THE BOOK--Playing The Percentages In Baseball

While I don’t disagree with any of your points, I don’t think the Phillies issued an IBB to Burrell to get to Conine, since they’re on the Phillies. What team was it that they faced? And do the Phillies still have a bad tactical manager, or were you referring to the other team still? It wouldnt surprise me if you meant the Cubs, thats for sure…

The hit and run,which I think should never be used.

It would be nice if we actually had data on when the hit and run was being used.  Does BIS track that?  Retrosheet seems to have stop using the “runner in motion” code in its PBP data.

Please educate me… why is it bad to bat lefties in a row?

Ian, the problem with batting lefties in a row is that it allows the other team to bring in a left handed pitcher to neutralize all of them at once (since lefty batters do less-well, on average, against lefty pitchers). If you dont clump the lefties, the other manager has to choose - either bring in lefty pitcher after lefty pitcher, burning up the bullpen, or leave the lefty in to face right handed batters also - who tend to do better vs. lefty pitchers.

1) Sac bunting for pitchers--In the “real world” I wouldn’t have a problem with them sac bunting, even in supposedly non-optimal situations. Pitchers have to focus on their pitching, and it may be advantageous to have a tangential part of their duties (batting) systematized as much as possible. But I would certainly insist that they practice bunting as much as possible.

2) The hit & run? I agree that it “looks bad”. But so did the bunt, before some of the more detailed analysis in The Book.

3) Bringing in a lefty reliever to face a lefty batter late in the game? I think every manager is well aware of such an advantage. So if they don’t make such a move, it’s probably because of such considerations as recent pitcher usage and workload. Before I would classify such a decision as a mistake, I would need to hear the manager’s attempted explanation.

Your point regarding runners going on a full-count and only 1 out is spot-on. If they are going to run, they need to run as if they are stealing the base outright, and not as if they live in some imaginary world where strikeouts are a rarity.
Along the same logic, I don’t understand why a sacrifice bunt (non-suicide-squeeze-variety) does not consist of the same technique as “bunting for a base hit”. Wouldn’t a the chance for an extra baserunner plus the chance for one less out be more appealing?

Not bringing in one of your better relievers (often the closer) in the 7th and 8th innings in a high leverage situation (when the game is on the line), and using closers in the 9th with 3 and 4 run leads.  I don’t have any data, but a strategy used by a lot of managers lately is to use his closer in 4-run games in the 9th.  I guess they are afraid to lose a 4 run game without their best pitcher in the game, which will occur about 1-2% of the time I would guess.

On using the closer in 4-run games in the 9th, I think part of that error is related to the common practice that when the manager warms up a closer, he wants to bring him into the game.

Usually, the times the closer comes into a 4-run game in the 9th is either when 1) the closer’s team had a save situation going into the bottom of the 8th or top of the 9th and scored runs to make it a non-saveable game or 2) the closer has not had an appropriate amount of work in the past couple of days due to their team blowing out or being blown out of games.

It seems to me that the managers rarely actively look to bring in closers to a 4-run game because they’re afraid another pitcher will lose that lead.

I think I agree with you on the 3-run lead problem, but I think that’s just the result of the save statistic “dictating” the usage patterns.

F

Well, I had a long post responding to the above comments and I just accidentally deleted it. (I know, Tango, I should write them in Word and then cut and paste).

I’ll re-write them later.

There are 2 more poor strategies that I forgot which really bug me.  They are partially the players’ fault and partially the managers’.

One is swinging at a bad or marginal pitch with a 3-2 count, especially when a baserunner is important (like 0 outs and no one on base) as opposed to a hit.

Obviously if a pitch is 50/50 whether a strike or not, the batter should usually take it with a 3-2 count.  That is not acceptable for most managers and players.  They do not want to get called out on a 3-2 pitch.  If I am managing a team and my players do not have lots of called strikes (and walks of course) on 3-2 counts I am going to straighten them out.  How often have you heard, “It was too close to take with 2 strikes,” even on a 3-2 count?  It should be the other way around on 3-2!  It was too close to swing at!

The other one is taking a 3-0 pitch with a runner in scoring position, especially with 1 or 2 outs.  Obviously with no one on base, especially with 0 outs, the walk is every important, but when the hit is important or the extra base hit is attractive, you want to have most of your batters (certainly your sluggers) NOT taking on a 3-0 pitch.  You see this all the time - a batter taking all the way on 3-0 and a runner on second and 2 outs.  This CAN’T be correct.

How about the batter that comes out and takes the first pitch after a pitcher just walks 2 guys or walks a guy on 4 pitches or throws 6 or 7 ballls in a row?  This also CAN’T be correct.  I don’t have any data, but I imagine that after a string of balls or after a pitcher has just walked a batter or two, he is likely to throw a strike and a fat one at that (not always a fastball - sometimes when they miss with several fastballs in a row, they switch to an off-speed pitch).

