Wednesday, October 03, 2007
Dept. of Timely Studies
I'm all in favour of mathematical analysis of hockey, but this press release from the U of Guelph is embarrassing (ÞOffWing):
Can anyone think of anything that might have changed between 2004 and now that might be relevant?
Shootouts were instituted two seasons ago. For two years, there has been a goal scored to break the tie in 100 per cent of overtime games.
I understand that there is a delay between research and publication, but this press release was dated today, and gives no indication that "now" does not in fact, strictly, refer to the present. Looking forward to seeing the media coverage.
Prof. Alfons Weersink, Department of Food, Agricultural and Resource Economics, examined the outcome of all the NHL games played from 1995 to 2004 to determine the impact the 1999/2000 season rule change has had on the way the game is played. Published recently in the Canadian Journal of Economics, his research shows that the current overtime award system is having a perverse effect on the way professional hockey is being played.
Can anyone think of anything that might have changed between 2004 and now that might be relevant?
"The rule change has been successful because now there is a goal scored to break the tie in about 50 per cent of overtime games compared with 25 per cent under the old rule," said Weersink.
Shootouts were instituted two seasons ago. For two years, there has been a goal scored to break the tie in 100 per cent of overtime games.
I understand that there is a delay between research and publication, but this press release was dated today, and gives no indication that "now" does not in fact, strictly, refer to the present. Looking forward to seeing the media coverage.
Comments:
I'm curious about the impact that changing to the shootout would have on his contentions. You're still guaranteed a point for making it to overtime, so potentially his contentions still stand with regards to conservativism in regulation. Does the shootout lead to conservative play, if you feel like you have stronger goaltending/stronger shooters?
I didn't know that a quarter of all games end tied (as opposed to a fifth previously).
While I would probably agree that, if the game is tied and there's five minutes to go, you're probably more likely to play conservative and head to OT, I question whether the point-to-all in OT would actually have an impact on the entire game. Further to that, it would be pretty much impossible to prove that any of this can be blamed squarely on the new OT system.
Not that I necessarily like the system, but this wouldn't be the reason I would pick to harp on.
The 2005 rule change does not mean the authors' conclusions about incentives are garbage; not at all. But the math -- the data -- is out of date, which is a pretty severe problem given that
(A) they don't acknowledge that in this press release, or in the June 2006 version of the paper I was sent (it was published in May 2007)
(B) without the hard math behind it, it's present value is no greater than a blog post. Like this one, that I wrote in July 2005, without the help of economists in two other countries.
Other note: the authors do demonstrate quantitatively that 3rd-period play became more conservative after the introduction of the loser point in 1999.
It's an admirable effort; I should stress here that I'm sure my quarrel is with U of Guelph PR more than the authors. Bad luck for them that the NHL went and changed the rules on them again in the middle of their research. That said: as best I can tell from reading it, it would be a fraction of the original amount of work to update it to include data for the new-new rules (equations are set up, theorems are already proved, etc.). I hope they issue an addendum.
Hate to make fun of you Matt but the post itself say it is a study of 1995-2004...no where in there does it say it includes the post lockout shootout era. That would mean that his study is only when games ended after OT ended, not currently with the Shootout format. The fact that you call this embarrasing without actually reading what it says is even more embarrasing.
I guess though you are used to seeing things how you want see them as you are a Flames fan who is probably still living in 2004 thinking that Calgary has a chance of following up on that miracle cup run.
Let's Go Oilers....Calgary Sucks!!!
Prof. Alfons Weersink, Department of Food, Agricultural and Resource Economics
Doesn't anyone else find the Prof's name incredibly funny/hoaxy? I suppose we've checked that Guelph actually has a Dept. of Food,Ag,&Res Econ? Now I'm certainly a fan of all learning and all knowledge and love hockey to death, BUT is it a good use of the Prof's time to be researching hockey results from 12 years ago?
For me, it's just the whole idea of causality vs. a correlation. Evidently they've proved there's some kind of statistically significant correlation between the newer OT rules and an increase in ties (which can potentially be called an increase in "conservative play"). To say that it's resulted in "less aggressive offence during the regular games" is a pretty difficult statement to back up, though. That's an entirely subjective assertion, really.
I hesitate to say this given the work that I do...but seriously, is there nothing else that he could study instead? What do you even do with information like this?
