THE BOOK--Playing The Percentages In Baseball

I’ll say up front that I was not real happy with my analysis.  The details are fine, but it seems to me that you really have to do this thing from a WE perspective and not a RE one, even early in the game.  That is not easy. It is much easier to work with RE. We can use WE tables for average players, but to construct custom WE tables or to modify existing ones (for various batters, pitchers, and situations), is not that easy to do, to say the least.  I wish I had tried though.  I also wish that I had at least given the actual win rate of the teams that issued the IBB’s in that situation in addition to the actual RE, which I did give.  Nevertheless, it is a fairly interesting read, if nothing else for the details, as far as when, how, and why.

My first comment.

Bear with me, please…

As i understand it, the real question isn’t how many runs scored after the decision (1.722 runs), it´s the percentage of time that the move actually worked.

Let’s say you are Jim Leyland and your team has had trouble scoring runs. It’s the 6th inning and your frame of mind (Even if it is wrong), tells you that you need to stop them, the game is on the line.

You think you can overcome a 2 run deficit past the 6th inning, but not a 3-4-5-6 run deficit. So you go “all in” in order to avoid any runs.

Of course, if your move does not work out, you are going to give up a lot more runs then the initial 2 (2nd and 3rd base runners), but maybe that doesn’t matter. You already lost the game.

P.S. Love the blog…

Alright hold on one second.  You can’t spend a year ripping into people who run imperfect analyses in their articles, and then proactively tag yours with a “I’m not real happy with this” disclaimer.  The honesty and insight is nice, but it’s kind of a cop out.

Mike, go ahead and rip away!  What else would you like me to do?  I don’t know what you mean by “cop out.” That I should not have had it put on the web site, if I didn’t think it was that good?  It has some value, I just didn’t think it was a great analysis without the WE.  If I were writing that for a book which people had to pay money for, I would have written something different.  If THT doesn’t like something, they have the option to not print it of course. It is up to them.  If they think their readers will get some value from it, then that’s fine with me. Who am I to tell them what to print or not?

Honestly, I was going to revise it and post it here (a link to Google Docs) or just let it die a slow death.  I gave it to THT a while ago (a month or so) and didn’t think they were going to put it up on the web site after all this time, but it popped up unbeknownst to me (with a glaring typo in the first sentence to boot).

One thing that might be good to look at the distribution of runs allowed that inning after an IBB occurs vs when an IBB does not occur.  If you are in a tied game in late innings it probably doesn’t matter if you allow 1 run, 2 runs, or 10 runs.  The outcome is very bad.  I’d guess that the IBB would increase the times you got out of the inning without giving up a run.  This in effect “keeps you in the game”.

I agree, it would be very useful to have a WE that easily took into account all variables including team construction, splits, etc. but that is probably way too hard.

#5, yes that is essentially what WE does.  It is true that in general a bases loaded situation gets you out of an inning with no runs a little more than with runners at second and third.  So, late in a game when it is imperative not to let up any runs, it may be correct to issue the IBB.  My assumption is that early in the game, that is generally not the case, that you care mostly about average number of runs.  But, that assumption may be wrong even in the 6th inning of a close game.  That is what I am unhappy about with my analysis - that I did not look into that.

I think starting to look at it with RE is legitimate.  Looking at WE is of course the next step.  It doesn’t de-legitimatize the original analysis, but only mitigates, somewhat, any conclusion.

So are you going to write a followup, answering the last question of the article?

#6.  Yeah I guess I suggested that because I thought it’d be easier to do and would maybe shed a little light on the actual outcomes of IBB.  WE and RE are great and extremely useful but I think they might hide a little of the info that would be interesting.  WE convolutes the sucess/failure of an IBB with LI which makes it useful for identifying when IBB are useful but does not tell us how runs score or with what distribution.  RE reduces the number down to an average value.  This hides the information about the distribution of runs scored.  If 50% of the time no runs score and 50% of the time 4 runs score this might be an acceptable risk in a close game.  If 100% of the time 2 runs score then the IBB is always a bad decision in a close game.

The route you took with RE is actually more useful than you give it credit.  It is very difficult to take into account platoon splits and player quality with most techniques.  I think it was a good effort but is such a complicated problem that no easy solution is apparent.

Another idea would be to look at each instance of an IBB and try to judge the validity of it using your RE/wOBA methods using actual player information.  Then we might see some instances where it makes sense to have an IBB.

This is part I of a two-part article!  I address platoon splits and quality of hitters (or at least attempt to), etc., in Part II.

