Thursday, May 26, 2005
Sabermetrics and hockey
In the comments on yesterday's post, Jes Golbez points to a study one person named pnep put on the web at hfboards.com on his metric for who should make the hall of fame. It contradicts my intuition picking Glenn Anderson ahead of Dale Hawerchuk. I believe this is due to a flaw in the methodology of the study, but at this point I am unclear on exactly what the methodology was. I hope to address this in detail later when I further understand this study.
The idea behind this study is one I would like to support. I am a big fan of sabermetrics in baseball. Sabermetrics is the statistical analysis of baseball. Jim Albert describes it here. Bill James is one of my favorite authors. He is the most accomplished sabermetrics author in the world today. I strongly recommend anyone who is interested read his works.
I have wondered many times if it was possible to apply sabermetrics to hockey. The best attempt that I have seen on the net is Daryl Shilling's hockey project. I think it is imperfect, but if we keep in mind the errors that go with numbers produced by such attempts it is a nice effort. I think that it is not possible to have nearly as precise a theory in hockey as there is in baseball (though I would love to be proven incorrect). To explain why, I will quote from Bill James in his Historical Baseball Abstract
The difference between a good statistical analyst and a poor statistical analyst is that a good statistical analyst ... understands this, and a bad one implicitly denies it.
A good statistical analyst, in studying the statistical record of a baseball season, asks three or four essential questions:
1) What is missing from the picture?
2) What is distorted here, and what is accurately portrayed?
3) How can we include what has been left out?
4) How can we correct what has been distorted?
We all know many things and many different types of things that are not reflected in the statistical record. Acknowledging this, a good statistical analyst is sometimes able to reach out and draw areas of the game which were previously undocumented inside the tent, inside the focus of the statistical record. Sabermetrics is sometimes able to invent a way to correct for one or another distortion of the statistical picture.
The bad statistical anaylst , ot the other hand, will assume that what the statistical tells him must be true and complete- and by making that assumption, will forfeit his ability to add anything significant to the record.
We must look at hockey statistics. Playing winning hockey at its simplest requires being good at scoring goals and being good at preventing your opponent from scoring goals. If your team scores the most and allows the least goals your team will win. We must look at individual statistics and see how good they are as a proxy for a given player's ability to score and prevent goals.
Preventing goals is almost impossible to quantify. Right off the bat, we hit a huge roadblock. There is little to no reliable way to quantify how many goals a player has prevented being scored.
Most hockey stats attempt to quantify goal production. And they do a good (although sometimes misleading) job of this. An example of a misleading conclusion relating to goal production would be two average players. One player plays with all star linemates. The other plays with crappy linemates. The player with all star linemates will be creditted with assists and goals when his linemate did most of the work and he was just along for the ride. The player with crappy linemates will not get the same chance to score or assist. Even if he sets up a sure goal, there is no guarantee his crappy linemate will score. On the surface, these two identical average players will not appear identical because the one with the better linemates scored more often. This is misleading. In principle, I think this can be corrected for, although I am not certain I have ever seen it done successfully.
Hockey is different from baseball. It is much harder to quantify. As a result, I don't think as precise a sabermetric theory is possible. I think because of the misleading nature of some hockey stats and because many important aspects of the game are unquantifiable, many results of attempted sabermetric hockey studies produce garbage. Its not easy to shuck off the garbage from the input statistics and make something meaningful come out.
The idea behind this study is one I would like to support. I am a big fan of sabermetrics in baseball. Sabermetrics is the statistical analysis of baseball. Jim Albert describes it here. Bill James is one of my favorite authors. He is the most accomplished sabermetrics author in the world today. I strongly recommend anyone who is interested read his works.
I have wondered many times if it was possible to apply sabermetrics to hockey. The best attempt that I have seen on the net is Daryl Shilling's hockey project. I think it is imperfect, but if we keep in mind the errors that go with numbers produced by such attempts it is a nice effort. I think that it is not possible to have nearly as precise a theory in hockey as there is in baseball (though I would love to be proven incorrect). To explain why, I will quote from Bill James in his Historical Baseball Abstract
The difference between a good statistical analyst and a poor statistical analyst is that a good statistical analyst ... understands this, and a bad one implicitly denies it.
A good statistical analyst, in studying the statistical record of a baseball season, asks three or four essential questions:
1) What is missing from the picture?
2) What is distorted here, and what is accurately portrayed?
3) How can we include what has been left out?
4) How can we correct what has been distorted?
We all know many things and many different types of things that are not reflected in the statistical record. Acknowledging this, a good statistical analyst is sometimes able to reach out and draw areas of the game which were previously undocumented inside the tent, inside the focus of the statistical record. Sabermetrics is sometimes able to invent a way to correct for one or another distortion of the statistical picture.
