Stats stats stats... any geeks want to share tricks?

[bjs73]

In my opinion, Batting Average and Splits are probably the worst thing you can use to make a decsion on keeping a hitter or not. I'm not saying that they're worthless though. But basing decisions off of that data ALONE can be self destructive. I like using BA and Splits to help finalize my decision on cards after I've alreay analyzed it in other ways. The way I do it isn't too difficult and time consuming but I have to say that I still say that 100 PA's minimum is what you need. Lately, I haven't been cutting very many hitters until after game 60 especially if they're my studs.

The blog article I have on what I do to predict hitters cards works fairly well and I usually try to point new people to it for study. However, I've found that it helps to give a visual example of what I've written. The following is a private topic that I am copying and pasting here from my records. It gives a good visual example of what I do:

[i:93cb00eb27]Copied from my Private Messages:[/i:93cb00eb27]

Question: I read your blog on taking samples from 100 ab. If I understand it correctly, a player with 100 ab can have those results x6 to give a rough estimate of how well they are most likely to do the rest of the season?

Answer: That's close but not quite it. I'll use your Singleton card to better express my theory. You really need to have at least 100 PA or better for a reliable statistical study of the trends. The distributions over 100 PA's on a statistical bell curve will begin to be more reliable since they will help negate the anomolies that occur. That being said, I track a position players 2B's and K's categories in making my predictions.

Here is how we do the math on Singleton:

1 double divided by 111 AB's = .091 rounded up. Take that number and multiply it by the exact number of AB's on each of Singleton's single season figures. Problem is that 1 double kind of sucks at this point because it doesn't really match up with much of anything so let's look at his K rate instead. (His lack of doubles doesn't point to anything except for bad luck. A classic example of why having more data is better.)

35 K's divided by 111 AB's = .315. Take that number and multiply it by the exact number of AB's on each of his single season cards:

1977 = 169 Projected vs. 101 Actual - Variance of 68
1979 = 180 Projected vs. 108 Actual - Variance of 72
1980 = 184 Projected vs. 94 Actual - Variance of 90
1981 = 114 Projected vs. 61 Actual - Variance of 53
1982 = 177 Projected vs. 93 Actual - Variance of 84

Singleton's data is not very good to this point. He's obviously been unlucky lately. But when you're dealing with a small sample size (and even 100 points of data is the minimum) you've got to take what you can get. As you can see, his distributions of 2B's and K's have been very skewed to this point but if you look at the closest variance from actual that you've got, it points to the 81 season.

Throw in the fact that you already know that, based on injury, it's either the 77 or the 81 season and that his platoon numbers really do point towards his 81 season, it's pretty much money in the bank for me to say he's in the 81 season.

Here are the reasons I use 2B's and K's above all other categories: There is no such thing as a "ballpark 2B" (only ballpark singles and HR's) and also K's work surprisingly well for hitters. (BB's you have to use PA instead of AB's and also your league's intentional walks tend to give you false information.) I'd use HR's as my third item to look at. BA/platoon numbers is usually the fourth item I look at but again that is more subject to slumps and the such. Not totally reliable by itself because of the stadiums and other league conditions.

2B's and K's are king, believe me. Try and crunch the numbers on some of your other hitters with data points over 100 AB's now. Write down your variances for 2B's and K's for them and make a prediction. Do this again around game 81 and look at what you predicted. Then at the end of the season, test yourself to see how close you were.

[yak1407]

Tried the 2B and Ks on Larkin and Parker.
The numbers on Larkin, after 331 ABs were very close for his worst season, which I had already guessed, no difference on Ks and 3 on 2Bs.
Parker, after 88 ABs, was harder to discern, 20 for 2Bs and 17 for Ks, leaving me with his 86 card for Ks and his '79 season for doubles, of course he only had one 2B as a sample.
But would platooning a player, who doesn't normally platoon work?
Not that it matters, it's a pretty straightforward approach and one I'm going to try going forward.
Thanks