THE BOOK--Playing The Percentages In Baseball

Although I think BA is fine as it makes the point, but OPS or wOBA (or something else that captures offensive production) would have been nice.  I think he uses BA just because it is easier for some of the readers to relate to and in this case we don’t really care about complete offensive production.

The point which is great and should be obvious is that there cannot be a gap because if there were, that gap would get filled up by the minor leaguers, so to speak, and there then would be no gap.

Of course another issue is that there are so few LH batters in the first place (15% of the population?), that the fact that there are so many LH hitters in baseball (30%?) tells you right away that they must have an advantage.

It is worth noting that given that most (70-80% I think) GB are hit to the pull field (and that most IF hits are to the SS and 3B), the fact that the IF hit % are almost the same means that “given where the ball is hit” that 2 step advantage means a lot.

an interesting article; I’ll focus on my point of doubt…

John’s paragraphs 5 & 6:

“...certain defensive positions are only open to right-handed throwers: catcher, second base, third base and shortstop. Three of these are premium defensive positions and we can expect reduced offensive production from them, since teams are willing to sacrifice offense for a strong glove at these key positions.

But since these below-average hitters tend to swing right-handed as well as throw right-handed, they will bring down the level of the average right-handed hitter. Does that make sense? The end result would lead to a higher batting average for left-handed hitters.”

That’s reasonable speculation, and consistent with a higher aggregate batting average by left handers irrespective of position. But what should we expect when bringing position bias into the question? Good hitting left handed throwers/hitters concentrated in 4 positions, while good hitting right handed throwers (most also batting right) can be spread among 8 positions. Those below average right handed hitters should be concentrated at C/2B/3B/SS, bringing down the right handed average in particular at those 4 positions. I’d expect these concentrations of good left handed hitters and weak right handed hitters to widen the gap between LHB and RHB at each position, not narrow the gap. I don’t understand John’s claim that the lack of a real gap between left and right handers inside each position confirms the speculation. It sure seems to me like the opposite of what we should expect.

Now maybe the difference does emerges if we measure offensive skill beyond batting average. I would hope so.

The real problem is that you cannot address the question well with such high level aggregated data.

If I’m guessing correctly, it seems like John used AVG because all the balls he’s considering are in play and will either go as ground outs or infield singles (or ROE, although it wasn’t clear if he was looking at those or not.  I suppose an infield double could happen as well...) So for this specific subset of plate appearances, AVG = OBP = SLG.

Joe,

I’m not sure I follow your reasoning.  You wrote:

Those below average right handed hitters should be
concentrated at C/2B/3B/SS, bringing down the right handed average in particular at those 4 positions.

Is this true because the higher premium for defensive excellence at these positions lowers the threshold for offensive production?  If so, why wouldn’t the same hold true for the left-handed hitters? In which, case I wouldn’t expect a gap between RHB and LHB once you take position into account.

MGL,

It is worth noting that given that most (70-80% I think) GB are hit to the pull field (and that most IF hits are to the SS and 3B), the fact that the IF hit % are almost the same means that “given where the ball is hit” that 2 step advantage means a lot.

It does mean a lot, but with a twist. It turns out that batters who hit balls to the opposite field do better in getting IF hits. In other words, on balls fielded by the 2B, RHB actually get a higher fraction of IF hits.  Here are the numbers:

+------------+------+---------------+---------+-----------------+
| fielded_by | bats | IF_GB_FIELDED | IF_HITS | if_hit_fraction |
+------------+------+---------------+---------+-----------------+
|          1 | L    |          7539 |     610 |          0.0809 |
|          1 | R    |         12807 |     896 |          0.0700 |
             ----------------------------------
|          2 | L    |           842 |     134 |          0.1591 |
|          2 | R    |          1469 |     118 |          0.0803 |
             ----------------------------------
|          3 | L    |         15696 |     605 |          0.0385 |
|          3 | R    |          6089 |     305 |          0.0501 |
             ----------------------------------
|          4 | L    |         23311 |    1368 |          0.0587 |
|          4 | R    |         18126 |    1352 |          0.0746 |
             ----------------------------------
|          5 | L    |          5090 |     826 |          0.1623 |
|          5 | R    |         34677 |    2947 |          0.0850 |
             ----------------------------------
|          6 | L    |         10710 |    1136 |          0.1061 |
|          6 | R    |         38801 |    3172 |          0.0818 |
+------------+------+---------------+---------+-----------------+

John,

Very interesting table in #5. Could that be positioning...that is, the first baseman and second baseman both play closer to second with a righthanded batter...so they tend to have slightly longer throws?

