Wednesday, October 22nd, 2014

Moneypuck Again

36

James Mirtle has a series of pieces in the Globe on the new hockey analytics: “The Moneypuck Revolution“, “Why Haven’t Advanced Stats caught on in the NHL“, and “Hockey’s New Numbers“. I think he was very clever to use the movie “Moneyball” to pitch a series he’s wanted to do. I am also happy to hear that Gabe Desjardins is making some money advising some NHL teams. I am sure he is worth it.

James knows me well enough to know that I’m offside on this issue. His series prodded me to address the issue in some depth. (Fair warning – this isn’t going to be short.) I even went over to the Sloan Institute that he referenced and watched the panel discussion on the subject. (It was an interesting, if inconclusive discussion.)

For the record, I’m not against people looking at any of this stuff. I played with it myself for quite a while in the 1980′s and then again in the late 1990′s. (When I was working for About.com, Dean Oliver was the basketball guy and we had several discussions about the subject.) The data has come miles since those days, but none of it solves the fundamental problem.

Hockey numbers are not useless. We can learn things with the statistics. We have already learned that hockey is heavily influenced by randomness, a fact that puts lie to most of what passes for hockey punditry. In fact, that’s the best argument for hockey analytics – the numbers are real, if often misleading, and lots of hockey narratives are fantasy. Hockey analytics are increasing in popularity at least partly because the usual alternative – media mythmaking – is worse. Stats can be used to skewer the ridiculous claims made by teams and the reporters who follow them. (See Tyler Dellow.) Stats can also help identify when a player (or team) result has been heavily influenced by luck. Finally, I’m intrigued by statistical comparables. (At age 22, Joe has played so many games and done this and that. Here is a list of all the players who did that well that young. The worst of them is pretty damn good. Therefore Joe will probably be pretty damn good. I like statistical comparables because a lot of my subjective evaluations are based on human comparables. Ryan Kesler is Bob Gainey with more offense. Jonathan Toews has a young Steve Yzerman skillset. Sidney Crosby is Peter Forsberg plus, and so on.)

Hockey analytics are not useless.

But.

Hockey statistics will never do what we want them to do, which is to effectively evaluate individual hockey players. To give us answers when considering a trade or a personnel decision. To tell us whether this third line winger creates more wins than that number four defenseman. Baseball statistics can produce answers to these kinds of questions, while hockey statistics can only produce more questions.

Why? Because baseball statistics describe what actually happens in baseball games. The stats add up to runs and to wins. Hockey statistics do not add up to goals. Goals or proxies for goals like shots, shots and attempted shots or even quality shots, underpin all the analyses. None of these statistics say anything about how the scoring chance, the goal was achieved. The actual activities that go into creating the chance are not recorded. The fundamental problem is that hockey has the equivilent of runs, but the hits and walks that create those runs are missing from the statistical package.

Offence in hockey involves moving the puck through open ice towards the opponent’s goal. Finding open ice in the offensive end usually results in a scoring chance. Defense in hockey is about eliminating open ice and forcing the offense to give up the puck. The team that best controls the ice surface – by skating faster, by being disciplined, by filling the passing lanes and pressuring the puck, by winning the puck battles, by moving the puck – will get the most scoring chances.

In the NHL teams are close enough in quality to ensure that neither team can control the ice for the entire game. When control shifts, momentum appears to swing. Suddenly, the opponent is finding the open ice and forcing a chip and chase game on the formerly dominant team. Sometimes neither team can find any room. Back and forth it will go, but the team that carries most of the play usually wins. The top players – the ones who can do the most with open ice – lead the way. When a team can’t find open ice, their top players are often invisible.

Individual players do not – cannot – control the ice. It requires a team effort to deny ice to opponents, to cut off passing lanes and to pressure the puck. The entire team has to transition instantly. While a single player might occasionally go end to end, moving the puck down the ice is also usually a collective effort. This quintessentially team activity – finding or creating open ice with the puck, denying that open ice to the opponent without it – can’t be measured. We can see it – if questioned even a non-fan could identify which team is carrying the play – but we can’t count it.

In fact, it pollutes all of the individual numbers.

