Jump to content

Providing independent coverage of the Minnesota Twins.
The same great Twins Daily coverage, now for the Vikings.

The Store


Photo

The Future of Analytics

  • Please log in to reply
17 replies to this topic

#1 jay

jay

    Senior Member

  • Members
  • 906 posts

Posted 06 February 2014 - 11:41 AM

Incredible article on Grantland about how some super smart guys have been able to create a model for the NBA that advances the ongoing efforts to evaluate the game. The model runs off an immense amount of data that tracks the position of each player on the court along with the different potential outcomes. That data is then used to find changes to the Expected Point Value (EPV) for the possession. Some folks criticize the use of expected run values in baseball because they generally go off an average, while individual situations often aren't average. This model goes as far as tracking each player and can then compare them against a more accurate, weighted average.

This type of system is incredibly similar to what could work in baseball to [finally] generate more accurate defense and baserunning stats. My mind started whirling about how it could measure the path an outfielder takes to a ball, the value of hitting the cut-off man, a baserunner's decision making, and so much more.

Give it a read if you can, I'd love to hear other opinions and insights.

http://grantland.com...-nba-analytics/

#2 ashburyjohn

ashburyjohn

    Twins Moderator

  • Twins Mods
  • 4,727 posts
  • LocationLake Tahoe, Nevada

Posted 06 February 2014 - 11:59 AM

the value of hitting the cut-off man


It's exciting. I highlighted the word value because for me that's what the key is. If you can sharpen the estimates of the value a player brings in all the facets of the game, the better your decision making will be. And applying analytics to where the value is highest, for instance in understanding and preventing injuries to prospects, is the most important of all.

#3 TiberTwins

TiberTwins

    Junior Member

  • Members
  • 24 posts

Posted 06 February 2014 - 12:37 PM

It's exciting. I highlighted the word value because for me that's what the key is. If you can sharpen the estimates of the value a player brings in all the facets of the game, the better your decision making will be. And applying analytics to where the value is highest, for instance in understanding and preventing injuries to prospects, is the most important of all.


Could this value you talk about be what the Twins value in scouting? (Pun Intended)

#4 ashburyjohn

ashburyjohn

    Twins Moderator

  • Twins Mods
  • 4,727 posts
  • LocationLake Tahoe, Nevada

Posted 06 February 2014 - 01:53 PM

Could this value you talk about be what the Twins value in scouting? (Pun Intended)


Analytics has a specific1 meaning these days in the computer world, but having an analytic mindset in the traditional sense can't be a bad thing to have in any aspect of decision making, including scouting.

[SIZE=1]1 Well, specific and all over the map, at the same time :)[/SIZE]

#5 thetank

thetank

    Senior Member

  • Members
  • 204 posts

Posted 06 February 2014 - 04:28 PM

If you are 30 games under .500 how can you have as a team an overall WAR of 14? The point is that WAR is not not quite right.

#6 jay

jay

    Senior Member

  • Members
  • 906 posts

Posted 07 February 2014 - 10:41 AM

If you are 30 games under .500 how can you have as a team an overall WAR of 14? The point is that WAR is not not quite right.


I'm not looking for a WAR debate, but even replacement (the R in WAR) level players would win baseball games. Replacement level is defined at Fangraphs as "one who costs no marginal resources to acquire". The expectation for a 0 WAR team is 47.7 wins on a 162 game season.

#7 jay

jay

    Senior Member

  • Members
  • 906 posts

Posted 07 February 2014 - 10:50 AM

Could this value you talk about be what the Twins value in scouting? (Pun Intended)


That's actually an interesting point. Excellent scouting is currently the best way to gauge and measure the many facets of baseball that can't be captured either very well or at all with the stats we have today.

This system, if implemented in baseball, could fundamentally change that. I hope the Twins are paying attention.

#8 ashburyjohn

ashburyjohn

    Twins Moderator

  • Twins Mods
  • 4,727 posts
  • LocationLake Tahoe, Nevada

Posted 07 February 2014 - 11:03 AM

The expectation for a 0 WAR team is 47.7 wins on a 162 game season.


Makes you marvel at how hard the Mets front office must have worked to construct that 40-win team in 1962.

#9 jay

jay

    Senior Member

  • Members
  • 906 posts

Posted 07 February 2014 - 12:11 PM

Makes you marvel at how hard the Mets front office must have worked to construct that 40-win team in 1962.


And those 2003 Tigers who got all the way to 43 by winning 5 of their last 6 games. This also makes me realize I may have under-estimated how bad the Astros have been for 3 years straight with 51 to 55 wins.

The fact that few teams have ever been able to win less than 47 games highlights some measure of accuracy in the replacement level figure.

#10 ashburyjohn

ashburyjohn

    Twins Moderator

  • Twins Mods
  • 4,727 posts
  • LocationLake Tahoe, Nevada

Posted 07 February 2014 - 01:06 PM

The fact that few teams have ever been able to win less than 47 games highlights some measure of accuracy in the replacement level figure.