"How about the batter that comes out and takes the first pitch after a pitcher just walks 2 guys or walks a guy on 4 pitches or throws 6 or 7 ballls in a row?  This also CAN’T be correct.  I don’t have any data, but I imagine that after a string of balls or after a pitcher has just walked a batter or two, he is likely to throw a strike and a fat one at that (not always a fastball - sometimes when they miss with several fastballs in a row, they switch to an off-speed pitch).”

I’d sure like to see some data on this before passing judgment. 

There’s a huge game theory issue here, too.  The batter knows the pitcher has been struggling, and might just groove a pitch to get back on track, and therefore be eager to swing.  But the pitcher knows the batter knows this, so he might throw an off-speed pitch and get the batter to chase.

I think the key is to lock into a certain area where you can punish the ball, and if the ball comes there, unleash hell, and otherwise, make the guy throw a strike to you.

I wouldn’t say there is a huge game theory element.  All pitcher/batter pitch confrontations involve game theory (at least they should).

Since most batters take in that situation, it is all the more reason for the pitcher to throw a fat pitch.

If a batter swings at the first pitch and makes out, or even gets a strike, after the pitcher has thrown several balls in a row to the previous batter or batters, he gets bad-mouthed.

It is the manager’s responsiblity to say, “Hey the pitcher seems to be struggling with his control.  Rather than go up there looking to take oin the first pitch, look for a certain pitch in the zone and hammer away if you get it.”

Sure, most managers use the LOOGY late in a close game, some a little more than others, so it is not really a sticking point with me.  With a RH closer in, most managers will not bring in a LOOGY even if they do not have a dominant closer (of course the manager may not know he does not have a dominant closer).  They sometimes will not bring in a LOOGY even with a RH setup guy in the 8th.  If you don’t have a dominant closer, I am a big believer in using a tandem closer (RH and LH) in late, high leverage innings.

I do not KNOW that the hit and run is a bad play.  I think it is, but I am not sure.  There is not enough good data on when hit and runs occur, and I have never seen any research on it.  Obviously you need to know how mnany “extra” hits the batter gets with the runner going, and how often the runner gets an extra base when a hit occurs, versus what the damage is from swinging at any pitch, especially when ahead in the count (hit and runs are usually done on 2-1, 2-0, and 3-1 counts, occasionally at 1-0 and 1-1).

That is another bugaboo, baserunners not running hard on hit and runs.  There is no such thing as a hit and run versus a run and hit anymore.  On a hit and run, the baserunner should run like he is stealing.

There is also not much game theory on the hit and run, as managers rarely pitch-out to thwart the hit and run as, as I said, they are usually executed on hitters counts (when you don’t normally pitch-out).  So if they are correct occasionally, they should be correct a lot.  In other words, there really is no surprise element involved which would make managers want to use it only occasionally.  Either it is correct in certain situations or it is not correct (or it is 50/50).  If it is correct, I would think that you would use it much more often.  Since it is not used that often, I suspect that it is incorrect (if you get what I mean).  Very few strategies should only be used very occasionally unless game theory is involved or they are only correct in rare circumstances.  For example, the only reason that you don’t see the suicide squeeze more often is that it is easy to thwart (and costly for the batting team) with the pitch-out.  If managers used it more often, the defense would pitch-out a lot and the offense would have to stop using it so much.

I watch a ton of games.  It is becoming common practice for managers to use closers in 4 run games, presumably because they are afraid to lose the game.  In this era of high offense, you see many more 4 and 5 run (or more) innings by teams.  Managers in their heads overstate the frequency of them.

If your closer needs work, then of course you bring him in regardless of the score, or with a 3 or 4 run lead.

The idea is to simply use your closer enough to keep him sharp, but not too much to overwork him and in as many high-leverage situations as possible.  Managers clearly do not do that and do not understand that concept.

And of course, you must give him time to warm up and perhaps it is an issue to have your relievers have some prior knowledge of what their role is going to be, although I think that issue is overstated (as is the issue of hitters “liking” and “not liking” certain slots in the batting order).

As far as bunting like you are bunting for a hit in a sacrifice situation, there is probably a place for that, depending upon the bunting skill and speed of the batter.  However, since a bunt that advances the runner(s) where the batter makes out is not that bad an outcome, it is also OK to try and keep the bunt fair as much as possible.  When you are bunting for a hit, you generally either bunt it foul or you bunt it perfectly, at least if you are bunting down the third base line.  So since the RE charts are different in a sac situation and when bunting for a base hit, the approach probably SHOULD be different.  However, OF COURSE the batter should be running as hard as he can when he is sac bunting.  You often see them NOT do that.  There is no excuse for that.  They figure that as long as they moved the runners over, they did their job.  Of course, they generally reduced their team’s chances of winning when making out and moving over the runners.  They don’t know that and neither do their managers.

When the hit & run works, its a thing of beauty.  I guess the question is how often it succeeds and fails, and what the cost/benefit is.

If I ever see a closer pitch the 9th with a 4 run lead, its almost always because he hasn’t pitched in a while and needs some work.  Can’t criticize that.  I’ve noticed a lot more closers this year going 4 outs, sometimes 2 innings for a save.  That’s a good thing, and maybe somebody’s reading The Book.

