THE BOOK--Playing The Percentages In Baseball

I should point out that we need to figure Bonds’ and Mayne’s win linear weights based on the impact of each event with the bases loaded, and 1 or 2 run deficit.  This may give us more margin for error to tip things towards Showalter’s decision.

For example, if you construct the win-specific LWTS or wOBA, you will get this for the Bonds’ situations:
wOBA = .5 x BB + 1.0 x 1B + 1.5 x 2B + 1.8 x 3B + 1.8 x HR
all divided by PA

With the Mayne situation, it would look more like:
wOBA = .8 x BB + 1.1 x 1B + 1.2 x 2B + 1.2 x 3B + 1.2 x HR
all divided by PA

The random wOBA has coefficients of: .72, .90, 1.24, 1.56, 1.95, respectively.

As you can see, Bonds’ biggest weapon (walk) is reduced in significance with the 2-run bases loaded situation.

In fact, every situation needs to be evaluated in terms of Linear Weights (or wOBA) by the particular game state.

This is an interesting topic that has been addressed in several forums before.  I have ran the scenario on my sim and while I don’t remember the exact outcome, it was a clear win for not walking Bonds.

The fact that Bonds’ lwts or wOBA (or OPS) is heavily laden with the non-IBB walk further bolsters the decision not to walk him, as Tango states.

I am not sure why we should give Showalter the benefit of the doubt.  In fact, I see no reason at all.

He (Showalter) has no idea what the win expectancies were with respect to either player hitting.  He simply was afraid that Bonds was going to “beat his team.” And of course, he had a 70% chance of looking like a genius.  All he knows is that Bonds was a great hitter, and Mayne was pretty bad.

Granted the decision was perhaps close enough that “something” unknown to us could have swung it in the other direction, but to speculate or worse yet presume that Showalter knew something that we didn’t, is silly, IMO.  We can make that “disclaimer” for just about every dumb decision a manager makes.

Managers make “mistakes” like this all the time.  While the costs of each mistake might be small (in this case, only .03 wins, although I think it is more if you run a sim), at the end of the season, I believe that they can add up to several wins.

I truly (and firmly) believe, and I would probably get lambasted in another forum for such an opinion, that if a manager did nothing but teach and motivate his players and left all strategic decisions and lineups up to a computer, his team would win 4-5 more games per season.

I’m quite certain Showalter had no idea what the win expectancy was, and he did as you said, that he didn’t want Bonds there.  And the unknown could have actually swung in the opposite direction of what I’m giving Showalter credit for. 

I think in this particular situation, a situation that pretty much has never existed before (the highest LI possible with the best hitter of his generation at bat followed by an average hitter), I’d give in to the “close enough”.

***

I agree that a purely “model-driven” process would net a team 4 or 5 wins.  Perhaps it would gain 6 or 7 wins on the pure “numbers” of it, and maybe lose a couple of games for lack of “human” understanding.

I agree that a purely “model-driven” process would net a team 4 or 5 wins.  Perhaps it would gain 6 or 7 wins on the pure “numbers” of it, and maybe lose a couple of games for lack of “human” understanding.

I like that!  Of course, a good model can utilize “human understanding” as well.  Or at least the practical use of the models can include some discretion by the manager, such as when your “IBB Bonds” chart says “go with gut,” although, as you say, there is no particular reason to assume that a manager’s gut adds any positive win expectancy to a pure “numbers” model.

I think what many people always forget is that the optimizer is a two-step process.  One step is to estimate the data inputs.  Garbage-in-garbage-out won’t help any optimizer, no matter how well constructed.  So, I would ask what the manager thinks is the player expectation in this particular situation.  The other part is the model.  The model can’t be as simple as just assuming everything is independent of everything else, or it can try to get into “human understanding”, like how the hole opens up between 1B and 2B, which a lefty can take advantage of.  Or that a particular hitter is a fastball hitter, and the pitcher is a fastball-only pitcher.

Once all the data points are known or estimated based on real data or a manager’s experience, it then becomes a simulation exercise.

If people don’t like the answer, then this must mean that they didn’t like the data that went in, or that the model itself wasn’t programmed to be complex enough.  The computer is only as smart as the guys who work on it.

I analyzed this question for many more situations using my Markov model and presented my findings at the annual SABR meeting three years ago. The presentation is on my web site:
http://www.pankin.com/sabr33.pdf
(or it can be reached from http://www.pankin.com, Fun Stuff, Baseball, When Should Bonds be Walked Intentionally if the above does not work)

Depending on the score, inning, and the following hitters, walking Bonds with the bases loaded can increase the chance of winning the game. I did not analyze the 1998 game, but my recollection was that he was followed by a weak hitter then, so walking him may well have been the “percentage” play.

As mentioned, it was Mayne, an essentially average hitter.

I also recommend to readers to check out Mark’s site:
http://www.pankin.com/baseball.htm

Especially interesting is the discussion on Markov:
http://www.pankin.com/markov/

Tom

I recall Tangotiger visiting this topic in the past and it was nice to see an update on this topic incorporating your latest research findings.  Thanks also for the link to Mark Pankin’s Markov webpage, I was just thinking I need to learn more about that and here it is, serendipitously.

I was wondering if it would be too much for you to revisit another old research you did previously on whether Felipe Alou hates the walk or not and “forces” his players to avoid the walk.  I recall your research showing that it was more that Felipe was supplied players with lousy OBP than him eschewing the walk - it came out about the time he was hired by the Giants.  Would it be possible to update that given his experiences as a Giants manager?  Some Giants fans seem to think Alou is “still” that type of manager and I was wondering what the facts, if any, show.  Thanks either way, nice content on your new blog.

This is the Alou article that OGC is referring to:
http://www.tangotiger.net/alou.html

Seeing that I am highly fond of Alou (after Tim Raines, my favorite baseball personality), perhaps I should give it better treatment.  I’ll put this in my to-do-eventually pile.  Thanks…