Hate to come down on you Matt but if you looked more into the story you would find out why this is relavent. As I mentioned before this was a study done to see if giving the extra point for a reg. tie resulted in more actual games being decided in overtime. The professor found that it did, twice as much in fact. The relevance of this story is that the NHL is considering the 3 point format where a regulation win would get you 3 points, 2 for a OT/Shootout win and 1 for a OT/Shootout loss. His arguement is that it resulted in more actual decisions (in the time frame of 5 years before and after implementation of the rule) after the rule than before. The whole relavance is that if they want more games decided before the extra frame they have a basis to say that when the extra point was free more games were decided then maybe a 3 point game would result in less OT's/Shootouts. If you are interested in the whole story and the relevance to why it has been re-released today check out the link posted
http://www.tsn.ca/nhl/news_story/?ID=219827&hubname=nhl
and then you will see why the study is not dumb or embarrassing.
Dare, I also hope that this comment shows what you can do with this type of information..more accurately hypothesis whether a 3 point game would result in less ties and more regulation ends.(as well as quit rewarding teams for losing in OT/Shootouts as had this rule been in effect last year the Wild, Thrashers and Stars who were all stellar shootout teams would not have had as good a record compared to a Ottawa or Detroit who took care of business in regulation)
Thanks for the find, Matt. I would like to see the full article, it sounds interesting.
Some of this wording seems a bit strong, but on the whole I'm sure that he's right. It would be hard to decide how much of that was caused by the increased incentive to play for the tie and how much was due to just generally more conservative tactics.
I think I would have looked at results when there is a one goal difference in score at the second intermission. Most teams just played to suck the joy out of the game at that point, the losing team is the more likely team to score, but scoring chances on the whole start becoming very scarce.
I'm sure that this trend reversed starting after the lockout. That's why the bookmaker's preseason team points lines (my subject du jour) for 05/06 totalled far less than all other projections from hockey publications and from pundits like you and me. We were adding up an extra point for every tie from 03/04 ... they were moving the lines on 3rd period scoring props and expecting a shade fewer OT games. And they were right.
It's regressed though. I haven't checked, but I watched at least a couple of hours of the ANA v VAN series from last year. That was enough. We're back where we started folks, at least in the west.
The eastern conference should still see the effect of the NHLs commitment to a more entertaining style of play. Especially the southeast conference. It should be reflected in third period EV scoring. And the props from your local bookie on "which period will the most goals be scored in?" should reflect that. The more conservatively coached team should drive that line as well. Or so I think. The boring team drives the pace, I'm afraid.
You'd want to weed out the empty net situation as well. And focussing on the EV goals over the timeframe just makes basic sense.
That's my best guess anyways, I'm fairly confident that the facts will support that. One blustery day this winter, if the mood strikes, I'll crunch the numbers on this.
I really think that this type of research is important. If the NHL wishes to improve the game for the fans, they have to determine what is wrong. And why.cpbtlqhy
Oilinator:
All great points. And I thought you just yelled at that fever4flames nutbar.
The "points for a regulation win" idea is key. That changes the whole risk/reward structure for every coach. Regardless of his other biases.
For the life of me, I don't know why this wasn't brought in when they first went to the OT "extra point" sytem. As I understand it, the original idea was just that, every single game would produce a total of 3 points.
Personally, I think the trend of quieter third periods, especially in the west, is reason for concern as well. I mean the new rules have made the game more fun to watch I think. But a lot of the increase in EV scoring rate (which wasn't huge anyways) was not the rules, but a renewed commitment of NHL teams to play a fun game to watch. Or at least that's the way it seems to me.
Hi all. Oilinator, your 2nd comment seems a lot more good faith, so let me try again to explain what the issue is (and is not) here.
I agree that incentives matter, as I wrote 2+ years ago. The study was not a waste of time, and was not poorly executed. But the point of the study was to examine a specific rule change in 1999 designed to address a specific problem: regular season OT was boring because everyone was playing for the tie during OT.
The NHL decided that the way to solve this problem was to give them both a point regardless, and award a 3rd point if and only if there was an OT winner. There were still games (170 of them in '03/04) that finished as ties.
The study looked at two things:
1) Did the change accomplish the goal of more OT scoring (discouraging teams from playing for the tie during OT)?