When Henrik Sedin is asked why he and his brother suddenly jumped to elite status at the age of 29, he says, “We’re the same players playing the same way we always have. The team is a lot better and so we are getting more chances and better chances.” If this is true – and I believe that it is – we have no way to tell how much credit to give Sedin for his numbers and how much credit belongs to the team. Ironically, one of the new statistics actually supports this hypothesis:

4. ZONE START

What is it?

The percentage of faceoffs players are on the ice for in the offensive zone. Neutral zone faceoffs are not counted.

What does it tell you?

How coaches use their different line combinations. Vancouver’s Sedin twins led the NHL in zone start last season, as they were on the ice for almost three times as many offensive zone draws as defensive zone ones.

Teammate Manny Malhotra, meanwhile, was in the opposite role, taking mostly defensive draws, which has a negative effect on plus-minus.

The most important thing this number tells us that the position of the puck when a player comes onto the ice has a very significant impact on his chance to score on the shift. This stat measures faceoffs, but the same dynamic must also be true for changes made on the fly. If Henrik comes onto the ice just as a teammate is moving up the ice with the puck, he is likely to do much better on that shift than if he steps onto the ice just as a Canuck turns it over.

The zone start stat reminds us that what happens on one shift has an impact on what happens in the next one. As the Canucks improved as a team from being good to being very good at controlling the ice, a significantly larger percentage of the Sedin shifts started with the puck in a favourable position for them. The result is that the same players are doing more scoring and have better underlying numbers.

Henrik Sedin’s numbers – traditional and new – do tell us that he is a very good player, but they can’t be more precise than that because we can’t tell whether he gets more help from his teammates than a star from another team gets from his. One of the reasons Jason Spezza’s scoring rate has dropped is because Ottawa as a team is no longer great at controlling the ice surface. A lot more of his shifts start with Spezza skating towards his own goal these days. That makes it a lot harder for him to score.

With lesser players – the real challenge in objective evaluation – the individual numbers mean even less because different players contribute in entirely different ways. There are reliable defensive players who seldom make a mistake, but provide almost no offence. There are other defensemen who can make a good pass and score once in a while, but make too many mistakes without the puck. There are forwards who can score but do little to help the team move the puck into scoring position. There are forwards who can win puck battles, kill penalties and even move the puck, but their hands are stone. Goaltenders don’t do anything except guard the net.

It is the mixture of all the skills, the collective skillset, that produces team strengths and team strengths win hockey games. Team speed. Team toughnesss. Team goaltending. Team defense. Team offense.

We can’t objectively sort out the individual contributions to those team strengths. Until we can find the hockey equivilent of singles, doubles and triples from the organized chaos of the game, the statistical evaluation of individual hockey players with disparate skills is a mug’s game.

Be Sociable, Share!

Comments

36 Responses to “Moneypuck Again”
  1. Magicpie says:

    In the defense of all the statistical analysis guys I’d say that I don’t think any of them claim that these methods can come up with that magic number that quantifies exactly how many wins each player contributes to the team. (I’m actually glad they don’t. When we finally figure out how to calculate that number hockey will become infinitely less interesting)

    As you said though, that doesn’t mean they’re useless or that we shouldn’t bother with them. They may not solve the problem but they still move things in the right direction. When you take into account that the traditional way of evaluating players statistically basically boils down to “how many points does this guy have,” having a few more advanced stats to look at doesn’t hurt.

    • Tom says:

      In the defense of all the statistical analysis guys I’d say that I don’t think any of them claim that these methods can come up with that magic number that quantifies exactly how many wins each player contributes to the team.

      I don’t think any of them would dare make this claim. But unless they can, the entire concept of Moneyball is out the window. The objective is to use statistics to find players who are undervalued or overvalued. Get good players cheap or unload expensive, but unworthy ones according to the numbers.

      None of the hockey statistics are useful for this. None of them will ever be able to do this.

      They may not solve the problem but they still move things in the right direction.

      They aren’t going in either the right direction or the wrong direction. There is nowhere to go.

      When you take into account that the traditional way of evaluating players statistically basically boils down to “how many points does this guy have,” having a few more advanced stats to look at doesn’t hurt.