The bad statistical anaylst , ot the other hand, will assume that what the statistical tells him must be true and complete- and by making that assumption, will forfeit his ability to add anything significant to the record.
We must look at hockey statistics. Playing winning hockey at its simplest requires being good at scoring goals and being good at preventing your opponent from scoring goals. If your team scores the most and allows the least goals your team will win. We must look at individual statistics and see how good they are as a proxy for a given player's ability to score and prevent goals.
Preventing goals is almost impossible to quantify. Right off the bat, we hit a huge roadblock. There is little to no reliable way to quantify how many goals a player has prevented being scored.
Most hockey stats attempt to quantify goal production. And they do a good (although sometimes misleading) job of this. An example of a misleading conclusion relating to goal production would be two average players. One player plays with all star linemates. The other plays with crappy linemates. The player with all star linemates will be creditted with assists and goals when his linemate did most of the work and he was just along for the ride. The player with crappy linemates will not get the same chance to score or assist. Even if he sets up a sure goal, there is no guarantee his crappy linemate will score. On the surface, these two identical average players will not appear identical because the one with the better linemates scored more often. This is misleading. In principle, I think this can be corrected for, although I am not certain I have ever seen it done successfully.
Hockey is different from baseball. It is much harder to quantify. As a result, I don't think as precise a sabermetric theory is possible. I think because of the misleading nature of some hockey stats and because many important aspects of the game are unquantifiable, many results of attempted sabermetric hockey studies produce garbage. Its not easy to shuck off the garbage from the input statistics and make something meaningful come out.
Comments:
Hornets
While the New Orleans Hornets haven't won much lately, their
woes pale in comparison to those of their next opponent.
The Hornets, losers in five of six, will try to get heading back
in the right direction by earning another road win over the
sputtering Minnesota Timberwolves on Saturday night.
New Orleans (15-32) had won six of seven at one point last
month, [url=http://wholesalenikeshoes.manifo.com/]wholesale nike shoes[/url] but its inability to contain opponents has hurt the team
over the past three weeks. The Hornets have given up an average of
104.8 points while dropping six of eight.
They let Denver shoot 52.4 percent in a 113-98 defeat Friday,
their third in a row, and that virtually negated a 52.8-percent
performance at the other end of the floor. Ryan Anderson had 21
points off the bench but no other teammate scored more than 13.
Eric Gordon had nine points on 3-of-8 shooting after sitting out
Wednesday's 104-99 defeat to Utah. It was a scheduled day off for
Gordon, who missed New Orleans' first 29 games with right knee
soreness and is currently not permitted to play in the second game
of back-to-backs.
The Hornets also gave up 32 points on 22 turnovers Friday, their
most in one game since a season-high 24 in a loss to Philadelphia
on Nov. 7.
"The goal for is ... to get our minds right to play (Saturday)
night," coach Monty Williams said. "We can't sweep this under the
rug by any stretch. We have to understand that you can't turn the
ball over against any team in the league that many times."
Gordon was one of six players in double figures with 16 points
in a 104-92 victory over Minnesota on Jan. 11, and Greivis Vasquez
had a team-best 18 points and 13 assists. Vasquez has 11 or more
assists in six of his last eight games but just missed extending a
three-game streak of double-digit performance Friday as he finished
with nine.
The Hornets could use Vasquez's help to continue their recent
success in Minnesota. They're 9-2 in their last 11 visits and have
won the past two.
New Orleans has also prevailed in four of the last six overall
matchups but lost 113-102 at home to the Timberwolves on Dec. 14.
Nikola Pekovic had a career-high 31 points on 14-of-18 shooting and
Andrei Kirilenko scored 21.
Minnesota, though, enters this game having lost 11 of 12 and a
season-worst six in a row. The Timberwolves (17-26) have also
matched another season high with three consecutive home
defeats.
They cut a 29-point deficit to four with 5:55 remaining on
Friday against the visiting Los Angeles Lakers, but would get no
closer in a 111-100 defeat. In his second game back from a sprained
ankle, Alexey Shved had a team-best 18 points and was one of five
players in double figures.
"It's hard," point guard Ricky Rubio said. "We've been losing a
lot. Even if we make the effort, sometimes it's not enough."
The Hornets are 1-3 on their five-game trip entering Saturday's
finale, while the Timberwolves are 0-2 on their season-long
six-game homestand.
<< Home
This is also what makes it difficult to come up with a good simulator. Just how to you quantify chemistry between players the leadership potential? The intangibles are too many to count.