I agree that RE is for starters and then you want to go to WE if you can or if you have to.

And Edgar (Edgar who for pres. - Martinez?) WE automatically takes care of everything.  You can show the distribution of runs scored if you want to, but it doesn’t add anything to the analysis.  WE is WE, which is all you ever want or need.

MGL:  Without doing full WE, maybe you can compare how often a team scored 1 or more runs after the IBB, vs. the % of times it would have scored from the -XX/1 out situation.  Presumably, the IBB is an effort to prevent ANY scoring, in a situation in which giving up 1 run is very damaging (but additional runs aren’t that important).

Wrote this at ballhype, but I thought I’d copy it here as well…

Neat article, MGL, looking forward to the next part.

Minor comment: seems as though another factor to consider is the quality of the pitcher.  I’d guess that the average pitcher in your “runner on 2nd and 3rd, one out” situations is worse than the average pitcher overall, simply because good pitchers don’t get into that situation as often.  This might be another reason that you’re seeing the higher-than-expected wOBA after the intentional walk.  And it’s another reason to be careful about comparing actual rates to average rates in this situation.
-j

Well, given that I only looked at the 6th inning and earlier, trying to prevent any runs at all cannot be the primary goal.  In any case, given that the batter being pitched to after the walk is an above average batter and given that the DP chance is only slightly increased, my guess is that they score one run a little more than the regular tables would suggest in an average bases loaded situation.  Plus, we can look at a typical WE table and see that in the 6th inning or earlier, walking the bases loaded is NOT a good idea, with average batters, etc.  But, as I said, I did not look at WE in the situations where batters were actually walked, so who knows?

MGL - I really don’t understand why you decided to do the analysis this way.  Don’t you have a simulator that can plug in both teams lineups with opposing pitchers and bench players and play out the reat of the game, both with and without an intentional walk?  That is the only accurate method for evaluating individual strategic decisions, and I thought from other threads that you were set up to do that.  If you want a more generic answer you could just input more generic information into the simulator.

Yes, I can do that for one situation or for a generic situation.  But I can’t really do that for IBB’s in general.

Part 2:
http://www.hardballtimes.com/main/article/when-to-walk-part-2/

When I get a chance, I’ll post some of Andy’s findings that were shown in The Book.

Andy, p. 296, The Book:

Issuing an intentional walk to an outstanding hitter will almost always increase the opposing team’s run production. The only base/out situation in which a walk reduces the run expectancy is when there are men on second and third, two outs, and an unprotected elite (Pujols-like) hitter at the plate.

It is very rare for there to be an “unprotected elite hitter” at the plate in general.  The times when you have that and runners on second and third with two outs is probably a couple of times per year for all 30 teams.
What I wish I had done (and maybe I still will) with the IBB analysis is to see when it is correct to issue that IBB fairly late in a close game, simply based on WE.  Managers do that all the time in a close game in the late innings with runners on 2 and 3 and one out, to “set up the DP.”

While I have shown that that significantly increases the RE, pretty much no matter who the batter and the next guy are, I don’t think I have shown when it might be correct from a WE perspective.  My (educated) guess is that it is often close but that it rarely clearly correct to issue the IBB, even in late game situations (not the 9th or later of course - it is often correct then), such that managers are still better off just not doing it unless it is obvious.  The reason I say that is that the common belief is that you significantly reduce the batting team’s chances of scoring at all after the walk (even though you increase the total average runs).  But that is not true, as you can see from the run distribution tables.  You only slightly increase the chance of the other team scoring zero runs.

Anyway, Part II is up at THT if anyone is still interested.

On the other hand, a manager knows that he will get crucified by fans and in the press, and perhaps even in his own organization, if he doesn’t walk an opposing team’s elite batter, the game is on the line, and he ends up getting beaten (you know the old, and silly, axiom, “Never let the other team’s star hitter beat you”).

Usually, if he issues the expected walk, he is on safe territory (from criticism), whether his team ends up winning or losing. So maybe it’s not worth the grief that might ensue. His call, I guess. Now, if managers want to make protection of their derrieres part of their decision-making process, it’s fine by me. I only wish that they would do it using the right information in the first place, so as to make an informed choice.

The manager’s main job is not, as the rest of us might mistakenly assume, to win ballgames; it’s to avoid being blamed for failures. Usually, an underperforming team can’t fire the people who actually underperformed, the players, due to long-term salary commitments and other issues. GMs rarely fire themselves for putting together an inadequate roster. The manager, then, becomes the fall guy.