      The traditional way is to acknowledge that points don’t tell us much about most players and look at the skills the player actually has. How fast does he skate? How strong is he? Does he see the ice well? Good hands? Good shot? Mistake prone? Quick stick? Decision making? Can he score? If he does not score, what does he do to help a team? Is he likely to improve? How do his skills fit on our team? Will he be better or worse on our team? Will he make us better or worse?

      • James Mirtle says:

        Talking to Gabe, I’m pretty sure he is able to make these stats do, to some extent, what you’re talking about. It simply takes more advanced analysis than many are capable of doing or willing to do.

        Probably including you and me.

        • Tom says:

          Talking to Gabe, I’m pretty sure he is able to make these stats do, to some extent, what you’re talking about. It simply takes more advanced analysis than many are capable of doing or willing to do.

          Well, I’m from Missouri. If Gabe is making this claim, it’s up to him to show it. I may not be willing to do the work to produce the stat, but I’m more than willing to critique the analysis and the methodology. Let’s see it.

          • James Mirtle says:

            So when you say “I am sure he is worth it” what the heck does that mean if you also don’t believe he can’t properly evaluate players?

          • Tom says:

            He can probably prevent some managers from making stupid mistakes. If Kevin Lowe had asked Tyler Dellow whether it was wise to sign Khabibulan, he might have learned that no other goaltender in hockey history snapped back well enough at age 35 to be worth his contract. If Kevin Lowe asked Gabe about contracts, he might stop signing guys to contracts that assumed their career year was sustainable performance.

            In other words, see paragraph four in the post. He can say – and support it with numbers – that player X is better than his traditional stats or player Y is not as good as his traditional stats appear to make him.

          • Hawerchuk says:

            Tom,

            I think we need to keep our official sports stats history straight, right?

            4000 BC – 1998: nothing in baseball can be quantified statistically
            1998 – 2002: ok, so hitting can, but fielding and pitching can’t
            2002 – 2008: ok, so baseball can be quantified, but that’s because it’s stationary; a fluid team game like basketball can’t be quantified
            2008 – present: ok, so basketball can be quantified, but hockey is somehow different, so it can’t be quantified

            We very much can put a value on player performance. As with all player valuation, regardless of sport, it has error bars on it. So be it.

            In fact, your statement “[h]e can probably prevent some managers from making stupid mistakes” contradicts your claim that “[n]one of the hockey statistics are useful” for “find[ing] players who are undervalued or overvalued.” You believe that statistical analysis can say with confidence that Nikolai Khabibulin is worth less than $3.75M per year. For the vast majority of players we can do quite a bit better than that and we can definitely find numerous undervalued players throughout the league.

          • Tom says:

            2002 – 2008: ok, so baseball can be quantified, but that’s because it’s stationary; a fluid team game like basketball can’t be quantified

            This is not true. Dean Oliver convinced me that basketball could be quantified in 1998. He had the basic unit – a possession – and everything that happens on a possession is recorded.

            The type of games that cannot be quantified to the extent of evaluating individuals are the games that battle over territory – Soccer, rugby, football. and hockey.

  2. Axeman says:

    Tom,

    I’m inclined to agree with some of your points, and it strikes me that this is a fair and rational look at analytics. But …. to parse some of your comments:

    “The fundamental problem is that hockey has the equivilent of runs, but the hits and walks that create those runs are missing from the statistical package.”
    But they are. We just don’t necessarily recognize al the huts and runs yet, and we don’t necessarily collect all the information. But Corsi, Fenwick, possession time, zone starts, those are the hits and walks. You next paragraph discusses the need to force the opposition to turn over the puck So possession time and turnovers are relevant statistics. I recognize that things like turnovers aren’t tabulated well, but at least someone is counting. Puck battles won and lost – we can count those. All these things are hits, woks, sac bunts (ugh!) and HBP.

    While you’re entirely correct in stating that “moving the puck down the ice is also usually a collective effort.” that doesn’t meant the individual contribution need be ignored. My recollection is that teams with Gretzky on the ice moved the puck up the ice, as a team, way better than teams without Gretzky. The Sedins move the puck up the ice because they have it most of the time they are on the ice. Teams that have guys who win puck battles in the defensive zone move the puck up the ice well because they have the puck to move up the ice. None of these things are silver bullets, but they all matter. And the fact that we maybe don’t count them as well as we count doubles and triples doesn’t mean they can’t be counted and evaluated.