The difference between hockey and baseball from a statistical perspective is that the latter consists of a series of completely individual actions (i.e. at bats), whereas in hockey each action has several components and several contributors. Hockey statistical analyses must find ways to incorporate peripheral statistics such as giveaways, takeaways, scoring chances, etc. into the metric.
# posted by 
: October 10, 2007 2:30 PM
: October 10, 2007 2:30 PM
I enjoyed reading your articles. This is truly a great read for me. I have bookmarked it and I am looking forward to reading new articles an i keep a nice post
[p]Clarke, who replaced Mark [url=http://www.fashionhermesbagsuk.co.uk]discount hermes bags kelly[/url] Anson in the post . getting a considerable element in leather-based productions, birkin replica are well-known using the exceptional and stylish taste . One of this [url=http://www.springhermeshandbags.co.uk]Hermes handbags UK [/url] company from Hermes is with Birkin handbag another quality product hand woven . Lv Handbags Outlet There we had been really glad for you to spray this incredible Randy Fenoli with Say Yes to clothing fame for own [url=http://www.fashionhermesbagsuk.co.uk]hermes bags kelly[/url] future screen print variant . Beckham, that may not completely do without having the ornament involving wine red regular sewing bag, plus the tote also continue the style of middle skirt [url=http://www.springhermeshandbags.co.uk]hermes handbags kelly[/url] of tall stomach . All of us who has ever previously reached a telephone call in a odd selection realizes just how shocking it can be when you are getting a good [url=http://www.fashionhermesbagsuk.co.uk]hermes bags birkin on sale[/url] hang up call up . Its golden lock isn't able to only protect itself, but catches me . Hermes is a prominent brand in clothing and luxurious accessories.[/p][p]These kind of Hermes bags tend to be outfitted using glowing man highlights and also signature hardware from Hermes Purses Wall socket shop . zaklyuchenie, Hermes bag, in all its glory, and continues to fascinate and even dreams will warm up and a strong desire . But do not [url=http://www.springhermeshandbags.co.uk]cheap hermes handbags birkin[/url] despair . However, [url=http://www.fashionhermesbagsuk.co.uk]hermes bags[/url] now the replica So Black Bags have been available in the market and actually have become one of the hottest items . It is also made of different materials . They have big types if you actually should pack many different concerns for an overnighter, or tiny dainty types just big adequate to hold your necessities . Merely seize the opportunity end up being the focus regarding picture ahead of time!
To manage anyone favorable totes nicely, the first step would be to adjust clinical cleaning methods.[/p]
[url=http://www.wisebread.com/anonymous-comment-redirection?nid=3825]we sale cheap handbags online 0x[/url]
[url=http://allieiswired.com/archives/2011/03/web-snob-rachel-zoes-new-baby-clinton-kelly-botkier-and-more/]They sale discount jewellery online 4x[/url]
[url=http://www.vectoropenstock.com/preview.php?id=359&error_captcha=1#ancla]we sale discount jewellery online 3k[/url]
[url=http://www.z1z9.com/vb/newthread.php?do=newthread&f=2]we sale discount handbags online 5x[/url]
[url=http://uxdesign.smashingmagazine.com/2009/12/08/principles-of-effective-e-commerce-search/?replytocom=426608]we sale cheap jewellery online 4k[/url]
To manage anyone favorable totes nicely, the first step would be to adjust clinical cleaning methods.[/p]
[url=http://www.wisebread.com/anonymous-comment-redirection?nid=3825]we sale cheap handbags online 0x[/url]
[url=http://allieiswired.com/archives/2011/03/web-snob-rachel-zoes-new-baby-clinton-kelly-botkier-and-more/]They sale discount jewellery online 4x[/url]
[url=http://www.vectoropenstock.com/preview.php?id=359&error_captcha=1#ancla]we sale discount jewellery online 3k[/url]
[url=http://www.z1z9.com/vb/newthread.php?do=newthread&f=2]we sale discount handbags online 5x[/url]
[url=http://uxdesign.smashingmagazine.com/2009/12/08/principles-of-effective-e-commerce-search/?replytocom=426608]we sale cheap jewellery online 4k[/url]
# posted by : March 22, 2013 4:44 AM
: March 22, 2013 4:44 AM
[p]"Starting in 1992, as we committed to cutting and polishing [url=http://www.tiffanysjewellerycanada.com]tiffany jewellery canada[/url] our own diamonds, we were buying directly at the mine head," explained Kowalski, "So we could identify exactly where more of our diamonds were coming from, at a time when the public was rightly horrified by the atrocities going on in Sierra Leona and elsewhere . Without [url=http://www.tiffanysandcocanada.com]tiffany & co canada[/url] any doubt, Tiffany Paper hearts may be the best option in your case . Jewelry is commonly [url=http://www.tiffanysandcocanada.com]tiffany & co rings
