Projecting WHIP

[WebHamster]

This thread contains comments on Bob Hoyng's Article "Projecting WHIP", which was posted in the Cafe's Point of View section on June 26, 2006. The full article can be read here.

[Madison]

Guess I'll have something to test out next offseason. Very nice writeup!

[ehk103]

Smoking egg-rolls can be harmful to your health, bob.

[mamorris]

Bob,

Nice article. A possible reason for having to reduce that multiplier from 3/7 to .4 might have something to do with outs created on the basepaths. I have seen some sabermetric formulae that multiply innings by 2.82 instead of 3 to get a more accurate estimate of batters faced (or balls in play in your case). This would make your new formula:

Projected Hits = (BABIP / [1 - BABIP]) * (2.82 * IP - K) + HR

It might be worth a look, even if it doesn't change the results much.

[The Loveable Losers]

Interesting thought on that. I'll look into those numbers at some point to see if they come out a bit closer while sticking to the .300 BABIP.

I also saw a recent article at The Hardball Times that talked about differing theories to Voros McCracken's (that BABIP is not controlled by the pitcher). They make enough of an argument to convince me that you would be well advised to take into account pitchers like Webb (who should have a slightly higher than average BABIP due to the large number of ground balls) and Eric Milton (best I could come up with off the top of my head that should have a slightly lower than average BABIP due to the large number of fly balls).

They found a slight correlation in the article between pitchers that gave up a high BABIP giving up a high BABIP the following year but did not find the same correlation for pitchers that have a low BABIP following that up with another low BABIP year. The correlations they found in the article were not overly convincing but even so I would definitely lend a bit of credence to high BABIP pitchers and therefore regress their numbers toward the mean some but not automatically bring them all the way back to .286 or .300 or whatever ends up being used there.

What it definitely looks like - although there just hasn't been enough information to state this convincingly - is that holding hitters to a BABIP of around .300 is a skill that pitchers do control. However, it's a skill that has a certain ceiling as well and almost all major league pitchers achieve that ceiling. There are outliers though - guys that just don't have the stuff to keep major league hitters from sitting on one pitch in one location - that simply can not hold hitters to that ceiling that almost all major league pitchers are capable of attaining. Those outliers are the guys that have the higher BABIP against and while they normally won't last in the major leagues and would rarely if ever find their way to a fantasy roster they do have to be taken into account when you're examining the data.

[The Loveable Losers]

Interesting results.

First of all, here's the two charts:

1941 total players (all pitchers from 2003-2005)
Within X%: 5%, 10%, 15%, 20%
Old Formula: 673, 1140, 1411, 1624
New Formula: 684, 1144, 1424, 1614

479 pitchers with 90 or more IP from 2003-2005
Within X%: 2%, 5%, 10%
Old Formula: 479, 90, 195, 327
New Formula: 479, 94, 214, 332

What we see here is that the two formula return very similar results but the new one seems to be a tiny bit better. More importantly though we've switched to using the league average .300 BABIP without causing any degradation to the quality of the formula. I'd say that while further research (especially by real statisticians) is in order I highly prefer the new formula. I didn't like making that adjustment from using .300 BABIP as a general rule of thumb to using .286...it definitely didn't feel right.

So the general format for the new formula would be as follows...
Projected Hits in Play:
pHIP = (BABIP / (1 - BABIP)) * (2.82 * IP - K)
pWhip = (pHIP + BB + HR) / IP

And the general formula for PHIP would be as follows:
pHIP = (3 / 7) * (2.82 * IP - K) or if you want the straight decimal values pHIP = 1.20857 * IP - 0.42857 * K

From a technical aspect the new formula gives a bit more weight to strikeouts or to but it more accurately it gives the correct weight to strikeouts. We were using too low of a BABIP number to compensate for using too high of a number for the IP. While this worked for the most part it would have skewed our results as we shifted the BABIP since IP were 3 times as important as a strikeout in the original formula when the correct weight is a 2.82-1 ratio.

[Mugrila]

Is it bad that all of this goes right over my head?

[The Loveable Losers]

Is it bad that all of this goes right over my head?

Nope...that's why I like to boil things down to formulas...almost anyone can use those.

If you take the number of innings a pitcher throws times by 1.20857 and then subtract 0.42857 * K and finally add in the home runs you're going to get pretty darned close to figuring out that pitcher's hits given up. Furthermore if a pitcher varies greatly from that (especially doing better than the projection) you can reasonably assume that he's going to fall off some and regress back toward your projected hits. Understanding the logic behind it is nice but as long as you can use a calculator you can use the result to figure out which pitchers are overperforming and which ones are overperforming when it comes to their hits given up and by extension their whip.