Economics Job Market Rumors Topic: question about baseball statistic WAR
https://www.econjobrumors.com/topic/question-about-baseball-statistic-war
Economics Job Market Rumors Topic: question about baseball statistic WARen-USThu, 11 Aug 2022 23:10:17 +0000http://bbpress.org/?v=1.0.2<![CDATA[Search]]>q
https://www.econjobrumors.com/search.php
Economist on "question about baseball statistic WAR"
https://www.econjobrumors.com/topic/question-about-baseball-statistic-war#post-538971
Thu, 11 Oct 2012 23:19:16 +0000Economist538971@https://www.econjobrumors.com/<p>Does anyone know of a good book to start with on this subject?
</p>Economist on "question about baseball statistic WAR"
https://www.econjobrumors.com/topic/question-about-baseball-statistic-war#post-538963
Thu, 11 Oct 2012 23:05:27 +0000Economist538963@https://www.econjobrumors.com/<p>Oh really? Could you elaborate or point me to where it's described
</p>Economist on "question about baseball statistic WAR"
https://www.econjobrumors.com/topic/question-about-baseball-statistic-war#post-538841
Thu, 11 Oct 2012 21:06:19 +0000Economist538841@https://www.econjobrumors.com/<p>WPA actually uses markov chains.
</p>Economist on "question about baseball statistic WAR"
https://www.econjobrumors.com/topic/question-about-baseball-statistic-war#post-538817
Thu, 11 Oct 2012 20:37:41 +0000Economist538817@https://www.econjobrumors.com/<p>Also, does anybody know if there is a command in Stata to obtain players' war?</p>
<blockquote><p>Sabermetrics folks are all about this WAR stat. My understanding is that it is a linear combination of a player's basic statistics. The vector that the stats are multiplied by is an estimation of the win probability added by each of his stats (i.e. an RBI improves your team's chance to win by .03 so if you have 100 RBI you get 3). But this win probability added vector is itself based off of estimation (someone correct me if I'm wrong). As in it's probably from a regression with dep. variable whether a team wins and independent variable the # of RBIs - maybe a bit more complex but that's the idea.<br />
My question is, clearly there ought to be a standard error associated with each of the win probability added components of the vector. Even if every component is statistically significant, some are more significant than others I'm sure. To then put a bunch of estimates in a vector and multiply this vector by something else, wouldn't you have to account for the standard error of the final statistic. As in, Mike Trout's WAR is 10.5, but maybe his standard error is higher than someone else whose WAR is 10.5 because a lot of Trout's war comes from his high stolen base count, whose component in the win probability added vector has a high standard error.</p></blockquote>Economist on "question about baseball statistic WAR"
https://www.econjobrumors.com/topic/question-about-baseball-statistic-war#post-538807
Thu, 11 Oct 2012 20:25:02 +0000Economist538807@https://www.econjobrumors.com/<p>also, why is rainfall as an iv so frowned-upon in sabermetrics?
</p>Economist on "question about baseball statistic WAR"
https://www.econjobrumors.com/topic/question-about-baseball-statistic-war#post-538788
Thu, 11 Oct 2012 20:02:04 +0000Economist538788@https://www.econjobrumors.com/<p>Sabermetrics folks are all about this WAR stat. My understanding is that it is a linear combination of a player's basic statistics. The vector that the stats are multiplied by is an estimation of the win probability added by each of his stats (i.e. an RBI improves your team's chance to win by .03 so if you have 100 RBI you get 3). But this win probability added vector is itself based off of estimation (someone correct me if I'm wrong). As in it's probably from a regression with dep. variable whether a team wins and independent variable the # of RBIs - maybe a bit more complex but that's the idea. </p>
<p>My question is, clearly there ought to be a standard error associated with each of the win probability added components of the vector. Even if every component is statistically significant, some are more significant than others I'm sure. To then put a bunch of estimates in a vector and multiply this vector by something else, wouldn't you have to account for the standard error of the final statistic. As in, Mike Trout's WAR is 10.5, but maybe his standard error is higher than someone else whose WAR is 10.5 because a lot of Trout's war comes from his high stolen base count, whose component in the win probability added vector has a high standard error.
</p>