    As far as team sports go, hockey is far more team oriented than baseball or basketball. The individual contribution is much harder to describe. But we have moved forward light years in this regard, and continue to advance.

    My problem with either the stats based or (shall we call it) eye of the beholder based approach is that some people think that only one way is correct. There is only one thing I hate – single issue people!

    In your response to Magicpie n the comments, you noted that “But unless they can (claim that stats approach is the be all end all), the entire concept of Moneyball is out the window.” I disagree. The objective at the end may be to find players who are undervalued; the objective right now is to move along a continuum that advances the discussion. The fact that the answer isn’t complete doesn’t mean it isn’t out there.

    • Tom says:

      But they are. We just don’t necessarily recognize al the huts and runs yet, and we don’t necessarily collect all the information. But Corsi, Fenwick, possession time, zone starts, those are the hits and walks.

      These numbers are not the same thing. Corsi and Fenwicks don’t add up to goals. They are proxies for goals, substituted because goals come too infrequently to be statistically significant for most players. If we did measure possession time, it might tell us something, but it would tell us something about the team, not the individuals on the team.

      I agree that any team with Gretzky on it will do better whenever he is on the ice at virtually all aspects of the game. The converse is also true though: Gretzky will do better when he is surrounded by other excellent players than when he is surrounded by 20 stiffs. We can’t know – and never will be able to tell – how much is Gretzky and how much is the team. The same this is true for every star. We can see they are good and the fact they are good is usually reflected in the statistics, traditional and untraditional. They aren’t precise but they are good enough to denote quality. Still, they are not precise enough to tell us whether Daniel Sedin is a better player than Ryan Kesler.

      The real problem comes when we want to know whether Andrew Alberts will help a hockey team more than Andrew Ebbetts, whether Raffi Torres is worth more than Jannik Hansen. Baseball statistics answer those questions. Hockey statistics do not and never will.

      • Axeman says:

        The real problem comes when we want to know whether Andrew Alberts will help a hockey team more than Andrew Ebbetts, whether Raffi Torres is worth more than Jannik Hansen

        On that I can agree with you 100%

      • The real problem comes when we want to know whether Andrew Alberts will help a hockey team more than Andrew Ebbetts, whether Raffi Torres is worth more than Jannik Hansen. Baseball statistics answer those questions. Hockey statistics do not and never will.

        Baseball stats do not answer this question. They can answer which of these players was worth more last year, but that doesn’t tell you what will happen next year or if things don’t change in another system or used in a different role.

        You are arguing that hockey sabermetrics cannot do things that baseball cannot do, but you think it can, therefore hockey fails.

        • Tom says:

          Baseball stats do not answer this question. They can answer which of these players was worth more last year, but that doesn’t tell you what will happen next year or if things don’t change in another system or used in a different role.

          I think this is semantics but fair enough. Never mind about the future. Hockey stats don’t tell me whether whether Raffi Torres was worth more than Jannik Hansen last year. In baseball I can determine whether Andrew Alberts created more runs than Andrew Ebbetts and I can determine whether Raffi Torres created more runs than Jannik Hansen. I cannot use hockey statistics to say whether Torres created (or prevented) more goals than Jannik Hansen.

          We are talking about moneyball here. Billy Beane used baseball statistics to choose players other teams did not want or valued less highly. It worked then, but I doubt if he can do it any more because players are no longer so grossly misvalued, but still, he was able to do it. He did not even have to watch them play. (I don’t understand what you meant about a baseball players ststistics changing because of system or role. They are influenced by normal randomness, but role? I don’t think so.)

          That can’t be done in hockey. It never will be able to be done in hockey.

          • I can use hockey sabermetrics to find an over or under valued player. I may not have the same confidence level on my pick as if it was a baseball player, but that does not mean it is impossible.

