ERA vs FIP (and BABIP too!)

[homer]

I was checking out some of the Cubs' starting pitching numbers and a few of them have some really stark differences between FIP and ERA. To wit:

 

Hendricks ERA 3.59 FIP 4.46

Lester ERA 2.22 FIP 3.98

Chatwood ERA 3.86 FIP 4.98

 

The BABIPs for these three chaps are .245, .238, and .299 respectively. I believe league average is around .300 for pitching BABIP.

 

So does that mean that Hendricks and Lester are getting really lucky on balls in play or could it mean that the defense behind them is really, really good? And if the latter is the case how do you find that amongst the stats?

[JosephC]

Really not an issue with Hendricks IMO. He's not a big strikeout pitcher so it would be expected that his FIP would be behind his ERA. So far in his career his ERA is more than 0.50 less than his FIP, so he has a considerable history of outpitching his FIP. Don't see it as an issue for him and I wouldn't expect a performance drop-off unless he suffers an injury.

 

Lester's 3.98 FIP is still a pretty solid number. The 2.22 ERA is not sustainable but I don't think he'll blow up to a 4.00 ERA this year either. From this point forward I'd expect the ERA to go up, the FIP to drop and he'll probably end up in the low 3's for ERA at the end of the year which is more than good enough.

 

Chatwood is the guy they should be worried about because his WHIP is bad and his walk rate is ridiculously bad. No way he will sustain a sub-4 ERA if those numbers continue. His BABIP number is better than league average but not crazy and could be sustainable. The HR/FB ratio is low, and if that corrects with as many guys as he's been walking, then the ERA will balloon rather quickly. I'm actually somewhat shocked that Chatwood's FIP is as low as it is when considering the walks, but he's done a pretty good job (or has just been lucky) of avoiding the home run ball.

[sveumrules]

Cubs currently rank 3rd in DRS, 2nd in UZR & 1st in defensive efficiency converting 73.9% of all batted balls into outs, so excellent team defense is definitely a contributing factor.

[sveumrules]

Just to add a little more context, there are six teams currently outperforming their FIP by more than a quarter of a run. I've listed them with their DRS, UZR & efficiency rankings noted...

 

Cubs (3.18 ERA 4.00 FIP -0.82) 3rd/2nd/1st

DBacks (3.43 ERA 3.92 FIP -0.49) 2nd/5th/7th

Brewers (3.53 ERA 4.00 FIP -0.47) 1st/3rd/2nd

Oakland (3.87 ERA 4.29 FIP -0.42) 12th/6th/3rd

Braves (3.68 ERA 4.05 FIP -0.37) 8th/8th/9th

Angels (3.65 ERA 4.02 FIP -0.37) 5th/4th/15th

[homer]

Thanks.

 

Fangraphs just had an article on Brewers success in 1 run games.

 

Here is the blurb about the defense:

The team’s 30-point improvement in defensive efficiency from 2017 (.684, eighth in the NL) to 2018 (.714, second behind the Cubs’ .719) is the league’s biggest jump. Newcomer Lorenzo Cain, the key to the team’s defensive turnaround, leads all center fielders in both UZR (4.7) and DRS (12), while Arcia is second among NL shortstops in UZR (3.8) and tied for first in DRS (9). Travis Shaw is tops among NL third basemen in DRS (7) and, in right, Santana has improved from -4.9 UZR and -5 DRS to 2.9 UZR and a league-high 7 DRS. With strong defensive contributions such as those, the position players’ standing in terms of WAR (9.3, fourth) is higher than in terms of wRC+ or scoring. Cain (2.5) ranks third in the league behind Freddie Freeman (3.2) and Nolan Arenado (2.6), while Shaw (2.0) is tied for 14th with Max Muncy (!).

 

https://www.fangraphs.com/blogs/the-brewers-and-one-run-success/

[sveumrules]

Yeah, Domingo's defensive numbers this season are definitely out of left (right?) field.

 

Stuff like that really makes me wish that DRS and/or UZR provided some kind of play log or opportunity log so we could have some kind of idea of how much is being credited or deducted with each individual opportunity.

 

For instance I'd love to know how DRS puts Braun at +28 for his career in LF when UZR has him at -23.

[Oxy]

The greatest flaw in advanced pitching stats is not accounting for situation. Pitchers have entirely different approaches to a batter depending on the number of outs, the guys on base, the quality of the batter and the quality of the batter(s) on deck and in-the-hole.

 

I'm sure that smart front offices like the Brewers' are well on top of this kind of thing, but I'm not sure that Fangraphs is. (I'm not sure that Fangraphs isn't, either).

[Lathund]

The greatest flaw in advanced pitching stats is not accounting for situation. Pitchers have entirely different approaches to a batter depending on the number of outs, the guys on base, the quality of the batter and the quality of the batter(s) on deck and in-the-hole.

 

I'm sure that smart front offices like the Brewers' are well on top of this kind of thing, but I'm not sure that Fangraphs is. (I'm not sure that Fangraphs isn't, either).

 

The problem with trying to fit all these things into one or a few different stats is that it's really, really hard to quantify most of these things. Or to get big enough sample sizes. Or to weigh their importance against eachother. I'm not saying you can't make stats like these, and indeed there are stats out there that try to do this: Baseball Prospectus' Deserved Run Average (DRA) is one. It adjusts for park, league environment, batter (quality and handedness), catcher, umpire, defense, temperature, pitch types and anything you could imagine. And it does have a better predictive value for future ERA than FIP, and of course far better than ERA itself. (However I don't think that'd go over well on this website, as it has Oliver Drake as the 4th best pitcher on the Brewers behind Perez, Hader and Peralta, making the Drake haters seriously question its validity )

 

A stat like FIP is simple, easy to caclculate, has good predictive value, and is easy to understand. High K% and low BB% tends to give up fewer runs than the other way around. It does a good job of isolating what the pitcher alone can control. But doesn't account for how the pitcher interacts with the park, the weather, the defense or quality of contact. Adding these things onto the stat is the hard part; even something seemingly as simple as quality of contact. Is it exit velocity alone, or based on EV combined with launch angle? How does in which direction it's being hit factor into it? And is a hard ball hit into the shift any better or worse than a weak one away from it? Average EV or median? Distribution of artibtraily decided soft/medium/hard %?

