Pitching staff xFIP

[Patrick425]

You're saying the difference between a good fielder and a bad one is a rare play here or there? Are you taking into account range, or even arm strength?

If you read the rest of my post, you saw that I acknowledged that pitchers with a better defensive behind them tend to have lower ERAs. However, I don't believe that discounting ALL defensive plays is the way to accurately measure this. It's like saying since some QB's have better receivers than others, we are going to completely discount all attempts and completions when assessing QBs and only count times that they score on there own, since that's all that they can truely control.

 

Taking the most extreme example from last season Florida's defense let in 102 more runs than you would expect from an average defense. Colorado let in 58 fewer runs than an average defense. Some of those are from unearned runs but the difference between those two numbers leads to a staggering difference in team ERA.

Wow, that all seems very subjective. "Florida's defense let in 102 more runs than you would expect from an average defense"....How the heck is that measured?? "Some of those are from unearned runs" How many? How many are routine plays that weren't made that were not scored as errors? Probably not very many if any at all. Fielding is so subjective, I sometimes think that peoples attempt to get there hands around it through statistical measures is futile, including attempts to dismiss it completely. One play I'll see a guy like Weeks totally botch an easy grounder. The next play he makes an outstanding top web gem play. How do you account for that?

[TerraceReserved]

Wow, that all seems very subjective. "Florida's defense let in 102 more runs than you would expect from an average defense"....How the heck is that measured?? "Some of those are from unearned runs" How many? How many are routine plays that weren't made that were not scored as errors? Probably not very many if any at all. Fielding is so subjective, I sometimes think that peoples attempt to get there hands around it through statistical measures is futile, including attempts to dismiss it completely. One play I'll see a guy like Weeks totally botch an easy grounder. The next play he makes an outstanding top web gem play. How do you account for that?

 

Those number for Florida were compiled using PBP data. Every ball hit into the field of play is given a specific zone, trajectory, and speed. Then using that information you can see how many balls were turned into outs in every play in the field. Lets say a hard hit ground ball hit in one of the zones between 2nd and 1st is made 60% of the time, and Rickie only gets it 50% of the time, and there were 20 balls hit like that over the year. So then Rickie would be credited with not making 2 plays that an average 2nd basemen would have. Then runs are attributed to those plays. A single on average is worth .77 runs above an out, so thats 1.44 runs that he gave up. To take it farther people on base and outs are then attributed to tweak the value of the hit given up.

[sbrylski]

You're saying the difference between a good fielder and a bad one is a rare play here or there? Are you taking into account range, or even arm strength?

If you read the rest of my post, you saw that I acknowledged that pitchers with a better defensive behind them tend to have lower ERAs. However, I don't believe that discounting ALL defensive plays is the way to accurately measure this. It's like saying since some QB's have better receivers than others, we are going to completely discount all attempts and completions when assessing QBs and only count times that they score on there own, since that's all that they can truely control.

It doesn't remove that component. It sets it to league average, or a neutral environment. If an average defense allows a BABIP of .300, and a pitcher has a .330 BABIP, it removes that generally uncontrolable factor.

 

xFIP is a simplification, definitely not perfect, as pointed out already. But it does a good job of removing luck, natural variation, defense, and park effects. Those are not the only factors that influence ERA, however, but they are large ones.

 

[Patrick425]

Those number for Florida were compiled using PBP data. Every ball hit into the field of play is given a specific zone, trajectory, and speed. Then using that information you can see how many balls were turned into outs in every play in the field. Lets say a hard hit ground ball hit in one of the zones between 2nd and 1st is made 60% of the time, and Rickie only gets it 50% of the time, and there were 20 balls hit like that over the year. So then Rickie would be credited with not making 2 plays that an average 2nd basemen would have. Then runs are attributed to those plays. A single on average is worth .77 runs above an out, so thats 1.44 runs that he gave up. To take it farther people on base and outs are then attributed to tweak the value of the hit given up.

I did read this in Moneyball. At the time I read it, I thought how is this done? Is there a computerized grid laid over the field, then synced with game video to record in what zone every ball was hit to. Or, is it just someone making judgements with their naked eye? And, even if it's being done with a computerized grid of some sort, how does it take into effect different camera angles. Again, seems subjective. Either way, if this data is being gathered, then why not use it to adjust pitchers ERA. It just seems like fip is a "lazy" method to account for the effect of defense on a pitchers ERA. Since there are some really good defensive plays and some bad defensive plays, then lets eliminate all defensive plays?? Why not just account for those "good" and "bad" defensive plays (since it appears that's all be charted out anyways) to adjust an ERA. That would make much more sense to me.