            (I don’t understand what you meant about a baseball players ststistics changing because of system or role. They are influenced by normal randomness, but role? I don’t think so

            A pitcher goes from the starting rotation to the bullpen or a batter loses the protection behind him in the batting order and starts to get pitched around or any number of other factors in the system or role can influence a player’s results.

            I think we could make a pretty convincing case that Hansen and Torres were approximately equally valued last year. That is what you are asking for in your example?

          • Axeman says:

            Hockey stats don’t tell me whether whether Raffi Torres was worth more than Jannik Hansen last year.

            Yet. When I started reading Bill James, with the first Abstract in 1976, people still noted that the Best Hitting Team was the one leading the league in BAVG that season. Which of course was shown to be patently untrue. The science evolved. I think when we are eventually able to reach the point baseball is at today (almost 35 years since James popularized sabremetrics) we’ll be able to state with authority what I (for one) already know – Hansen is the more valuable of the two.

          • Tom says:

            Yet. When I started reading Bill James, with the first Abstract in 1976, people still noted that the Best Hitting Team was the one leading the league in BAVG that season. Which of course was shown to be patently untrue.

            The first annual was actually in the 1982, I think. The ’84 Annual was the first to get widespread distribution. It took a long time for the ideas to actually influence the sport, but that’s when the discovery happened. Within a short time, Klein and Reif were trying to do the same with hockey.

            I think when we are eventually able to reach the point baseball is at today (almost 35 years since James popularized sabremetrics) we’ll be able to state with authority what I (for one) already know – Hansen is the more valuable of the two.

            Can we imagine how we will do that? I can’t. I do agree, though. Hansen isn’t just more valuable, I think he is a lot more valuable.

          • Tom says:

            I can use hockey sabermetrics to find an over or under valued player. I may not have the same confidence level on my pick as if it was a baseball player, but that does not mean it is impossible.

            Well, okay, make a case.

  3. PSH: I can use hockey sabermetrics to find an over or under valued player. I may not have the same confidence level on my pick as if it was a baseball player, but that does not mean it is impossible.

    TB: Well, okay, make a case.

    Ok. Here is my top 20 corsi list adjusted for team and zone effects. Any player on the list that you dont recognize as a highly paid star (Mikhail Grabovski, Frans Nielsen, Daniel Winnik as examples) are undervalued. On the converse we have my worst list and players on it who command some respect in the league like Cam Fowler or Robyn Regehr are overvalued.

    • Tom says:

      Before I look at the list, I’ll need to know more about how well the adjusted Corsi actually reflects winning hockey. When I add up all the adjusted Corsi for the Vancouver Canucks, does the result predict their actual goals for and goals against? If I add up the adjusted Corsi for the entire Northwest Division does the prediction improve?

      • Tom you are moving the goalposts. You asked for a player who is over or undervalued as shown by hockey sabermetrics. I told you I can give you some but with a lesser confidence interval than I could in baseball. I delivered.

        Now I will try to answer your questions

        . When I add up all the adjusted Corsi for the Vancouver Canucks, does the result predict their actual goals for and goals against?

        Corsi is the shots attempted by a team minus the shots taken by a team. It has two adjustments made to give it value for individual players. First a team adjustment is made (it is easier to have a good Corsi on a good team than a bad one and this is taken into account). Second we adjust for the individual circumstances of how players are used by adjusting for their zone starts. Players who start more shifts in their own zone will have more shots against than players who start in the opposing zone.

        So to answer your question. No after the adjustments there is no reason whatsoever to expect that an adjusted Corsi should predict goals for or against. Raw Corsi ratings with no adjustment track with goals for and against quite strongly. Other factors can also be important – goaltending, special teams, shot quality and all of these can be taken into account somewhat. The adjustment is an attempt to individualize a team statistic. There is no logical reason whatsoever to add up adjusted ratings over a team. It won’t lead to anything useful.

        If I add up the adjusted Corsi for the entire Northwest Division does the prediction improve?

        I think my answer to the first question shows that this second question is pointless.

        • Tom says:

          Tom you are moving the goalposts. You asked for a player who is over or undervalued as shown by hockey sabermetrics. I told you I can give you some but with a lesser confidence interval than I could in baseball. I delivered.