2) Did the fact that a team was guaranteed at least one point by finishing 60 minutes tied affect play in the 3rd period?
It's interesting stuff. But it is, as things stand now, a historical footnote. In 1999, the NHL changed the way they award points to order to make ties less common. In 2005, they changed the rules so that ties never happen ever. A study examining the 1999 rule change can be an interesting window into how incentives affect team play, coaching, whatever, but it cannot be reported using so much of the present tense. Thing #1 is not applicable as the rules stand now BECAUSE YOU CAN'T FUCKING PLAY FOR A TIE EVEN IF YOU WANT TO ONE TEAM IS ALWAYS GETTING TWO POINTS.
Look at the 2nd paragraph of the TSN story you linked to: "The study found that the league achieved its goal of having more games end in overtime by awarding a point to losers in the extra session, but that it did so at the expense of having games end in regulation time."
I have an inkling that some of the communication breakdown here is over the meaning of "overtime". The meaning of "having more games end in overtime" in the context of this study is "having fewer games end in ties". The league subsequently achieved this goal much more directly, by legislating ties away.
Please read my previous comment (#2) again -- I don't wish to slime this study's authors. But the entire marginal value of having these conclusions come from professors of economics rather than amateur statisticians is the specific math regarding the specific incentives. I, or a 10-year-old with an NHL Yearbook and a calculator, could have told you that after the loser point was instituted, Ties went down but Games Tied Thru 60 went up. What is interesting (and only quantifiable by smart folks like the authors of this study) is how the 3rd period play is affected.
And when the expected payoff of being Tied Thru 60 changes as drastically as it did in 2005, the old calculations on this are no damn good anymore, and you can't report on them -- whether you're the U of Guelph PR department or the Canadian Press -- as if they are.
Point taken Matt. Hope you took me seriously when I said no Offense in my comments. Now that you have cleared it up your comments make much more sense to me.
I also am not commenting to agree with the Prof. as I don't. I agree with you that if they did institute a 3 point game it would just lead to a lot of boring trap hockey in the third period. Teams would be in such need of the 3 point win that every team would have to play defensive with the lead. Kind of a double edged sword. 3 point games happen lots already making a regulation win not much of a big deal but a actual 3 point win would just result in boring third periods. Sorry to come at you and think, things could always be worse...you could be a Flames fan..oops you are. Let's Go Oilers!!!
The paper intended to give a analytical framework to deal with rule changes. Once the rule is changed you can incorporate into the framework and do a similar analysis again. The paper is a delayed publication (thats how academia work since everything has to be peer reviewed). The statistical analysis is incidental and should be looked as a verification of the theory. Only then we can use the framework to project our analysis into unknown rule changes.
A informal publication was done in 2004 as University of Southampton Discussion Paper here is the link:
http://eprints.soton.ac.uk/33785/
Hope that helps
Post a Comment
<< Home
I'm curious about the impact that changing to the shootout would have on his contentions. You're still guaranteed a point for making it to overtime, so potentially his contentions still stand with regards to conservativism in regulation. Does the shootout lead to conservative play, if you feel like you have stronger goaltending/stronger shooters?
I didn't know that a quarter of all games end tied (as opposed to a fifth previously).
While I would probably agree that, if the game is tied and there's five minutes to go, you're probably more likely to play conservative and head to OT, I question whether the point-to-all in OT would actually have an impact on the entire game. Further to that, it would be pretty much impossible to prove that any of this can be blamed squarely on the new OT system.
Not that I necessarily like the system, but this wouldn't be the reason I would pick to harp on.
The 2005 rule change does not mean the authors' conclusions about incentives are garbage; not at all. But the math -- the data -- is out of date, which is a pretty severe problem given that
(A) they don't acknowledge that in this press release, or in the June 2006 version of the paper I was sent (it was published in May 2007)
(B) without the hard math behind it, it's present value is no greater than a blog post. Like this one, that I wrote in July 2005, without the help of economists in two other countries.
Other note: the authors do demonstrate quantitatively that 3rd-period play became more conservative after the introduction of the loser point in 1999.
It's an admirable effort; I should stress here that I'm sure my quarrel is with U of Guelph PR more than the authors. Bad luck for them that the NHL went and changed the rules on them again in the middle of their research. That said: as best I can tell from reading it, it would be a fraction of the original amount of work to update it to include data for the new-new rules (equations are set up, theorems are already proved, etc.). I hope they issue an addendum.
