THE BOOK--Playing The Percentages In Baseball

Is that new or something you found in the archives?

Interesting stuff.  Ted Williams (who didn’t suffer from strikeouts or lack of power) advocated the two strike approach to hitting:  Swing from the heels until you have 2 strikes, then change your approach a bit, choke up a little, and try to hit it back through the middle.

Ted knew what he was talking about.

Brand-new.  Just did it right now.

Neat study.  Any significant age difference between the Kleskoyans and Jedmonds?

Good point.  I meant to look at it, and then forgot about it.

If we look at the guys in the first of the four years as being 25 or younger as YOUNG, and 30 or older as OLD, we get about 30% in each group.

Among the 100 guys whose K rates dropped, 49 were young and 16 were old.

Among the 100 guys whose K rates increased, 26 were young and 38 were old.

If we focus only on the young guys:
- the 1430 guys in the control group had a wOBA go from .337 to .343 (K rate stayed flat), a 6 point increase
- the 49 young guys with a drop in K rate increased wOBA by 19 points
- the 26 young guys with an increase in K rate dropped wOBA by 1 point

That’s a 20 point swing.

If we focus on the old guys:
- the control group went from .353 to .345 (8 point drop)
- the old guys who lowered their K rate: +13 points
- the old guys who increased their K rate: -15 points

That’s a 28 point swing!

Among the 26-29 year olds in that first year, the guys who dropped their K rates: +9 wOBA, the guys who increased their K rates: -2 wOBA.  Control group: -2 points.  So, an 11 point swing.

Great study, Tango.  Wasn’t this Mike Schmidt’s point to Pat Burrell?

Absolutely enlightening. You have so many of these you need to set up some sort of easily accessible index for us to peruse ...

More evidence for the thought that walk and strikeout aren’t opposites.  Strikeouts and walks actually have a lot in common, and the opposite of those two is putting the ball into play.  Good work on this.

Tango, I would love to see what “year 5” looks like for both groups and to then compare that to a Marcel.  And to also break that up into young, old, and in-between groups.

For example, let’s say that Marcel predicted the decrease in K group should have a wOBA in year 5 of .300 and they actually had a wOBA of .310 - that would be evidence that players who decrease their K’s are a subset of players who defy Marcel.  IOW, they improved their true talent more than we typically see among all players.

IOW, we are trying to see whether an increase or decrease in wOBA accompanied by a substantial increase or decrease in K rate is more indicative of a change in true talent than a generic increase or decrease in wOBA.

Any chance to run some numbers on that?

Great study! A++ (I was just on E-bay.)

Yes, a nice little study on an interesting topic. Tango was surprised by one of the results, but that didn’t surprise me at all. Makes sense if you think about it.

I think Teddy Ballgame/Rally had it exactly right--don’t worry about the strikeout until you have 2 strikes. At that point, be willing to change your approach. I wonder, if “by count” stats are available for (some of) the Kleskos and Edmonds, if this 2 strike thing can be confirmed as the actual change that was made.

"I wonder....if this 2 strike thing can be confirmed as the actual change that was made.”

I’m gonna guess that virtually all of Edmonds’ Ks came on a 2-strike count......

---"I’m gonna guess that virtually all of Edmonds’ Ks came on a 2-strike count.....”

HaHa? Do you think you caught me saying something stupid? I shouldn’t have to point out that not all 2 strike counts result in Ks. We should be able to detect a change in approach with 2 strikes, with the appropriate data.

Of course EVERY batter’s swing (and pitcher’s approach) should be tailored to the count since the value of the various outcomes changes (a swing and a miss is not that bad until 2 strikes).  Any batter that does not do that is probably not operating optimally even for him.  Now how much to change one’s swing with the count can depend entirely on the batter himself.

Just a joke, David.  Guess I should have used a smiley face.  What’s your theory:  that the reduced K hitters had just as many 2-strike counts, but avoided the K more in those counts?  And would we expect something different from hitters who reduce their K rate? 