I’m not sure that could cause all the difference though. And having to hold runners probably limits the variance in first baseman positioning also.

Does this include sac bunt attempts?

#6

bunts are included.

You can control for the 1B positioning bsaed on whether there is a runner on 1B.

Probably a better proxy for 1B positioning is whether there was a throw over to first, which is in the RS data.  A runner on first may be so slow that it’s pointless to hold him on.  Or the teams might be playing out a 17-2 slaughter-fest and be in “let’s just get this over mode.”

Pitchers don’t throw over to first as often as they used to (I don’t think), plus everyone gets held onto first base (Ortiz, Manny, Molina, Giambi, Griffey, etc.) except under rare circumstances, so throw overs would NOT be a good proxy for holding runners on.

If there is a runner on 1st, 95% (maybe more) of the time he is being held on.  If you want, you can remove 9th inning big (2 or more) leads and MAYBE pitchers on first.

The reason why opp field ground balls are more likely to be IF hits is two-fold.  One, fast hitters tend to hit more to the opp side (even GB) either by design or because they are slap hitters in general.  Two, opp field GB tend to be slower, such as “cue balls.”

Well, you can definitely remove the bottoms of the 9th and later innings, if there is another runner on base, other than 1B, that represents the winning run.  But, can’t be too many of those!

back to the positional bias itself…

Using John’s criteria (pitchers,DHs,switch hitters and pinch hitters excluded), totals for 2000-2006 are as follows:
Bat: L Th: L .277/.354/.454 (142,627 AB)
Bat: L Th: R .273/.349/.442 (166,736 AB)
Bat: R Th: R .270/.333/.434 (562,911 AB)
Bat: R Th: L .243/.325/.411 (4,239 AB)

If you stick to batting average, LHB are +8 pts in CF, +7 at 2B and +7 C. RHB are +12 pts at SS. RHB are 1 point better at each of the other 4 positions.

Nearly half of all left handed batters are also left handed throwers; effectively,they can only compete for playing time at 1B or OF, whereas right handed throwers can play any position. This raises the level of competition at 1B and the OF and lowers it at C/2b/3b/ss. It’s not just “defensive premium” which separates the two groups.

John suggests that an equilibrium occurs: “if there were a discernible gap between left- and right-handed batters in the major leagues, that gap would be filled by minor league lefties.” I don’t doubt that, but that equilibrium affects the floor of talent, players who get few at bats anyway. That wouldn’t explain the “disappearance” of the left handed advantage within each position, only its slight erosion. I think relative equilibrium within several positions has more to do with regular and semi-regular players moving along the defensive spectrum.

Leaving switch hitters aside, right handed batters get 92% of the PA at ss and 87% of the PA at c. Meanwhile LHB get 61% of the PA at 1b. Moreover the lefties are concentrated in higher offense positions which tend to bat more often at the top or the middle of the order. The 4 right-handed-thrower only positions get only 45% of the PA.

The advantage for left handed batters disappeared in the unweighted comparisons John presented. Weight them and the advantage would reappear.

Dan Fox unearths some more of his archives:

http://danagonistes.blogspot.com/2007/11/irreducible-essence-of-platoon-splits.html

Good job to Dan, to give wider exposure to his work.

I think I agree with Joe A. here.  The equilbrium idea would mean that the worst LHHs and worst RHHs are comparable hitters.  And I imagine that’s generally true.  And clearly MLB goes much deeper into the LH talent pool to reach that equibrium (only about 10% of population is LH).  But that doesn’t mean the overall means have to be identical.  The very best LHHs should be better than the very best RHHs, and so on down the line until they converge at replacement level. 

This difference disappears when you control for position because in the 1B/OF pool you are (virtually) comparing the top 40% of RHHs to the entire LHH population (79%), and in the C/IF pool you’re comparing the bottom quintile of LHHs to the lower 60% of RHHs.