 

Like I said before, any pitching stat that does not give credit to a pitcher for making a pitch that results in a weak ground out or easy pop-up and only gives credit when he Ks a batter seems somewhat faulty imho.

[Patrick425]

It doesn't remove that component. It sets it to league average, or a neutral environment. If an average defense allows a BABIP of .300, and a pitcher has a .330 BABIP, it removes that generally uncontrolable factor.

Ok, I guess that makes a little more sense. As I said in an earlier post, I'm not sure that I understand fip completely. Every explanation of it that I heard of it up until now, pretty much stated that it takes out all fielding plays and only accounts for BBs, Ks, and HRs. Your post, if I'm understanding correctly, makes it sound like that's not the case.

[TooLiveBrew]

xFIP is a simplification, definitely not perfect, as pointed out already. But it does a good job of removing luck, natural variation, defense, and park effects. Those are not the only factors that influence ERA, however, but they are large ones.

 

This is a sound bit of advice. As with any stat, don't trust it on its own, but (x)FIP is a significantly more accurate stat imho in evaluating pitching, since it doesn't gloss over major considerations the way ERA (and ERA+) does.

 

This happens very very rarely. A routine play in my opinion is an easy ground ball hit directly at a fielder. If a fielder misplays it, it goes down as an error. Unless the official scorer is blind or something. People are acting like this happens all the time. It hardly ever happens.

So then the question you have to ask yourself in terms of preference is, 'Since it happens rarely, does that mean I/you/we should penalize a pitcher for it?' If your answer is 'yes', since in your estimation it's a negligible occurrence, then you'll probably be happy to stick with ERA/+, etc.. However, the example provided of the 2007 Marlins v. 2007 Rockies (the 2007 Brewers certainly fall in as a strong example, too) seems to indicate that this happens much more than 'hardly ever.'

[sbrylski]

It doesn't remove that component. It sets it to league average, or a neutral environment. If an average defense allows a BABIP of .300, and a pitcher has a .330 BABIP, it removes that generally uncontrolable factor.

Ok, I guess that makes a little more sense. As I said in an earlier post, I'm not sure that I understand fip completely. Every explanation of it that I heard of it up until now, pretty much stated that it takes out all fielding plays and only accounts for BBs, Ks, and HRs. Your post, if I'm understanding correctly, makes it sound like that's not the case.

It only uses BBs, Ks, and HRs. But it assigns a value to them based upon a neutral environment. I guess that's where the confusion is.

[TooLiveBrew]

And keep in mind that "neutral environment" includes an average defense factored in behind the P.

[Ennder]

It doesn't remove that component. It sets it to league average, or a neutral environment. If an average defense allows a BABIP of .300, and a pitcher has a .330 BABIP, it removes that generally uncontrolable factor.

Ok, I guess that makes a little more sense. As I said in an earlier post, I'm not sure that I understand fip completely. Every explanation of it that I heard of it up until now, pretty much stated that it takes out all fielding plays and only accounts for BBs, Ks, and HRs. Your post, if I'm understanding correctly, makes it sound like that's not the case.

 

It isn't taking the plays out, it is assuming a completely average number of them are made. The good thing about xFIP is that it correlates well from year to year. You are basically pulling all of the luck out of ERA. The bad part is it doesn't take into account things like GB%, pitchers who have above average LOB% year after year etc. qERA is a little better version of FIP that does include GB% but I haven't found a place to get that stat on a regular basis.

This entire branch of stats is relatively new and they are doing a lot of work on it but it has been shown that you can predict future success better by looking at stats like K/9 than you could by looking at ERA. ERA is extremely luck based.

[logan82]

You know, this is a good thread to have under Statistical Analysis for future reference. Of course I don't think as many people would see it there.

[TooLiveBrew]

Ha! I was thinking this earlier, logan... "why is it that most of the good stat discussions take place in the ML forum?"

 

If only more traffic would go there. Oh well, this is a reasonable substitute.

[logan82]

I think this fits perfectly under "good place to archive stat oriented threads" that Russ talked about in that thread.

[Patrick425]

Well, I know these stats are not put together by 7th graders, but by guys who are much smarter than me. So, there must be some validility to them. I just have a hard time wrapping my simple mind around the concept.

ERA is extremely luck based.

