THE BOOK--Playing The Percentages In Baseball

Two points

1) The move to IBB the batter, while strange, I think was also done as much to force the Marlins to burn their last position player, but also to get them to kill another reliever.  I’m not sure if Gardner was their last reliever, but he was their 7th reliever of the night; it was close.

If the decision to walk or not walk is essentially a coin flip, there’s extra value in forcing the other manager into a tough situation.  (Good luck quantifying that one!)

2) The Cordero thing was tough to watch, but the way the game flowed, it was more unluckiness than anything.  Cordero relieved Rauch in the 8th after he started hemorrhaging hits.  The pitcher’s spot was due up 5th, and the Nats don’t really have the roster flexibility to bring in someone who isn’t a butcher.  They rolled the dice, thinking the spot wouldn’t come up and that Cordero could hold it.  They were effectively wrong on both counts, but I’m not really sure if it’s a strategic error, per se.  Sometimes the actual game situation forces your hand in ways that aren’t otherwise optimal.

Bottom of the last inning, score tied, runners on 1B, 2b, 2 outs, according to The Book on page 306: Andy says to never issue the walk.

Interesting that MGL says to do so if the pitcher is coming up.

I’ll chime in later today with my own numbers.

Kinda sorta related: I’ve been debating the ninth inning of the Yankees/Red sox game last night with some friends. Down by one with two outs and Abreu on first. Okajima (lefty) is pitching. Kevin Thompson (righty) is batting. Wil Nieves (righty, I think; can’t hit regardless) is on deck. Posada and Giambi are out of the game; only Josh Phelps (righty) is available to pinch hit.

So Torre lets Thompson hit. Phelps is on deck to hit for Nieves. Thompson ends up striking out. My friends insist Phelps should have hit for Thompson. I insist Torre was right.

To me, Thompson is more likely to walk or double or triple. Their expected OBPs are probably about the same. Phelps obviously is much more like to put the Yankees ahead with a home run. But with Abreu’s speed and two outs, Thompson might be more likely to tie the game with a double.

My thinking is that with a runner only on first, you’re most likely going to need two batters to tie it anyway. (Phelps would probably go for extra bases 10% of the time, based on his career stats). Considering their OBPs and gap hit rates, it’s better to go with Thompson-then-Phelps instead of Phelps-then-Nieves.

Any thoughts?

Bottom of the last inning, score tied, runners on 1B, 2b, 2 outs:
http://www.insidethebook.com/li.shtml#18

LI = 4.4 ... implies a runs per win (RPW) of 11/4.4=2.5

win prob = .617

Bases Loaded, LI= 6.4 ... implies RPW of 1.7.

win prob = .665

A walk to an average hitter will add +.048 wins with an averag hitter coming up.  So, in order for the IBB to be break-even, the batter coming up must be -.048 wins per PA.  The RPW is 1.7, mean the runs per PA is -.048 x 1.7 = -.082 runs per PA.  That means a wOBA of around -.094 relative to average.  So, if the average is .340, then the breakeven is .246, more or less.

I didn’t know that Treanor was a pinch hitter.  Yes, it could be that forcing the last pinch hitter and another reliever is optimal.

Also in the game, Fick bunted in the later innings with a runner on first.  A very good rule of thumb for a manager (as stated in The Book) is to rarely if ever bunt with a guy with little speed and who is not a good bunter.  I don’t know if Fick is a good bunter or not, but I doubt it.  I don’t think he is fast, and in any case, he hit into a DP after jogging to first base.

Anthony, I don’t know for sure (at all), but I think you are right.  The “kicker” in these decisions is usually that once you inlcude the pinch hit penalty (see The Book - essentially a pinch hitter hits substantially less than he would if he had been in the whole game), pinch hitting is NOT the correct strategy.

Tom, that sounds about right.  I came up with an OBP of around .270 for the break even point.  Ramirez is an above average hitter (we mostly care about BA with him at bat, since almost any hit wins the game), so that would raise your BE point from .246 to something higher.