The best way to inoculate yourself from scapegoating is to avoid as much criticism as possible and avoid giving management something they can point to to justify firing you because they put together an inadequate team.

I saw a quote by R Dempster, on last night’s Cub game, about his bad inning where he loaded the bases and allowed a grandslam. He seemed to be saying that it was he, not Lou P, that decided which batter he would rather face. Lou went along with it, whether he agreed or not.

I think that happens on a significant proportion of IBBs. The pitcher’s preference is given a significant weight by the manager on the grounds that, even if he might be wrong, a pitcher will do better in a situation where he is more comfortable.

So, how much to weight that factor?

I think that happens on a significant proportion of IBBs.

I don’t think it does.

A manager will sometimes go out and talk to the pitcher before a potential IBB.  Whether he asks him what he thinks, I don’t know. If he does ask him, how much he weights that opinion, I don’t know.

But, I am pretty sure that a vast majority (75% maybe) of IBB’s are given from the dugout without the manager or pitching coach going to the mound.

So, if 75% of the decisions are made without going to the mound and lets say that of the 25% of the time they go to the mound, the “pitcher makes the decision” 1/3 or even 1/2 of the time, that means that around 90% of the time the manager makes the decision.  Even if you tweak those numbers a little in favor of the pitcher making or helping to make the decision, I think you will find that the manager makes the decision all by his lonesome more than 80 or 85% of the time.

And since the manager does and should have the ultimate decision, I don’t think it makes any difference to anything whether the pitcher gives his input or not, other than “for the record.”

As far as the manager “covering his ass” to preserve his job, I don’t think any manager does this intentionally (consciously), but there is no doubt that that is how some of these sub-optimal strategies evolve over the years.  The other principal way that strategies evolve over the years is by the way the success or failure is perceived based on the results.  That can be very misleading to the naked eye and naked brain.

Here is an example:  In blackjack, the correct play with a 16 versus a dealer upcard higher than a 6 is to hit of course.  Against a 10, it is almost break even.  Against a 7, it is a terrible play to stand.  Yet, some players stand on a 16 versus any dealer high card.  The “reason” is that you break over 61.5% of the time when you hit, and thus have no chance to win no matter what the dealer does, giving the illusion to some people that standing is better than hitting.

I think that almost all managers make their decisions on what they think will help their team win.  They just don’t have any way of knowing what the correct decision is in those fairly close situations, or even in some situations which aren’t even close (at least on the scale of what we in sabermetrics consider “not even close").

---"I think you will find that the manager makes the decision by his lonesome more than 80-85% of the time. “

I didn’t mean to overstate the impact of the preference of the pitcher. To me, the pitcher’s preference ruling, right or wrong, 10 to 15% of the time is a significant factor, to be included in the analysis, since it hasn’t been mentioned thus far.

I am not sure how that figures into the “analysis,” other than if the decision is wrong, we can assign 15% of the blame to the pitcher.

I guess it doesn’t figure into the “analysis”, then--unless you believe that the pitcher’s confidence in the situation makes a difference. They say it does, but of course I have no idea how real the effect is.

Well, whether the pitcher’s confidence figures into the situation or not, the results are what they are, which is why in my article I made sure I looked at the actual number of runs scored.

In fact, that was one of the main points of the whole article, which got a little lost in the shuffle.  People (saber-critics) often say that managers (and the players, I guess) “know” things that we don’t know, which may make their decisions have different outcomes than we expect based on our models.  Of course, by and large, I think that is hogwash.

And if that were true, with respect to the IBB at least, we would see run scoring after the IBB lower than we expect from the models.  We DON’T in fact see that.  We see run ACTUAL scoring almost exactly what any of the models I tested expected.  IOW, from the actual runs that were scored after an IBB in that situation, we see ZERO evidence that the manager (or the pitcher) somehow “knows” that the matchup against the next batter after the IBB is somehow better than the one with the batter being walked.  Of course, we can never know how many runs would have scored if the batter was NOT walked, but we DO know that the number of runs scored after the IBB is exactly what we would have expected using any of our models.  Which means, as I just said, that there is no evidence that the manager knows anything special about the matchup he chooses to have (like that batter being pitched to is slumping, or the pitcher is really keeping the ball down and thus should get a lot of DP’s, or the batter/pitcher matchup is particularly favorable, or he has particular confidence in his defense to turn the DP, or he thinks that the batter being pitches to will strike out).