Anyone else get velocity as a negative (non significant) predictor of ERA when they try to run a regression controlling for all attributes and no interaction terms? My method was quick and dirty so lots of holes could be poked in it but it also might reveal that in the sim logic swings and misses aren’t worth the extra toll on stamina or something of that nature.

I got the same in my (similar) regression.

There's a lot of evidence that the sim decides hit vs. out very early; if so, velocity might only serve to convert other outs to strikeouts. There are good data showing that substituting K's for other outs has little impact on scoring, probably due to DP avoidance.

I don't honestly understand the math of what you guys are doing. I come at things the other way, I look at the players themselves.

Generally speaking, when you look at the way players are generated by the system, a great Overall pitcher has all the "critical" ratings you're looking for along with great Velocity. It goes along for the ride. But the system, working on what I assume to be some kind of a bell curve, can't generate ALL great pitchers. And my observation is that when you see a just-very-good pitcher with high Control, he's usually got low Velocity. And I also see a lot of pitchers with great Velocity and poor vsL/vsR.

So while I never say never, when I evaluate pitchers, honestly Velocity is the very last rating I look at; sometimes not at all.

Same here. I just cancel the rating entirely and don’t use it to predict.

I actually think it is negative in guys who have great GB. You want to generate those double play balls, NOT strike guys out.

But I’ve never gotten a significant p value.

Correction on my part; I find velocity as a positive predictor of ERA. I just thought that it wasn’t intuitive and wondered if anybody else has found the same thing with different data pulls. Would be interesting to play with interaction terms.

cubcub113 - I agree with your logic, and so does a model with the VEL*GB interaction, which has a tiny positive coefficient (p = 0.36) and the VEL main effect becomes negative (p = 0.71). An interaction term might not perfectly capture your statement "you want to generate those double play balls, NOT strike guys out," but it agrees with it.

I also continually get P3 as a positive predictor of ERA, and sometimes P5. It would probably be better to break those down into buckets, but I don't think I pulled enough data to do that.

I create a single pitch rating in regressions. I highly recommend you do this.

(7*(P1-40)+6*(P2-40)+5*(P3-40)+4*(P4-40)+3*(P5-40))/5

A P-40 term that is negative (if a pitch is less than 40) I count as 0


I have never used interactions but I feel I should try this

Cubcub, the only problem with that formula is, if a pitcher has a P4 and P5 of exactly zero, that is not a bad thing. Especially if he is a reliever. The only negative is if a pitch rating is below 50 and ABOVE 0.

Posted by bjschumacher on 12/29/2019 10:42:00 AM (view original):
Cubcub, the only problem with that formula is, if a pitcher has a P4 and P5 of exactly zero, that is not a bad thing. Especially if he is a reliever. The only negative is if a pitch rating is below 50 and ABOVE 0.

You're saying a pitch of 40 is worse than 0? Where is the break-even point? You say 50 right?

Yes, that's what I am saying. If it's 0, he will not throw that pitch. For a reliever, would you rather have his pitch ratings be 90-75-0-0-0 or 90-75-50-45-40?

You would want the first?

Yes. I think anything under 50 is a negative.

What about stadium effects? If you are playing half your games in Durham, Sante Fe, Albuquerque or Colorado wouldn't you want to have more strike outs as opposed to balls put in play? Or are you guys saying that the sim determines out vs. hit before it determines what type of out it is? If that is the case, what is the point of the strikeout at all?

What do you think the sim determines first?
1. Out vs. hit
2. Out vs. ball put in play
3. Something else, please explain

Also, if P3-P5 have more significance than most people think, i would think that those DITR pitchers that have four pitches above a rating of 60 would be pretty valuable.