Originally Posted by mlbfan30
It's amazing how behind some of you could be at math. 40% of a sample size is enough data to notice tendencies of that pitcher. All if is is a fraction of the total, but the composition of that fraction would be roughly the same no matter how large of the sample size. You don't need pitches from every start to get the general idea of the way he throws. It's not like he would go from 70% fastballs to 55% fastballs in 1 year. It just doesn't happen. The range of error increases with a smaller sample size, but the 400 or so pitches is enough data to get a range of +- 2% I would say. That's the difference of throwing 68 or 72 fastballs per 100 pitches, which when you think about it is faily significant over a full season. But say if every type of pitch was +-3%, that is still enough info to find out the general idea of what he throws. Speed would probably have a smaller range of error.
Sample size is important, but the context matters a lot. 80 IP is generally (2.5 months) enough to see how good a pitcher is. You would most likely expect similar peripherals for each 80 IP increment of a pitcher say over 1 year. That's also 40%
80 IP is good enough to see how good a pitcher is? No. No way. I'm not sure you're considering the impact of randomness in the game of baseball. Less than half of a full season might indicate something about behavioral tendencies, but it says little about actual performance quality.
Here's a data set involving 56.2% of a MLB hitter's season in 2006:
.275 BA/.353 OBP/.445 SLG
And here's another from 2007 (50.99% of the season):
.270 BA/.353 OBP/.391 SLG
Here's a data set from 44.7% of a pitcher's season in 2007:
4.04 ERA/1.47 HR per 9/10.89 K per 9/2.02 BB per 9
The first player is Mark Teixiera from pre-ASB 2006. The second is Edwin Encarnacion from pre-ASB 2007. The last player is Johan Santana from Post-ASB 2007.
I didn't look hard to find those samples. They were the first three players that popped up in my head. Single-season samples above 40% don't necessarily clue us in to actual propensities or ability. And you don't have the data to tell us that a player can't possibly go from 70% Fastballs to 55% fastballs- particularly when working with lower sample sizes. If you do have the data to tell us that, then by all means show your work. Otherwise, that's an intuitive reach on your part, and it has nothing to do with math (good, bad, or otherwise). You're guessing. That's not math. It's lazy.
Also saying it fires blanks isn't really a big deal at all. Say it misses 50 pitches. The compositions of those pitches would be roughly the same as that of the entire 2000 sample size. If it misses only a few, it might affect a sample size of less than 100, but at 500+ pitches the 5 that it missed is only 1%
Do you know the margain for error for tracked f/x pitches or the actual percentage of pitches the system misses? If not, then your claims of being a superior mathemetician really don't matter because your methodology starts from a position of assumption rather than fact. Sorry, but I'm not really interested in that kind of pseudo-logic.