"the pinch hit penalty (see The Book - essentially a pinch hitter hits substantially less than he would if he had been in the whole game)”

Is there a pinch-hit penalty if the pinch-hitter stays in the game and bats a second time?

Not that this applies to Thompson in this case since he entered as a pinch runner.  Of course the first pinch-hit penalty might apply.

This game (Marlins/Nationals) had all sorts of situations that had me searching through The Book afterwards.  I suppose that’s bound to happen as a game extends, allowing for more situations and making one demand optimal strategy. 

Between Olivo’s two failed bunt attempts following Boone’s leadoff double (especially upsetting considering how bad a 2 strike hitter Olivo is) and Borchard’s dreaded streakout with a man on third and one out, this was not a fun set of extra innings for Marlins fans either.

Streakout?  Not even sure what caused that one.  Strikeout, obviously.

I don’t know if a pinch-hit penalty diminishes as the player stays in the game.  I would assume so.

Those sac bunts are tricky as far as determining whether a particular instance was correct or not.  If you read The Book, you will see and understand that in any “close” bunt situation (where the bunt may or may not be correct) it is usually incumbent upon the offense to sometimes bunt and sometimes not, to keep the defense honest.  As well, since the hitting team gets to see how the defense is playing, changing course in mid-stream is also often the correct thing to do.

Even in late inning close game play, like a tied game in extra innings, a bunt by a bad bunter and/or slow runner is probably rarely correct as there always needs to be a requisite number of singles and errors and a maximum number of failures in order for the bunt to be correct.

Also, the whole idea of a bunt being “automatic” in certain situations is ludicrous, even with a good bunter, poor batter, as the defense can take too much advantage of that knowledge.

Other than pitcher bunts (which managers get wrong all the time), the two biggest mistakes that managers make all the time with respect to sac bunts is bunting too much with slow or bad bunters and bunting too much when it appears to be “automatic.” My guess is that even under the most ideal bunting situations, the correct bunt % is somewhere around 50%.

Managers have little idea though of the importance of game theory and randomly mixing up whether you bunt or not, although I suppose if you explained it to them, most could understand what you meant.

The problem with managers, GM’s and most people in general, is not that they are not smart enough to understand expert advice - it is that there is this strong inertia associated with human nature to resist anything “different” and to avoid admitting that one is wrong or doesn’t know as much as they think they know.

It’s funny, as I actually remember reading earlier this year that Acta is something of a stathead.

Let me see if I can find that story…

(click on name)

Basically it sounds like he understands things like sample size and how precious outs are.  He gets the sac bunt wrong, but does seem to show a genuine curiosity and willigness to listen to sabermetricians.

You say you don’t know anything about Acta.  I seem to remember reading somewhere that he read your book (The Book) along with Baseball Between the Numbers and some other sabermertic-inspired books over the offseason.

So one would assume he’s not totally clueless.  But who knows?

That was Ned Yost.

Weird, I swear I had a reply in this thread earlier that somehow got deleted.

James Cole is right that Acta has made statements in the past have me convinced he’s at least open to sabermetrics.  He talked about how his team wouldn’t make outs on the basepaths just to please some “moron” fans who want him to be aggressive.  And he said at one point that he had been convinced by those who showed him the numbers that a player is more likely to score from first with no outs than a runner at second with one out.  Which is wrong, of course, but he seems to be at least open to new ideas.

Let me see if I can find that article again…

Here it is:

http://federalbaseball.com/story/2007/2/21/235947/639

Well, after Robinson, who in my opinion, was one of the worst managers I have ever seen, given the relatively small samples of Nats/Spos games I saw the last couple of years.

One of my next projects is to take The Book, add and subtract some things, and write a “manual for managers,” without all the math and explantions.  Seriously.  Then I might sell it to the highest bidder, since it does no one any good if every manager uses it.  Or at least try and convince a team that it is worth well over a million bucks, which it will easily be, considering the value of a marginal win.

