tRA Changed to tERA
As a few of you have noticed, tRA was on hiatus for a couple days. Now it’s back, but as tERA. Quite simply, tRA is on a runs allowed scale and not an earned runs allowed scale, which I believed caused some confusion, especially when I placed it right next to xFIP, which is on an ERA scale.
It seemed the natural thing to do was to compare tRA to ERA/FIP/xFIP and with tRA on a runs allowed scale that didn’t quite work, but now it should.
To convert back to tRA, just divide by .92.
David Appelman is the creator of FanGraphs.
can we get it on the ‘leaders’ page? makes for easy comparison/sorting.
This will come in a later update along with getting wRC+ on the leaderboards.
I’m ashamed that I didn’t even know it was on a RA basis to begin with. I had long noticed the higher values for tRA, so I took it with a grain of salt (or at the least only used it for comparison).
How’s about we kill one of these? tERA vs FIP in a duel to the death. Two metrics enter, one leaves. Too many metrics!
FanGraphs could sell tickets. Pay per view, maybe.
RA is the superior scale. FIP and xFIP should be adjusted the other way. There is no reason for us to continue to rely on the outdated scaling system of ERA.
This is purely a usability thing. For most people ERA is the scale they’re familiar with. I very rarely see anything on a RA scale. I very rarely even see people talk about RA in terms of pitchers, so until RA becomes synonymous with a pitcher’s performance, we’ll stick with ERA.
Unfortunately, David it won’t become influential until you use it. You are the leading edge. RA is not hard to understand, and it’s infinitely more useful. Pitcher X gives up 4 runs per complete game. Very intuitive. Instead, we are stuck with the convoluted earned run designation.
I understand your decision, but I disagree.
Well, I won’t disagree with you that the ER is convoluted, but I think saying RA is infinitely more useful is a bit of a stretch.
You realize a correlation of ERA and RA has a .95 R^2 since 1980. They’re very very similar, though there are always going to be some fringe cases.
Update: I should note that r^2 was with pitchers >= 100 innings in a single season.
Infinitely more useful in the sense that earned runs have no meaning the outcome of a game, ie they are useless. If a pitcher has some “control” over his unearned runs, then why even do we even bother separating the two? Runs are what we care about in a game, season, career.
Agreed.
I’d also appreciate it if this site switched to the metric system and all of the articles were written in Esperanto.
Purity of form is more important than having your readers understand you.
I think you have too little faith in the readers. Sortable leader boards clear up any confusion that may come with an unfamiliar scale. Take for example OPS: it has a completely meaningless definition (ie without units), bizarre scale, and many years ago, no one used it. Now it’s ubiquitous. What changed? Leading promoters of the stat said it was meaningful, and we can all sort descending and agree that it does a good job of sorting out hitters. That’s all it takes for this particular audience, imo.
Pure hilarity! I actually see both sides to this argument. GW – the reason for the switch to OPS wasn’t just that “leading promoters… said it was meaningful”, it was that leading researchers proved it was more integral to a team winning than BA was.
The strong correlation between ERA and RA indicates that we don’t really gain all that much information by going to RA. Whereas with BA vs OPS, we got a LOT more information about a player.
I just dont think the gains outweigh the negatives (less comprehension, means I have to translate numbers when I email them to friends). However, GW, I do wish that we had started with RA, as it’s slightly superior.
Yes, poor choice of words Travis, it was written quickly. To your point, though, there was no shortage of alternative methods to BA even at the time. (Total average and its variants come to mind.) OPS stuck despite having absolutely no frame of reference. RA does not suffer from this problem, it’s runs per game! How much simpler could it be?
When talking to the uninformed about a specific player, just include a note about the relative position to the rest of the league. This is what we all do when a new metric is introduced: sort descending, look for the leaders and the average.
I actually wish we would shift FIP and xFIP to a RA scale. ERA is on a weird scale and tends to overrate extreme ground ball pitchers and guys who don’t miss many bats b/c there will be more errors made behind them. The game is all about runs allowed and runs scored and that is how pitchers should be judged, imo. We can regress a pitcher’s defensive support in far better ways then errors. In the future, I’d like to see RA used over ERA and FIP and xFIP shifted back to a RA scale.
I don’t really see what the problem is. ERA might overrate groundball pitchers, but the scale itself doesn’t. They could put it on any scale, and it would have the same application. It is on the ERA scale, so that someone can look at a pitcher and easily know what 3.00 is in relation to 4.00. For any hardcore sabermetrics, you’re probably going to need to adjust it into a raa or other metric anyway, so the scale doesn’t matter.
I realize we should be trying to change people’s ideas about things, but I don’t think putting something on a slightly more intuitive scale at the cost of possibly turning off people on the fringe between sabermetrics and traditional stats is a good trade off.
As far as RA vs. ERA, RA is obviously the best metric when describing the pitching + defense of a team. I’m not so sure, though, that RA is better than ERA for describing pitcher’s talent. Like I said, RA describes what the pitcher + defense did, and ERA takes us a little bit further toward an FIP like metric. Not very far, but a little bit. Yes, a groundball that Luis castillo doesn’t get to is worth just as much as a groundball through his legs. The thing is, is that the ball through his legs is made by other defenders 98% of the time. A ball that he doesn’t get to is made a much lower percentage of the time, maybe 10%. (Note, I am talking about ALL balls he doesn’t get to, not specifically the one that he narrowly misses. All of these balls look the same on the stat sheet. Also, 10% is a number I pulled out of my ass.) So, if you are trying to separate fielding and defense, it is better to give credit to the pitcher for the ball that almost never gets fielded, and give credit to the fielder for the ball that almost always gets fielded.
I’m not trying to say that ERA is a good stat, as errors are subjective, and the range of the fielders is not taken into account. I’m just saying that if I’m stuck on a desert island with RA and ERA, I’m going to take the metric that at least goes part of the way towards assigning credit to the pitcher and the fielder for their contributions.
RA is not biased against any particular kind of pitcher. ERA is. Certainly, RA is a flawed metric. But, it tells you exactly what happened when a pitcher was on the mound with his defense behind him. ERA doesn’t. Felix Hernandez last year had 15 unearned runs made behind him and the RA gap is far greater than the ERA gap was between him and Greinke. Considering Hernandez is a GB pitcher and will have more errors made behind him, we should partially penalize him for those extra errors.
I think it’s funny that everyone is complaining about the ERA scale. It’s a SCALE people. It’s only real purpose is for usability. Scale it with whatever coefficient you want.
ERA is the right scale for a site that wants to be user friendly because ERA is the most recognizable scale. That doesn’t make it a tacit acceptance of ER vs. un-ER or pitcher’s control over either.
“It’s only real purpose is for usability.”
Agreed. And RA is more usable, it’s just not as widely used. So why not correct the imprecision of the past, however marginal? Especially for an audience that is not only amenable to different ways of describing the game and its players, but also influences the broader public in these new descriptions.
“RA is more usable”
According to who? Again, it’s a scalar coefficient. Get over it.
Anyone interested in tallying runs for any purpose.
“Get over it.”
No thanks.
Poor Graham…
David –
Can FanGraphs include *tRA (or *tERA) in the same was as xFIP is included along with FIP? I find those adjusted versions to be more helpful in certain situations, and would love to have them all in one place!
Any update on whether tERA is going to make it to the leaderboard?
Why have you started removing tERA from the stats pages? Sure, it doesn’t have the predictive value of SIERA, but my understanding has always been that it’s the best indicator of how a pitcher actually pitched.