The probability that a pitcher outperforms his FIP to the extent that Matt Cain does each year, is pretty small, and is nowhere close to 50%.

Fangraphs doesn’t have an ERA-FIP stat, so I’ll use BIP-wins as proxy. In 2012, Cain was 4th out of 88 pitchers for BIP wins. Assuming a Uniform distribution (which probably isn’t true, it’s most likely normal), he has a 4.5% chance of being 4th or higher.

In 2011, he was 8th out of 94 or 8.5%. In 2010, he was 4 out of 92 or 4.3%. Since I’m lazy, I’m just going to average all three seasons and say a player has a 5.8% chance of posting BIP-wins that high from random fluctuation.

The probability of doing this 7 years in a row?

0.000000002

Now, for this to be Matt Cain being an extreme example because so many pitchers have entered the league, we would have to have seen about 10 million pitchers enter the league over the last 7 years.

And this is using a uniform distribution. If I used a normal distribution, the probability would be even lower.

]]>As for ERA vs FIP, you bring up Matt Cain, which is a fabulous example of a pitcher with a relatively normal pitch repertoire consistently outperforming the ERA that his FIP says he should generate (not outperforming his FIP, FIP is FIP) However, when we look deeper into his peripherals, the reasons become evident. Yes, Matt Cain does appear to have a better than expected BABIP which is probably due to the scouting reported and pitch F/X supported theory that he gets exceptionally good late movement on his pitches, particularly his fastball, but the biggest difference comes when we look at his home vs away HR/FB% which pretty clearly shows that his home ballpark (where he has pitched just north of 50% of his games) helps him drastically limit his homers allowed while giving up the same amount of fly balls. This gives us very good evidence that if Matt Cain pitched in, say, Fenway, he wouldnt have nearly the same career ERA – FIP split. We arent “getting rid of stats because they dont work for one person” we are using all of the other data we have to try to figure out why it doesnt work for that one person.

]]>http://postimage.org/image/hjllo9cej

Sample was small, basically just the 10 knucklers with the most games started + Dickey and minus Cicotte since he pitched a century ago.

]]>We say that Matt Cain has made it past the “lucky” stage now, 7 years into his career. But a coin will land heads 7 times in a row with a probability of 1/128. There have been many, many pitchers that began careers over Cain’s time frame. That’s not an insignificant probability.

There’s obviously the magnitude of his over-performance that should be accounted for. I was only hoping to use simple probability to shed light on the possibility of the survivor bias, not to attack Matt Cain, specifically.

]]>I was treated to the usual “you just don’t understand” BS from the sabre guys I was talking too, but I thought what I was saying was as clear as day. Pitchers who threw certain pitches, or pitched in certain areas (high and tight, low and away, or on the very inside of the plate) would induce weak contact, and would have a naturally lower BABIP.

I just randomly pulled 3 guys off the top of my head, Moyer, Glavine, and Maddox. I checked on baseball reference and low and behold they had lower BABIP’s than the league average by about .010. I then went on to look up knuckleballers, and it was even more extreme than that.

I finally came across Tom Tippet’s article, and I was confused because it was from 2003. Obviously someone else thought about what I was saying way before, and had proved it to be true. Yet no one had come up with a different or revised stat for showing this.

I don’t believe this phenomenon applies only to knuckleballers, but also to players with exceptional changeups, inside movement, and the prototypical “crafty lefty” nibblers. I’m sure there are other instances as well that it applies as well. I think looking into building a second WAR, or different stat all together for pitchers might be a good idea.

]]>I don’t think Dickey will require much adjusting to the dome.

]]>