[/url] flaunted with any robe and it is the capability to offer you finesse in your persona . Sterling [url=http://www.tiffanysandcocanada.com]tiffany & co jewellery[/url] silver jewelry stays as brand new as at all times and offers a stylish, sexy look every girl wants . Some brides completed [url=http://www.tiffanysandcocanada.com]tiffany rings[/url] their wedding shopping on top-tier wedding day necessities (and some luxuries!)-- at as little as 35 percent of the retail cost . , a company of [url=http://www.tiffanysjewellerycanada.com]tiffany sale[/url] silversmiths . Withal, [url=http://www.tiffanysjewellerycanada.com]tiffany jewellery[/url] Tiffany promises to give affordable and competitive price rate, where you can buy the promising rings to adorn your beauty . The food: delectable.[/p]
[/url] flaunted with any robe and it is the capability to offer you finesse in your persona . Sterling [url=http://www.tiffanysandcocanada.com]tiffany & co jewellery[/url] silver jewelry stays as brand new as at all times and offers a stylish, sexy look every girl wants . Some brides completed [url=http://www.tiffanysandcocanada.com]tiffany rings[/url] their wedding shopping on top-tier wedding day necessities (and some luxuries!)-- at as little as 35 percent of the retail cost . , a company of [url=http://www.tiffanysjewellerycanada.com]tiffany sale[/url] silversmiths . Withal, [url=http://www.tiffanysjewellerycanada.com]tiffany jewellery[/url] Tiffany promises to give affordable and competitive price rate, where you can buy the promising rings to adorn your beauty . The food: delectable.[/p]
# posted by : April 07, 2013 11:14 PM
: April 07, 2013 11:14 PMHornets
While the New Orleans Hornets haven't won much lately, their
woes pale in comparison to those of their next opponent.
The Hornets, losers in five of six, will try to get heading back
in the right direction by earning another road win over the
sputtering Minnesota Timberwolves on Saturday night.
New Orleans (15-32) had won six of seven at one point last
month, [url=http://wholesalenikeshoes.manifo.com/]wholesale nike shoes[/url] but its inability to contain opponents has hurt the team
over the past three weeks. The Hornets have given up an average of
104.8 points while dropping six of eight.
They let Denver shoot 52.4 percent in a 113-98 defeat Friday,
their third in a row, and that virtually negated a 52.8-percent
performance at the other end of the floor. Ryan Anderson had 21
points off the bench but no other teammate scored more than 13.
Eric Gordon had nine points on 3-of-8 shooting after sitting out
Wednesday's 104-99 defeat to Utah. It was a scheduled day off for
Gordon, who missed New Orleans' first 29 games with right knee
soreness and is currently not permitted to play in the second game
of back-to-backs.
The Hornets also gave up 32 points on 22 turnovers Friday, their
most in one game since a season-high 24 in a loss to Philadelphia
on Nov. 7.
"The goal for is ... to get our minds right to play (Saturday)
night," coach Monty Williams said. "We can't sweep this under the
rug by any stretch. We have to understand that you can't turn the
ball over against any team in the league that many times."
Gordon was one of six players in double figures with 16 points
in a 104-92 victory over Minnesota on Jan. 11, and Greivis Vasquez
had a team-best 18 points and 13 assists. Vasquez has 11 or more
assists in six of his last eight games but just missed extending a
three-game streak of double-digit performance Friday as he finished
with nine.
The Hornets could use Vasquez's help to continue their recent
success in Minnesota. They're 9-2 in their last 11 visits and have
won the past two.
New Orleans has also prevailed in four of the last six overall
matchups but lost 113-102 at home to the Timberwolves on Dec. 14.
Nikola Pekovic had a career-high 31 points on 14-of-18 shooting and
Andrei Kirilenko scored 21.
Minnesota, though, enters this game having lost 11 of 12 and a
season-worst six in a row. The Timberwolves (17-26) have also
matched another season high with three consecutive home
defeats.
They cut a 29-point deficit to four with 5:55 remaining on
Friday against the visiting Los Angeles Lakers, but would get no
closer in a 111-100 defeat. In his second game back from a sprained
ankle, Alexey Shved had a team-best 18 points and was one of five
players in double figures.
"It's hard," point guard Ricky Rubio said. "We've been losing a
lot. Even if we make the effort, sometimes it's not enough."
The Hornets are 1-3 on their five-game trip entering Saturday's
finale, while the Timberwolves are 0-2 on their season-long
six-game homestand.
# posted by : May 15, 2013 3:36 AM
: May 15, 2013 3:36 AM
I think that it is quite interesting to check his flaw. but one thing about prediction and decision making. Discussion brings a lot of things.
Post a Comment
<< Home