Hate to make fun of you Matt but the post itself say it is a study of 1995-2004...no where in there does it say it includes the post lockout shootout era. That would mean that his study is only when games ended after OT ended, not currently with the Shootout format. The fact that you call this embarrasing without actually reading what it says is even more embarrasing.
I guess though you are used to seeing things how you want see them as you are a Flames fan who is probably still living in 2004 thinking that Calgary has a chance of following up on that miracle cup run.
Let's Go Oilers....Calgary Sucks!!!
Prof. Alfons Weersink, Department of Food, Agricultural and Resource Economics
Doesn't anyone else find the Prof's name incredibly funny/hoaxy? I suppose we've checked that Guelph actually has a Dept. of Food,Ag,&Res Econ? Now I'm certainly a fan of all learning and all knowledge and love hockey to death, BUT is it a good use of the Prof's time to be researching hockey results from 12 years ago?
For me, it's just the whole idea of causality vs. a correlation. Evidently they've proved there's some kind of statistically significant correlation between the newer OT rules and an increase in ties (which can potentially be called an increase in "conservative play"). To say that it's resulted in "less aggressive offence during the regular games" is a pretty difficult statement to back up, though. That's an entirely subjective assertion, really.
I hesitate to say this given the work that I do...but seriously, is there nothing else that he could study instead? What do you even do with information like this?
Hate to come down on you Matt but if you looked more into the story you would find out why this is relavent. As I mentioned before this was a study done to see if giving the extra point for a reg. tie resulted in more actual games being decided in overtime. The professor found that it did, twice as much in fact. The relevance of this story is that the NHL is considering the 3 point format where a regulation win would get you 3 points, 2 for a OT/Shootout win and 1 for a OT/Shootout loss. His arguement is that it resulted in more actual decisions (in the time frame of 5 years before and after implementation of the rule) after the rule than before. The whole relavance is that if they want more games decided before the extra frame they have a basis to say that when the extra point was free more games were decided then maybe a 3 point game would result in less OT's/Shootouts. If you are interested in the whole story and the relevance to why it has been re-released today check out the link posted
http://www.tsn.ca/nhl/news_story/?ID=219827&hubname=nhl
and then you will see why the study is not dumb or embarrassing.
Dare, I also hope that this comment shows what you can do with this type of information..more accurately hypothesis whether a 3 point game would result in less ties and more regulation ends.(as well as quit rewarding teams for losing in OT/Shootouts as had this rule been in effect last year the Wild, Thrashers and Stars who were all stellar shootout teams would not have had as good a record compared to a Ottawa or Detroit who took care of business in regulation)
Thanks for the find, Matt. I would like to see the full article, it sounds interesting.
Some of this wording seems a bit strong, but on the whole I'm sure that he's right. It would be hard to decide how much of that was caused by the increased incentive to play for the tie and how much was due to just generally more conservative tactics.
I think I would have looked at results when there is a one goal difference in score at the second intermission. Most teams just played to suck the joy out of the game at that point, the losing team is the more likely team to score, but scoring chances on the whole start becoming very scarce.
I'm sure that this trend reversed starting after the lockout. That's why the bookmaker's preseason team points lines (my subject du jour) for 05/06 totalled far less than all other projections from hockey publications and from pundits like you and me. We were adding up an extra point for every tie from 03/04 ... they were moving the lines on 3rd period scoring props and expecting a shade fewer OT games. And they were right.
It's regressed though. I haven't checked, but I watched at least a couple of hours of the ANA v VAN series from last year. That was enough. We're back where we started folks, at least in the west.
The eastern conference should still see the effect of the NHLs commitment to a more entertaining style of play. Especially the southeast conference. It should be reflected in third period EV scoring. And the props from your local bookie on "which period will the most goals be scored in?" should reflect that. The more conservatively coached team should drive that line as well. Or so I think. The boring team drives the pace, I'm afraid.
You'd want to weed out the empty net situation as well. And focussing on the EV goals over the timeframe just makes basic sense.
That's my best guess anyways, I'm fairly confident that the facts will support that. One blustery day this winter, if the mood strikes, I'll crunch the numbers on this.