          According to a statistic, these players are undervalued. I just spent about 2,000 words explaining why I don’t buy a number like the Corsi. I don’t think it is moving the goalpost to say you have some obligation to link the number you assign to the players to winning hockey games. Why are you using Corsi instead of shots or goals? Does it correlate best with winning? Even if the team Corsi does correlate to winning at some convincing rate how do you justify giving all the players on the ice equal credit (or the blame) for each shot?

          • Why are you using Corsi instead of shots or goals? Does it correlate best with winning?

            Yes. Corsi correlates better with winning than shots or goals.

            Even if the team Corsi does correlate to winning at some convincing rate how do you justify giving all the players on the ice equal credit (or the blame) for each shot?

            This is why Corsi requires contextual adjustment. With some sensible adjustments you can better gather individual skill levels. That is what I have done with these adjusted Corsi ratings.

            This method (and others) deliver players who are over and underrated. That was what you asked for. It doesn’t matter if you buy it or not.

          • Tom says:

            This method (and others) deliver players who are over and underrated. That was what you asked for. It doesn’t matter if you buy it or not.

            Actually we are looking for hockey playing value are we not? What exactly does this number purport to show? That Sean Bergenheim is more valuable than Corey Perry, Daniel Sedin or Nik Lidstrom? Never mind value, I think it is a stretch to suggest that Bergenheim does a better job of generating and preventing attempted goals than any of those players. Is that the claim?

            Do you expect the number for Bergenheim to stand up this year? Do you think that the fact that he did so well last year by this metric justifies the contract Tallon gave him?

            The issue isn’t whether I buy your formula. The issue is whether it works or not. I think you have an obligation to show that it works.

  4. Tom

    Nobody in their right mind thinks that one number is the be all and tell all statistic of the NHL. You are smarter than thinking such a thing exists and yet you bring up that strawman to try to argue against hockey sabermetrics in general.

    • Tom says:

      Nobody in their right mind thinks that one number is the be all and tell all statistic of the NHL. You are smarter than thinking such a thing exists and yet you bring up that strawman to try to argue against hockey sabermetrics in general.

      This is exactly my position. There is no statistic that will evaluate individual hockey players. That has been my position from the first paragraph of the post. If you are saying that adjusted Corsi cannot be validated in any way but it still imperfectly says something about the players, I have two questions:

      1) What exactly is it saying?

      2) How do you decide whether Sean Bergenheim is ranked so highly is the result of his play or the result of imperfection in the statistic?

      • 1) What exactly is it saying?

        Adjusted Corsi is the best number I know of to show puck possession ability of a given player.

        2) How do you decide whether Sean Bergenheim is ranked so highly is the result of his play or the result of imperfection in the statistic?

        Lat year Sean Bergenheim was very good at driving puck possession in Tampa Bay. That is a result orf his play.

        The problem, you have is you cannot accept this number unless it is a one number ranking that is the be all and end all of hockey statistics. That all or nothing approach is just obstructionism. We can see over and under valued players through puck possession numbers because puck possession correlates well with winning hockey games and because it is not something that is looked at in mainstream hockey analysis.

        • JS says:

          The first annual was actually in the 1982, I think.

          No, 1976. I have it, and everyone since. He home published in the era of the gestetner, and sold through the back pages of The Sporting News. About 75 copies a year at first. Ballantyne started publishing with the 1982 edition, I think it was – that was when they got wider attention. I think Okrent had a lot to do with that at the time.

        • Tom says:

          Adjusted Corsi is the best number I know of to show puck possession ability of a given player.

          What is the puck possession ability? I have been watching hockey for more than 50 years and I have no idea what puck possession ability is. Can you define it and tell me how Corsi measures it? Is there a way to recognize it in a player when I watch a game?

          Last year Sean Bergenheim was very good at driving puck possession in Tampa Bay

          Driving it? Bergenheim was number 11 in ice time among Tampa forwards, despite being one of the best players in the league at driving puck possession. Either his coach had the same problem I do in recognizing puck possession ability, or it isn’t very important.

          I think Bergenheim is a pretty good player. He reminds me of a lower case Samuelsson, a disciplined winger whose offensive philosophy involves shooting at every opportunity.