BTW, what a joy it is to type in a real word and number rather than some ridiculous series of letters that a human being can’t even read without a magnifyng glass and some “spy paper” like the red celophane that used to come in the home version of password…

Most teams didn’t want to spend 18.95$.  I doubt you’ll find one to spend 1,000,000$.

Just to clarify, Manny Acta never called anyone a moron or threatened to slit his wrists or whatever it was I wrote in that post. Somebody else had transcribed Acta’s statements during a press conference, and rather than repeat that blogger’s work, I didn’t quote Acta directly. Those statements are just my gloss on what Acta said.

I don’t know about anyone else, Basil, but I figured that…

Figured as such, MGL. But this was the second or third time I’ve seen a link to that post at least seeming to indicate I was quoting Acta, so I thought it best to clarify that I wasn’t.

Regarding batting average with the bases loaded.  It is indeed higher than the overall average, but that is entirely driven by the sac fly issue.  If you count sac flies as ABs, there is no increase in batting average with the bases loaded.

Here’re the numbers (2000-2006):
BA: .2653
BA (w/ bases loaded): .2846

Counting sac flies as ABs:
BA: .2635
BA (w/ bases loaded): .2614

LaRussa made a similar move last night: tied game in top of 9th, 2 out, men on 1B and 2B, issued an IBB to Atkins in order to face Hawpe. Hawpe then walked to force in go-ahead run.

Can the difference between Atkins (hitter platoon advantage) vs. Hawpe (pitcher platoon advantage) be big enough to make this the right move?

Doubt it, but maybe.  OK, I ran it throught the sim.  With Atkins hitting versus Flores (a good lefty reliever), STL wins .5087 in 100,000 games.  With Haupt, COL wins .5230.  So the IBB cost around .03 wins, not too terrible I guess.  I don’t know whether that is small enough to say “flip a coin” or “do whatever you want.” Plus, if it is close, maybe there are things that LaRussa knows that might close the gap or even put it in his favor.  I put these in the class of things that a manager would have no idea whether it is correct or not, so I give them a “pass.” When a manager does something that is obviously wrong even to a “civilian” using a little common sense and maybe some high school math, then I consider him a moron, like Manuel letting Eaton hit in the 7th inning last night in a tied game with 2 outs and runners on 1st and 3rd or having a batter who obviously can’t bunt real well, attempt a sac bunt.

The LI situation of Guy/20 is 4.5.  This means the runs per win converter was 11/4.5=2.44

MGL is reporting an extra .0143 wins to issue the IBB and face Hawpe.  (MGL, wby did you say .03?).

So, .0143 wins means it was .0143x2.44=.03 RUNS.
If .03 wins was the correct number, then that’d be .07 RUNS.

My general rule of thumb is I’ll grant the manager a “he knows something I don’t” margin of knowledge of .05 RUNS per PA.  That is rather huge (difference between the worst hitter on the team, and an average one.  I sometimes will even grant him a .10 RUNS per PA margin of knowledge (difference between one of the best hitters in the league and an average one.

Tango:  the diff is .5087 vs. .4770 for StL, hence the .03 win difference.  Seems like a pretty big difference to me.

MGL:  Just curious—does your sim account for the various ways the runner on 3B might score w/o a BB or H (balk, PB, WP, ROE)?  And how much of a wOBA/OBP boost does Hawpe get from batting with bases loaded?

Ok, so that’s .0317 wins, which is .077 RUNS.

As I said, a .050 RUNS margin of knowledge is my rule of thumb.  Anything over .100 RUNS is indefensible.  Some would even argue only a .020 RUNS margin is enough (12 runs per 600 PA).

I guess it would be interesting to study the issue for one week, and declare “These were the 10 most ridiculous moves of the week”, to see how much margin there really is.