Introducing TRX

As a lover of sabermetric pitching analysis, two of my favorite stats are unsurprisingly xFIP and tRA. xFIP wisely takes into account the factors that a pitcher has control over, while tRA looks at the “quality” of batted balls given up. Both are summed up into nicely-packed numbers that resemble ERA (although one must multiply tRA by .92 to get tERA). In a mild attempt to see what happens when you mix two completely different statistics, I entered my laboratory and came up with a crazy concoction: TRX. It’s literally just the average of each starting pitcher’s xFIP plus tERA. Can one simply look at both stats in gauging a pitcher’s performance? Sure, but isn’t it more fun/interesting/easy to have one number? The sample size used is the the seventy-five starting pitchers that come up on Fangraphs for league leaders. Here are the results:

2009 Top 10 TRX

Tim Lincecum    2.74
Justin Verlander  2.92
Zack Greinke    3.03
Chris Carpenter  3.08
Javier Vazquez   3.10
Jon Lester      3.22
Felix Hernandez  3.24
Roy Halladay    3.24
Josh Johnson    3.27
Adam Wainwright 3.32

Nothing all too surprising, although Greinke a “distant” third in any ranking is odd, and we get to see how good Verlander really is. According to TRX, the bottom five pitchers in baseball last year were Jeremy Guthrie (worst), Braden Looper, Doug Davis, Trevor Cahill, and Kevin Millwood. I ultimately think this could be a nice quick-and-easy way to evaluate a pitcher for those who love DIPS theory, but also value knowing just how the ball came off the bat.

TRX has an R^2 value of .45 when compared with ERA (similarly, FIP has a .47 correlation and xFIP is .31, both based on the same seventy-five pitchers used).

trx.xls


Print This Post
Pat Andriola is an Analyst at Bloomberg Sports who formerly worked in Major League Baseball's Labor Relations Department. You can contact him at Patrick.Andriola@tufts.edu or follow him on Twitter @tuftspat
Sort by:   newest | oldest | most voted
drew
Guest
drew

I already do this to evaluate pitchers, except I use tERA* and I add in qera as well.

drew
Guest
drew

(xfip + qera + tERA)/3

Lucas A.
Guest
Lucas A.

I like this!

dan
Guest
dan

Another option to consider.  Are you planning on making this available?

Pat Andriola
Guest
Pat Andriola

Dan,

I’ve attached the excel file with the 75 pitchers from 2009 in the link at the bottom of the article.

Red Sox Talk
Guest
Red Sox Talk

Pat, thanks for the post. I too have thought about consolidating things to try and minimize the weaknesses of any one given method.

In my own projection system, I also use a similar method, but I have elected to include historical ERA in there as well. There are some pitchers that FIP and tERA just can’t peg well, like Javier Vazquez.
http://www.beyondtheboxscore.com/2009/12/22/1212557/yankees-reacquire-javier-vazquez
http://fantasybullpen.com/fcg/javy-vazquez/

Patrick
Guest
Patrick
Urk! Pat, No!  TRA DOES include SO, BB, and HR, as well…  It’s a superset of the data in FIP. So here’s what you’re actually doing with this stat, as opposed to actually combining seriously different sources of information: You’re taking two stats with empirically derived coefficients on their variables and combining them.  But they SHARE most of the variables, and in fact TRA uses a superset of those in FIP! The net effect here is to very heavily weight the components that are in both – SO, BB, and HR, though in the case of HR, you’re using the… Read more »
wpDiscuz