 

What I'm getting at I guess is that there's a choice between precise but limited (FIP) or imprecise but extensive (DRA). DRA is already good, and will likely improve. But there's the problem of not really understanding what the number means. It's like a black box spitting out a number, but you don't know how you got there or what it's based on even if it's accurate. To take an example: Knebel has a 3.38 ERA, 4.38 FIP. Houser has 1 ERA, 2.76 FIP. They have the same DRA (3.20 vs 3.29). Now in that particular case it's likely to do with leverage/quality of batter and HR/FB rate (Knebel isn't going to continue having 40% of his flyballs go for HRs, nor is that going to stay at 0% for Houser), but that's because we know these other stats. It's an improvement, it can at a glance compare pitchers both in results and with some predictive value. But since you still need other context to understand it, it doesn't offer as much beyond (x)FIP as you'd hope.

 

So it's a bit like the greatest supercomputer in the universe spitting out "42" as the answer to the great question about the meaning of life, the universe and everything. Great, but what is the question?

[zurch1818]

I know I'm a little late to the party, but I did just look up the 2018 woba/2018 xwoba/2017 xwoba for those 3 pitchers.

 

Chatwood 0.345/0.369/0.331

Lester 0.294/0.352/0.308

Hendricks 0.303/0.341/0.306

 

All of them have a much higher expected than what they are currently performing at. They also have a lower expected last year than this year's.

 

I do question how xwoba is calculated...very black box. I also think defense could play a role. Hard to quantify. Past performance does not always predict future performance.

 

All I know is I would look to trade all 3 pitchers in fantasy. Those numbers do scare me. Luckily I enforce a rule of having no Cubs or Cardinals on my team.

 

The guy to buy on is Jon Gray.

[Oxy]

The greatest flaw in advanced pitching stats is not accounting for situation. Pitchers have entirely different approaches to a batter depending on the number of outs, the guys on base, the quality of the batter and the quality of the batter(s) on deck and in-the-hole.

 

I'm sure that smart front offices like the Brewers' are well on top of this kind of thing, but I'm not sure that Fangraphs is. (I'm not sure that Fangraphs isn't, either).

 

The problem with trying to fit all these things into one or a few different stats is that it's really, really hard to quantify most of these things. Or to get big enough sample sizes. Or to weigh their importance against eachother. I'm not saying you can't make stats like these, and indeed there are stats out there that try to do this: Baseball Prospectus' Deserved Run Average (DRA) is one. It adjusts for park, league environment, batter (quality and handedness), catcher, umpire, defense, temperature, pitch types and anything you could imagine. And it does have a better predictive value for future ERA than FIP, and of course far better than ERA itself. (However I don't think that'd go over well on this website, as it has Oliver Drake as the 4th best pitcher on the Brewers behind Perez, Hader and Peralta, making the Drake haters seriously question its validity )

 

A stat like FIP is simple, easy to caclculate, has good predictive value, and is easy to understand. High K% and low BB% tends to give up fewer runs than the other way around. It does a good job of isolating what the pitcher alone can control. But doesn't account for how the pitcher interacts with the park, the weather, the defense or quality of contact. Adding these things onto the stat is the hard part; even something seemingly as simple as quality of contact. Is it exit velocity alone, or based on EV combined with launch angle? How does in which direction it's being hit factor into it? And is a hard ball hit into the shift any better or worse than a weak one away from it? Average EV or median? Distribution of artibtraily decided soft/medium/hard %?

 

What I'm getting at I guess is that there's a choice between precise but limited (FIP) or imprecise but extensive (DRA). DRA is already good, and will likely improve. But there's the problem of not really understanding what the number means. It's like a black box spitting out a number, but you don't know how you got there or what it's based on even if it's accurate. To take an example: Knebel has a 3.38 ERA, 4.38 FIP. Houser has 1 ERA, 2.76 FIP. They have the same DRA (3.20 vs 3.29). Now in that particular case it's likely to do with leverage/quality of batter and HR/FB rate (Knebel isn't going to continue having 40% of his flyballs go for HRs, nor is that going to stay at 0% for Houser), but that's because we know these other stats. It's an improvement, it can at a glance compare pitchers both in results and with some predictive value. But since you still need other context to understand it, it doesn't offer as much beyond (x)FIP as you'd hope.

 

So it's a bit like the greatest supercomputer in the universe spitting out "42" as the answer to the great question about the meaning of life, the universe and everything. Great, but what is the question?

 

This was an excellent post, but basically ignored my post.

 

Statistics ^mostly^ treat every AB the same and judge a player according to that standard. However, pitchers themselves DO NOT treat every batter the same way. Sometimes, they pitch to contact and sometimes they try to strike a guy out. It all depends on the score and the situation. Some guys are better at manipulating this situation than others (If you are up by 5 in the 3rd inning, giving up a leadoff HR is ACTUALLY better pitching than giving up a leadoff BB. If a guy is trying to K a guy, he is probably more likely to BB a guy as well...hmmm....think about those concepts and then think about the validity of FIP and K/BB ratio...)