I really think that this type of research is important. If the NHL wishes to improve the game for the fans, they have to determine what is wrong. And why.cpbtlqhy
Oilinator:
All great points. And I thought you just yelled at that fever4flames nutbar.
The "points for a regulation win" idea is key. That changes the whole risk/reward structure for every coach. Regardless of his other biases.
For the life of me, I don't know why this wasn't brought in when they first went to the OT "extra point" sytem. As I understand it, the original idea was just that, every single game would produce a total of 3 points.
Personally, I think the trend of quieter third periods, especially in the west, is reason for concern as well. I mean the new rules have made the game more fun to watch I think. But a lot of the increase in EV scoring rate (which wasn't huge anyways) was not the rules, but a renewed commitment of NHL teams to play a fun game to watch. Or at least that's the way it seems to me.
Hi all. Oilinator, your 2nd comment seems a lot more good faith, so let me try again to explain what the issue is (and is not) here.
I agree that incentives matter, as I wrote 2+ years ago. The study was not a waste of time, and was not poorly executed. But the point of the study was to examine a specific rule change in 1999 designed to address a specific problem: regular season OT was boring because everyone was playing for the tie during OT.
The NHL decided that the way to solve this problem was to give them both a point regardless, and award a 3rd point if and only if there was an OT winner. There were still games (170 of them in '03/04) that finished as ties.
The study looked at two things:
1) Did the change accomplish the goal of more OT scoring (discouraging teams from playing for the tie during OT)?
2) Did the fact that a team was guaranteed at least one point by finishing 60 minutes tied affect play in the 3rd period?
It's interesting stuff. But it is, as things stand now, a historical footnote. In 1999, the NHL changed the way they award points to order to make ties less common. In 2005, they changed the rules so that ties never happen ever. A study examining the 1999 rule change can be an interesting window into how incentives affect team play, coaching, whatever, but it cannot be reported using so much of the present tense. Thing #1 is not applicable as the rules stand now BECAUSE YOU CAN'T FUCKING PLAY FOR A TIE EVEN IF YOU WANT TO ONE TEAM IS ALWAYS GETTING TWO POINTS.
Look at the 2nd paragraph of the TSN story you linked to: "The study found that the league achieved its goal of having more games end in overtime by awarding a point to losers in the extra session, but that it did so at the expense of having games end in regulation time."
I have an inkling that some of the communication breakdown here is over the meaning of "overtime". The meaning of "having more games end in overtime" in the context of this study is "having fewer games end in ties". The league subsequently achieved this goal much more directly, by legislating ties away.
Please read my previous comment (#2) again -- I don't wish to slime this study's authors. But the entire marginal value of having these conclusions come from professors of economics rather than amateur statisticians is the specific math regarding the specific incentives. I, or a 10-year-old with an NHL Yearbook and a calculator, could have told you that after the loser point was instituted, Ties went down but Games Tied Thru 60 went up. What is interesting (and only quantifiable by smart folks like the authors of this study) is how the 3rd period play is affected.
And when the expected payoff of being Tied Thru 60 changes as drastically as it did in 2005, the old calculations on this are no damn good anymore, and you can't report on them -- whether you're the U of Guelph PR department or the Canadian Press -- as if they are.
Point taken Matt. Hope you took me seriously when I said no Offense in my comments. Now that you have cleared it up your comments make much more sense to me.
I also am not commenting to agree with the Prof. as I don't. I agree with you that if they did institute a 3 point game it would just lead to a lot of boring trap hockey in the third period. Teams would be in such need of the 3 point win that every team would have to play defensive with the lead. Kind of a double edged sword. 3 point games happen lots already making a regulation win not much of a big deal but a actual 3 point win would just result in boring third periods. Sorry to come at you and think, things could always be worse...you could be a Flames fan..oops you are. Let's Go Oilers!!!
The paper intended to give a analytical framework to deal with rule changes. Once the rule is changed you can incorporate into the framework and do a similar analysis again. The paper is a delayed publication (thats how academia work since everything has to be peer reviewed). The statistical analysis is incidental and should be looked as a verification of the theory. Only then we can use the framework to project our analysis into unknown rule changes.
A informal publication was done in 2004 as University of Southampton Discussion Paper here is the link:
http://eprints.soton.ac.uk/33785/
Hope that helps
Post a Comment
<< Home