Sure. Once in a while there can be an outlier. Also, in 1962 talent wasn't available in the same way as now, for a variety of reasons, chief of which would be the reserve clause; it could be that a replacement level team would win 30-35 in that environment.

#11 Thrylos

Thrylos

    Yes

  • Members
  • 4,278 posts

Posted 07 February 2014 - 06:29 PM

Analytics has a specific1 meaning these days in the computer world, but having an analytic mindset in the traditional sense can't be a bad thing to have in any aspect of decision making, including scouting.

[SIZE=1]1 Well, specific and all over the map, at the same time :)[/SIZE]


Yeah. "Analytics" is all over the place and not only in the computer world :)
It became a catch phrase about the same time "big data" became a catch phrase and those two things are intimately related.

Outcome-based predictive analysis (or "decision support" whatever you want to call it) has been around for a while, and its product ("best practices") even longer. Even before computers. Just with computers it is much easier to work with larger data sets and really examine outliers and groups of outliers (and why those might be so) which makes things much more exciting. The sunny day scenario can very easily be determined with pen and paper :)
-----
Blogging Twins since 2007 at The Tenth Inning Stretch
http://tenthinningst...h.blogspot.com/
twitter: @thrylos98

#12 twinsfan34

twinsfan34

    Paul DeVos

  • Members
  • 673 posts

Posted 07 February 2014 - 07:48 PM

If you are 30 games under .500 how can you have as a team an overall WAR of 14? The point is that WAR is not not quite right.


Another factor is that some players will have positive WAR and others will have negative WAR instead of it just 'zero-ing' out.

So the Astros probably had guys who were negatives as well as a few guys who had positive, but probably pretty close to 4 or 5 WAR to add to that 47 WIN mark.

But yes, agree, it's an impressive stat, but not without it's 'flaws' as far as perfection goes.

#13 twinsfan34

twinsfan34

    Paul DeVos

  • Members
  • 673 posts

Posted 07 February 2014 - 07:59 PM

Just checked...

Fangraphs hitters and pitchers = 2.2 + 1.7 = 3.9 WAR

Baseball-Reference = 5.4 and 3.2 = 8.6 WAR


Houston Astros 51 Wins...

so Fangraphs would come out the winner there...

How about the Angels, who have a big time player in Mike Trout...does that mess with the stats?


Fangraphs = 26.3 + 10.4 = 36.7 WAR

Baseball-Reference = 25.6 + 9.0 = 34.6 WAR

Angels had 78 Wins...

Fangraphs WAR would say they had 84 wins, B-R 82 wins.


How about a team like the Cardinals, with a lot of wins...


Cardinals...97 wins..

Fangraphs: 23.3 + 16.8 = 40.1 WAR or 87 Wins...

B-R: 19.9 + 21.5 = 41.4 WAR or roughly 89 Wins...

#14 Hosken Bombo Disco

Hosken Bombo Disco

    not a good guy in the clubhouse

  • Members
  • 1,171 posts

Posted 07 February 2014 - 10:05 PM

I like the 47.7 baseline for a 0 WAR team but in general am uneasy about that particular stat.

At least Fangraphs and Baseball Reference can agree on what a home run is..

#15 jay

jay

    Senior Member

  • Members
  • 906 posts

Posted 08 February 2014 - 01:03 AM

I like the 47.7 baseline for a 0 WAR team but in general am uneasy about that particular stat.

At least Fangraphs and Baseball Reference can agree on what a home run is..


I can understand the angst. The concept of WAR is excellent -- create a singular stat that you can use to compare players that encompasses all aspects of the game from hitting to running to defense to everything.

The catch phrase "you only get out what you put in" holds true today. We don't have the tools and data needed to very accurately measure all of the aspects of the game and then assign individual value to each player from each play. The system proposed in the article would go a long way towards making those aspects more definable and hence improve the accuracy of WAR.

#16 AM.

AM.

    Senior Member

  • Members
  • 363 posts

Posted 08 February 2014 - 05:50 AM

It's an interesting piece on basketball, but having spent many hours trying to figure out WAR contributions of various Twins players, I think it is doing a very good job of factoring in the various contributions of baseball players. It factors in different hitting skills, including on base and slugging abilities. It factors in fielding contributions, albeit still rough, but at least a good estimate. It factors in chances, by weighting positions differently (playing time logged at SS gets an adjustment over RF). It factors in innings-eating pitchers, as well as quality of the pitcher. It factors in leverage of relievers in late game key situations.

The article talks about how box scores credit shooters and assists, and the need to use data to better assign value added. That to me is like crediting RBIs and Wins in baseball, and needing to better assign value added. WAR is doing that.

#17 AM.

AM.

    Senior Member

  • Members
  • 363 posts

Posted 08 February 2014 - 05:53 AM

And thanks for posting the link, jay...good piece.

#18 thetank

thetank

    Senior Member

  • Members
  • 204 posts

Posted 08 February 2014 - 02:02 PM

I wonder if WAR gets boosted if you are playing on a winning team versus knowing you have a long season ahead? Sure you want to keep your stats high to get a better paycheck, but, when your morale is low....