tRA on FanGraphs
Good news everyone. tRA has made its way to FanGraphs thanks to Graham MacAree of Lookout Landing and statcorner.com. tRA is currently located in the player pages under the batted ball section of stats.
Directly from the StatCorner glossary:
tRA involves assigning run and out values to all events under a pitcher’s control and coming up with an expected number of runs allowed and outs generated in a defense and park neutral environment. tRA is on a R/9 scale and does not involve any regression of the rates.
There are a couple things which are different between the StatCorner version of tRA and the version implemented on FanGraphs. The main difference is we’re using Baseball Info Solutions batted ball stats instead of Gameday batted ball stats. The other difference, though probably not as major is we’re using different park factors.
tRA will not be available in the leaderboards or team pages until this winter most likely because it’s going to take a minor overhaul to the code in order to handle the park adjustments.
Graham will be stopping by later to give a better overview of the stat.
David Appelman is the creator of FanGraphs.
Wooooooo. Thank you Graham and David. Fangraphs gets even better.
All right, thanks everyone involved in FanGraphs and Graham for creating the stat! It’s ironic because I was just reading the pitcher wins value thread that Dave Cameron did when discussing that addition and how some people clamored for tRA. Will pitcher win values using tRA at some point be in place?
Huzzah!!!
Could someone give me a quick-and-dirty explanation of why this is better than FIP?
tRA has linear weights for batted ball data. It also jumps through some hoops calculations wise that FIP already does in the formula more or less (I haven’t read up on how Tango came up with FIP), but that’s the gist of it.
It’s also park adjusted.
FIP was good. It was quick and easy and started us on the right path to understanding pitching talent. tRA is a step forward. Like OPS to wOBA, we are improving how we can get a more accurate measurement of talent.
Thanks guys. I appreciate it.
How much more accurate is tRA than FIP?
Hah, Colin, didn’t you just do a study on that? You probably know more about that than anyone.
Right; tRA (scaled to ERA) and FIP had nearly identical RMSEs in split-halves – 1.95 for tERA versus 1.92 for FIP. There really is no evidence that tRA is a more accurate measure of talent than FIP.
Colin, that study was very interesting.
If you’re looking for a stat that tells you the value of his contributions in a given time period (as opposed to trying to get a handle on true talent level), I wonder what are the arguments in favor of tRA over FIP? (This question is not addressed to Colin.) How big of a differences does the batted ball data make?
It’s also worth noting Colin’s comment from his ERA estimators article, that both tRA and FIP are component-based, and as such, abstract from the order in which the various events occur. So if you’re looking for a stat to capture the value of a performance, either could pose a problem.
Good stuff!
Three cheers for Graham!
Very cool. I can’t wait to hear from Graham later going over the intricacies of tRA. I would also be very interested to hear Graham or perhaps another writer compare the respective merits of FIP, xFIP, and tRA.
Also, I’m curious if the site will also list tRA+ or perhaps the league-average tRA for each year. I find tRA a bit difficult to get a handle on because it’s scaled to R/9, unlike FIP, and it’s a bit hard to judge what numbers are good, bad, average for the league.
I just got hard.
Great, great news. I knew this day would come.
Well, this is awesome. I assume this is a step towards making pitch values defense independent.
Niiice, i like it. Now, if we could only get K/PA and BB/PA for hitters and pitchers, it’d be perfect.
I think I speak for everybody when I say “bitchin’…”.
Holy cow this is amazing. Beyond awesome. Wow.
This is great news but we should all proceed with caution…the fangraphs version (using BIS data) and the original tRA version (using gameday data) can goof things up…Erik posted about Aaron Harang’s value today and check this out..
Harang Fangraph tRA 2008: 5.78
Harang Fangraph tRA 2009: 5.28
Harang Statscorner tRA 2008: 5.35
Harang Statscorner tRA 2009: 4.74
ummmm…there’s a sizable difference here folks…..proceed with caution.
The presence of both is going to cause a lot of disagreement. Win Values still use FIP, but a lot of people claim tRA is better.
A problem with BP’s stat section is that they have so many stats that attempt to measure similar things. That’s not what’s happening right now, but I hope it doesn’t happen in the future.
Ah yes, but what does tRA stand for???
[I hope that we get WAR to include tRA instead of FIP as an alternative for those of us who prefer tRA. That would be awesome.]
I always assumed it stood for True Runs Allowed, but I have no clue if that is correct.
Graham once told me that it simply stood for “the Run Average.” A modest name for such an elegant stat!
Yaaaay finally.
Why not park adjusted version though?
Shoot, there needs to be an edit button here or something.
I meant regressed version*
The regressed version is ohmygod annoying to program.
Really? I don’t understand…maybe I don’t understand how the regression is taking place, but isn’t it that you can throw in xRuns at the league average tRA and average that in?
Can someone explain to me why Felix has a lower tRA than Roy? What am I missing when I look at the batted ball #s? Plus park adjusted, I would assume SafeCo is a better place to pitch overall. Not to mention Roy pitches in the AL East fulltime. Thx.
I’d guess a 5% difference in LD rate is really going to push tRA up. Less home runs, more strikeouts…
The only thing Halladay has on Felix is 1 as this stat is calculated is less walk an inning and a better park factor.
Thanks. tRA is ONLY batted ball stats? When I look at there numbers, overall I see Roy has better control of course but very similiar (if not better) numbers across the board. His BB/9, K/BB etc are all second to none. Both are great by an measure, but Roy seems in a class to himself. I still like to use FIP as well, but would love to get a better understanding of tRA before I judge it! Seems odd he is a half run worse than Felix using tRA. I am sure division/league doesn’t get calculated in “officially” but I always take that into account anyway.
I don’t know if I can post a URL here without it getting stuck in a spam filter, so I won’t, but try googling “tra primer” while setting the site to statcorner. You’ll get their “about tRA” page.
tRA does count BB, K’s, and HR’s, but in place of using the same value for all balls in play, it substitutes values for various types of BIPs: groundballs, line drives, outfield fly balls, and infield fly balls.
If Graham drops by (or if anyone else knows). Is there a description somewhere of how stat corner calculates wOBA? I wonder why the pre-park adjusted statcorner wOBA differs from fangraphs wOBA.
IIRC, Statcorner doesn’t include SB/CS in wOBA, while Fangraphs does. Also, Statcorner uses a different method of figuring out the base LWTS, although I don’t know how much of a difference that’s going to make.
Statcorner uses reached base on error as part of wOBA and does not use SB/CS
In addition to what Colin said, Statcorner includes RBOE and adjusts for things like intentional walks and sac bunts.
Does it really include intentional walks in wOBA? Does it use the same denominator and the exact same method of converting wOBA into wRAA (or bRAA on statcorner?)
The nice thing about not including IBBs in wOBA is that when you convert to wRAA using plate appearances you get a custom value for IBBs that is the average value of a plate appearance for that particular player. The value of an IBB is very different from player to player.
So then FanGraphs wOBA and WAR underrates Pujols! I never thought that was even possible.
Yay!
Conceptually, I like that tRA includes batted ball data in the calculation. One of the complaints I hear from traditionalists is that FIP is a “useless” stat because, beyond home runs, it ignores the pitcher’s ability to avoid hard hit balls. We can argue to what degree pitchers have that ability but it’s an issue that always comes up when I try to explain FIP.
On the other hand, I’m not yet convinced that tRA increases the accuracy of ERA prediction enough to make the extra complexity worthwhile. I like that you have it here though because it gives us a chance to study it more carefully side by side with FIP. So, thanks for the addition.