* *

Clearly, reducing Ks is better than increasing them.  The interesting question here is how much can hitters choose to make these changes?  I’d guess the increased-K group didn’t choose more Ks: their ability to avoid Ks declined, often because of age, and they adjusted by increasing HRs and BBs but not enough to offset the Ks.  The decreased-K group, however, may reflect a change in strategy that other hitters could emulate.  Is there any way to tell?

I have to say, I’m not so sure about this study. You’ve proven the correlation, but I’m not sure we see causation. I would assume that the Edmonds group has gotten worse because their bat speed has slowed. Meanwhile, the decrease in strikeouts and accompanying increase in other stats for the Klesko group could be due to any number of reasons: Park, hitting coach etc. That does not mean that if Adam Dunn tried to cut down on his strikeouts right now by himself, he would be successful.

(Note: I know there is no increase in the rate of any of these numbers, but simply an increase in balls in play. But I figure that the rate probably decreases when you try to put more balls in play, so they must have improved to keep up the same BABIP, etc.)

David/14: I don’t know that you could say it was a change in bat speed. In post #4, I broke it down by age, and the same overall pattern emerged.  Whether it was a conscious decision to change the bat speed, or a decision to change the hitting approach, or an influence from the hitting coach to make changes, I don’t know.  But, there was a huge shift in K rates, the one thing that the player has most control over.

Park would be the one interesting effect to look at, especially since some parks affect visibility more than others, and therefore, something for the hitter to deal with.

***

MGL: I’ll look at year 5/6 later today.

Tango, the same pattern appeared in terms of the wOBA result of increasing/decreasing Ks, but the increasing K group was MUCH older.  So it’s reasonable to think declining bat speed (or slower reactions) is at work for some/many of the older players.  For the younger hitters with increasing Ks, we may be seeing cases where ML pitchers discover a “hole” in their swing (this group is only 1.7% of young hitters in your sample).  Again, it doesn’t necessarily mean the hitters made a conscious change in strategy that resulted in more Ks (indeed, given the small gain in HRs and large drop in BA, that seems unlikely).

We don’t know if there was a conscious change in hitting approach by these guys, but since the low K group decreased their HR total, and the high K group increased theirs, it seems to me there is more going on than just overall improvement or decline.

DSG - You’re right, just because Dunn decides to cut down on his strikeouts does not mean he’ll be successful at it, but Dunn could be a very good test case - he has made nosie this spring about changing his approach to hitting and reducing his k’s.  At one point he was joking about being a 275 pound Ichiro.  We’ll see how this turns out.

"We don’t know if there was a conscious change in hitting approach by these guys, but since the low K group decreased their HR total, and the high K group increased theirs, it seems to me there is more going on than just overall improvement or decline.”

Maybe, but you have to think about the survivor bias in this study.  You’re looking at hitters who saw a huge spike in their K rate, but STILL got to play regularly in the majors.  We would expect this (very small) subset of players to be those who made successful adjustments to partially offset the Ks.  The other hitters whose K rates increased—likely a much larger group --lost PT or went to the minors and didn’t make the study.  (Of course, this means that the true impact on wOBA of a large increase in Ks is much larger than Tango’s data suggests). 

We also don’t know how many players reduced their K rate at some point, but paid such a high price in reduced HRs or BBs that they quickly abandoned the experiment, or actually were demoted or lost PT (and fell out of the study).  This is probably a smaller group, but we don’t really know.

It would be interesting to divide the reduced-K group by increase in wOBA, and see if those who most improved have distinctive qualities.  Were they high-BABIP hitters to start with?  Power hitters?  (Might first want to expand this to top 200 in reduced K-rate to give you decent sample.)

I said this:

Among the 100 guys whose K rates increased, 26 were young and 38 were old.

Are you suggesting that “the young guys had holes and the old guys were getting slower”, as the cause?  I’m not disagreeing necessarily.

***

Guy/18 is right about survivor bias.  After all, I’m imposing an 800 PA over a two-year span.  Hopefully this study can spur others to take a more careful look.  Indeed a study at the minor league level may be more fruitful.

MGL asked about what happens afterwards.  I’m only setup right now to keep looking at twoyear chunks of at least 800PA.