          • Driving it? Bergenheim was number 11 in ice time among Tampa forwards, despite being one of the best players in the league at driving puck possession. Either his coach had the same problem I do in recognizing puck possession ability, or it isn’t very important.

            When Bergenheim was on the ice, Tampa Bay had the puck a significant amount of the time. The numbers say that. Largely Bergenheim did this while playing on the Tampa second line. I am sure Guy Boucher was well aware of this. I am sure that the competition level had something to do with it. Stamkos and St Louis drew the top checkers on the opposition and this allowed Bergenheim to do as well as he did. Bergenheim was key to this, when he wasn’t on the ice Tampa did not have the same puck possession.

            As for why you never recognized this, I expect it has to do with not watching enough Tampa Bay games.

      • JS says:

        This is exactly my position. There is no statistic that will evaluate individual hockey players

        Of course there isn’t, and I don’t recall seeing anyone claiming otherwise. But to use the baseball examples again, BA doesn’t tell us everything about a hitter — but it tells us something. To me, the idea should be to find ways to advance our knowledge about why and how Jannik Hansen is better than Torres. Corsi may be one part of that, time of possession another, minutes another, QofA another …. it all adds up.

  5. Roberto says:

    The traditional way is to acknowledge that points don’t tell us much about most players and look at the skills the player actually has. How fast does he skate? How strong is he? Does he see the ice well? Good hands? Good shot? Mistake prone? Quick stick? Decision making? Can he score? If he does not score, what does he do to help a team? Is he likely to improve? How do his skills fit on our team? Will he be better or worse on our team? Will he make us better or worse?

    This comment is interesting to me in that it goes to the heart of what Beane disliked so much in player evaluation. He himself had a great skill set, yet was not a very good baseball player, and it’s probably this recognition that drove him to use only statistics to evaluate baseball players for his team. It’s the very reason he didn’t watch his own team play. “Never trust your eyes.”

    I don’t think it’s absurd to think this can be achieved in hockey, albeit with larger error bars. It may even be able explain why guys with seemingly all the tools can’t seem to get it done at the NHL level. (Daigle comes to mind here). In a salary cap era where an under performing player on a large contract can seriously harm your team’s ability to compete, this matters. If Gillis can use this to avoid a Daigle-like signing, I’m all for it.

  6. Tom says:

    This comment is interesting to me in that it goes to the heart of what Beane disliked so much in player evaluation. He himself had a great skill set, yet was not a very good baseball player, and it’s probably this recognition that drove him to use only statistics to evaluate baseball players for his team. It’s the very reason he didn’t watch his own team play. “Never trust your eyes.”

    I don’t know anythying about Beane as a player, but I doubt if his own experience drove him to use statistics. I can understand the impulse though because it is impossible to tell who is good in baseball without the numbers. I dobn’t watch much baseball any more, but I tune in at this time of the year. I don’t know the players and if the announcers didn’t tell me, I couldn’t pick out the stars watching them bat. The numbers have always been important in evaluating baseball players. What sabermetrics did was produce better numbers.

    I don’t think examining Junior statistics would make any difference to a case like Daigle. His secondary numbers would be every bit as good as his primary ones. I think Daigle’s problem was his hockey brain. It worked fine in Junior when he had lots of time, but he did not think fast enough for the NHL game. He was forever forcing things or making a pass too late. He might have been okay if he had a couple of years in the minors, but that wasn’t on for him.

  7. Wow, that’s what I was searching for, what a stuff! existing here at this blog, thanks
    admin of this website.

Trackbacks

Check out what others are saying about this post...
  1. [...] TOM BENJAMIN’S NHL BLOG: Tom takes on James Mirtle’s recent series on advanced statistical analysis in hockey. [...]

  2. [...] Tom Benjamin, CanucksCorner: “Hockey statistics will never do what we want them to do, which is to effectively evaluate individual hockey players. To give us answers when considering a trade or a personnel decision. To tell us whether this third line winger creates more wins than that number four defenseman. Baseball statistics can produce answers to these kinds of questions, while hockey statistics can only produce more questions. [...]



Speak Your Mind

Tell us what you're thinking...
and oh, if you want a pic to show with your comment, go get a gravatar!