However, I will always think that a statement like this is an overstatement. There might be some luck involved in ERAs, but I don't think it's extremely luck based. That's like saying it's not a valid stat at all, or that Jake Peavy consistantly posts lower ERAs every year than Bret Tomko because of extreme luck.

However, the example provided of the 2007 Marlins v. 2007 Rockies (the 2007 Brewers certainly fall in as a strong example, too) seems to indicate that this happens much more than 'hardly ever.'

My impression is that the difference between those two teams was not because of "routine" plays that were not made by the Marlins that were not recorded as errors. The arguement made before was that Braun does not make easy plays and they are not recorded as errors. I questioned the frequency of this occurance, and I still do. If there were really 100+ routine plays not made and not recorded as errors then the Marlins might want to look into getting a different official scorer.

[TooLiveBrew]

That's like saying it's not a valid stat at all, or that Jake Peavy consistantly posts lower ERAs every year than Bret Tomko because of extreme luck.

 

Well, it's probably more accurate to look at how much significantly better Peavey has been than Tomko/others in the non-ERA categories -- which is part of what FIP attempts to address.

 

I questioned the frequency of this occurance, and I still do

 

Well, to be sure the numbers are highly subjective, but when there's a gulf as wide as the FLA-COL gap, you do have to admit there's (at worst) some relevance there. You're right to remain cynical, especially since defensive metrics are still very new - but don't just toss it out since it seems unlikely to you. Check it out, and if you still feel as strongly that there's too much weight being thrown on it, go with numbers that make more 'real-world' sense to you.

 

All I can say for (x)FIP is that I've found it much more useful than ERA/+, but it's not like ERA/+ is the only other choice.

[rluzinski]

However, I will always think that a statement like this is an overstatement. There might be some luck involved in ERAs, but I don't think it's extremely luck based. That's like saying it's not a valid stat at all, or that Jake Peavy consistantly posts lower ERAs every year than Bret Tomko because of extreme luck.

 

As the sample increases, a stat like FIP doesn't tell you anything more than a stat like ERA. In fact, the opposite it true. Where FIP is very is small sample situations, like a year's worth of stats from a reliever. It assumes a league average rate for the most volatile stats (BABIP, LOB%, etc...). While that's not completely corret, it's better than assuming a pitchers's BABIP over 50 IP perfectly represents his talents.

[Ennder]

That's like saying it's not a valid stat at all, or that Jake Peavy consistantly posts lower ERAs every year than Bret Tomko because of extreme luck.

 

Tomko - 2004 - 4.04 ERA. 5.1 K/9, 3.0 BB/9, 42% GB. xFIP 4.79. That year Tomko had a lot of luck with BABIP and HR/FB and it drove his ERA way down. Overall he pitched almost exactly like normal except some luck.

 

Peavy - 2006 - 4.09 ERA. 9.9 K/9, 2.8 BB/9, 38% GB. xFIP 3.82. Peavy had a slightly lower GB% than normal and a little bad luck with HR's.

 

Looking at those 2 stat lines and judging by ERA they are the same pitcher. xFIP clearly shows that Tomko wasn't nearly as good as him though and showed that if you were say handing out contracts that Tomko's 2004 was a fluke, not a new level for him. Looking at a few other stats along with xFIP would have shown even more support for Peavy not being a fluke and I drafted him fantasy without a worry in the world in 2007.

[Fatter than Joey]

Lets basically just look at Walks, HRs, and K's...huh?

 

Well -- that's not really it -- As other pointed out -- FIP is a stat that tries to predict what is going to happen in the future, counting stats just cannot do that, for example, Jeff Suppan went 12/12 last year in W/L -- what does that tell us about 2008? Nothing. Certainly it could give you some insight as to how he pitched in 2007, but it does nothing for 2008.

 

In order to generate a predictive stat, you have to remove as many variables as you can -- such as:

 

1.) Fielding -- Don't focus entirely on errors/non-errors -- as there are other things as well. Yostian shifts, super plays by guys like Scott Rolen, park dynamics, like the wall at Fenway, fast infields, etc.

2.) Who a pitcher faces, generally teams with high payrolls, are going to have guys that when they hit the ball, are going to hit it harder (AL East v. NL Central).

3.) The bouncing of the ball. Some days 8 line drives are outs and 8 ten-hopper groundballs are hits.

 

As Ennder points out ERA is a horrible stat to use for RPs -- If a guy like Brian Shouse gives up runs of his own, he probably has already given up someone else's runs. Plus they just don't get the IP that SPs get to even things out a bit.