Of the 100 guys who dropped their K rates substantially, 71 survived the study for years5/6.  Their K rates progressed as follows for years 1/2, 3/4, 5/6 (with around 1100 PA in each group):
20.2%, 13.8%, 15.1% (regressed 20% back)

54 of the 100 guys who increased their K rates survived the study (a bit under 1100 PA):
13.9%, 19.7%, 17.9% (regressed 30% back)

The control group:
12.6%, 12.4%, 12.5%

If I only look at the young players:
36 of the 49 “lower K” survived
19 of the 26 “higher K” survived

Their rates:
19.2%, 12.8%, 13.7%
14.1%, 19.7%, 16.9%

The other 1158 young guys:
13.0%, 12.6%, 12.4%

So, the young guys who dropped their K rates regressed back 18% (after adjusting by the control group), while the young guys who upped their K rates regressed back 43%.  Who knows, maybe those punks were cocky, and decided to reign it in?  Here are those guys, if you want to study them:

K12 K34 K56 playerID
22% 28% 20% allendi01
21% 28% 24% cansejo01
19% 25% 23% vaughmo01
19% 25% 19% howarfr01
19% 26% 22% preslji01
15% 22% 16% evansdw01
15% 20% 19% washicl01
15% 20% 17% vanslan01
15% 21% 21% jonesan01
14% 19% 19% vanslan01
14% 19% 19% mantlmi01
13% 18% 14% oneilpa01
13% 18% 12% deshide01
12% 17% 14% blairpa01
11% 16% 12% venturo01
9% 14% 11% rodriiv01
9% 14% 13% henderi01
8% 14% 12% knoblch01
7% 12% 12% adcocjo01

Are you suggesting that “the young guys had holes and the old guys were getting slower”, as the cause?  I’m not disagreeing necessarily.

Yes, I think the old guys are mostly slowing down (and adapting to that).  On the young guys, I have no strong intuition, but could be ‘holes’ in some cases.  Does this group have much PT prior to year 1 of your study?

the young guys who upped their K rates regressed back 43%.  Who knows, maybe those punks were cocky, and decided to reign it in?

Or, selection bias.  Those who didn’t find a way to cut back on the Ks didn’t survive.

Of the 7 young guys who pumped up their K but didn’t survive the study in years5/6, here they are:
yr4 K12 K34 wOBA12 wOBA34 playerID
1974 12% 17% 0.331 0.335 colbena01
1995 20% 25% 0.354 0.329 coopesc01
1986 17% 22% 0.330 0.323 hendeda01
1982 12% 18% 0.332 0.317 jonesru01
1998 13% 18% 0.337 0.314 mabryjo01
1967 14% 20% 0.334 0.300 wardpe01

(Canseco actually is in the original study twice, as was Van Slyke.  Both Van Slyke survived, and one Canseco didn’t)

We can look at b-r.com for these 6 guys.  (Little know fact… use this playerID in the b-r.com search box.)

Nate Colbert: forced to retire due to injuries

Scott Cooper: missed the entire 1996 season, came back in 1997 for one last season

Dave Henderson: missed part of the 87 season, and came back still pretty good in his 30s

Rupert Jones: missed part of the season, still strong rest of his career

John Mabry: part-time player by age 28, and probably should have been for his whole career

Pete Ward: missed surviving because he had 725 PA in 1968/69 (I was looking for at least 800).

So, the guys who didn’t survive wasn’t because of the jump in their K rates or sudden drop in production.

I have a different causality question. Can you assume that cutting down on K’s caused the improved BA by, in effect, giving the hitter more chances to put the ball in play? Or did the improved BA reduce the chances that the player would strike out (i.e., he put the ball in play and therefore couldn’t strike out)?

All you can say, it seems to me, is that these are “competing risks” (just like actuaries calculate competing risks between causes of death)—different ways to end an AB—but you can’t impute causality with the type of analysis you’ve done.

You of course can’t cut down on K’s.  Ks are results.  What you can do is try to make contact with the ball more, shorten your swing, hit earlier in the count, etc.  Whatever it is that these batters did resulted in what I showed.

Bumping…