data. To be honest, I wouldn’t have much faith in any statistical evaluation of a player using any currently known technique using just 1 season of data. It is just too flawed. One year of data we can probably separate good players from bad players but probably not reliably separate good players from average players.

    You claim Toni Lydman is the second best defenceman in hockey and David Backes is a Hart Trophy candidate. In reality all you are doing is reporting on statistical noise.

    Maybe, but I am sure that a significant number, if not most, of NHL award winners won on some sort of statistical noise. I never claimed that either of those guys are the most talented, I am just claiming they had among the best statistical seasons in the NHL, above and beyond just individual offensive statistics which is what award voting is usually based on (Corey Perry won based on statistical noise that resulted in him getting 50 goals). I don’t expect Lydman to have as good a set of statistics next season as this season and I doubt Backes will either but then it is probably safe to assume Corey Perry won’t score 50 goals nor will, I think your pick for Norris, Visnovsky score 18. If I was building a team from scratch and wanted the best team, none of those guys would be among my top picks (None of Perry, Visnovsky or Lydman would be my top pick from the Ducks, Getzlaf would). But that doesn’t diminish the great statistical seasons they had in 2010-11.

  7.  

    You make a distinction between a great season and a great statistical season. I think that is because you know that your +/- analysis says little about reality and you know Toni Lydman and David backes did not have great seasons – they had great statistical seasons by the way you chose to look at statistics which is not too well correlated with reality.

  8.  

    You make a distinction between a great season and a great statistical season.

    You are playing semantics. I am conducting a statistical analysis and by all accounts Lydman had a great season. Either you trust what the statistics say, or you don’t. If you don’t, then why are you bothering me and why are you discussing statistics on your blog. If you do, then you have to accept what they are saying and what they say is Lydman and Backes had very good seasons at both ends of the ice and excellent seasons overall.

Sorry, the comment form is closed at this time.