I don’t think a closer pitching more than 2 innings is ever a good idea unless its late season, pennant race do or die, or a playoff elimination game.  The season is too long and you’ve got to balance workload with the desire to obtain the most leverage in game situations.

Re: taking a pitch on a 3-2 count, are you overlooking the effect of fouling off pitches?  We’ve all seen those great at-bats where a hitter fouls off close pitch after close pitch before finally working a walk after ten pitches or so.  If you’re one of those lucky hitters with the talent to foul pitches off consistently, shouldn’t you be swinging at anything you think might be in the strike zone?

As a whole IBB’s are about at the break even point for both leagues for 2003-2005 so if some managers are losing runs for their team on them then others are gaining runs.

As far as hit and runs are concerned a lot depends on how a team defines a hit and run.  Is the runner running as if it is a straight steal?  He should be.  Is the batter required to swing at the pitch no matter where it is?  He shouldn’t be.  A hit and run that is a straight steal with the batter having the option of swinging should be an effective strategy a lot of the time.  DP’s would happen on LD’s to the infield and on some of the LD’s to the RF, but they only occur on 3% of hit balls.  DP’s would also occur on a portion of P outs to the infield, but pop outs only occur on about 7.3% of hit balls.  Compared to the advantage that occurs on the 45.8% of the hit balls that are grounders and the 16.8% of the hit balls that are FB and LD 1B’s and 2B’s and I think the potential for the hit and run looks pretty good.  Would be nice to have the actual data though.

Most strategic choices are highly situation-dependent. The problem with much analysis, as I see it, it the tendency to define a situation in a way that makes it easier to analyze. This might not correspond to the reality of a game situation, which can change pitch by pitch.

Let’s take the hit and run. The benefit on offense is in getting the defense moving and out of position, the chance for a base hit goes up, the GIDP is more difficult to turn, and the baserunner has a better chance of advancing. The downside is the batter doesn’t have the option to take a bad pitch, the baserunner is committed to take second base if the pitch is missed, and won’t have time to get back to firstbase if the batter lines to an infielder. So when might the H&R be a good play?

Well, you wouldn’t want to do this with a power hitter who can drive the runner in from firstbase. How about with a singles hitter at the plate and the count in his favor. Or against a control pitcher who doesn’t have a high K rate. Perhaps a groundball pitcher vs. a groundball hitter matchup. Managers don’t typically have power hitters throughout the lineup. They have to work with the talent they have on the roster and find ways to take advantage of what the players can do. An Ozzie Guillen type batter, who didn’t work the count but could put his bat on anything close, would probably be a good H&R man.

There are, no doubt, plenty of situations when the H&R is a poor option. But I wouldn’t write it off as a bad play.

Peter, yes by definition, the hitter must swing on a hit and run, which is the problem of course.  As I said, the question is whether having to swing at a bad pitch and the runner occasionally getting thrown out (50% of the time or so when the batter misses or does not swing - I forgot the actual number) outweighs the extra hits and extra baserunner advances.  I suspect it does, but I am not sure.  One reason I suspect it does is that if it did not, it would be correct for all non-power hitters with decent bat control to virtually ALWYAS hit and run.  As I said, it is not really a strategy that needs a surprise element and it is NOT very situationally dependent (it is SOMEWHAT, since, as Steve said, you probably don’t want a power hitter at the plate, and you probably don’t want someone who is going to swing and miss a lot).

Mylons, sure if a batter can foul off enough pitches, then swinging at a close 3-2 pitch is more of an option.  Even so, batters still swing at too many 3-2 pitches, mostly for the reason stated (they are taught to swing at anything close with 2 strikes, but no one told them it depends also on the number of balls AND they don’t like to be called out on strikes with a 3-2 count).

Peter, what makes you think that overall IBB are at the break even point for the league?  I would think that would be really tough to figure out.  My guess is that the average IBB has a significant loss expectancy (theoretical or actual WE expectancy with the IBB minus the theoretical WE without the IBB).

One thing I have been wondering lately about the steal attempt.  Have sabermetricians been overlooking the value of the steal attempt when the batter puts the ball in play.  As Steve suggests with the above data, the frequency of the line drive (and occasional short fly ball) DP should be a lot less than the extra base hits from the infield being out of position and the extra baserunner advances.  Even if this value is a little, it might increase the value of a steal attempt quite a bit, maybe even reducing the BE point to 60 or 65%.

Speaking of steal attempts, as we imply in the chapter on baserunning in “The Book,” batters should NOT be taking pitches (that they might ordinarily swing at) with a basestealer on first.  This should be obvious anyway.  If the steal attempt is usually around BE anyway (more for the better basestealers of course), then clearly you don’t want the additional disadvantage of the batter deliberately taking pitches.

I even heard an announcer (on the CIN broadcast I think) say something like, “The batter should take a pitch OR TWO” in order to let the runner steal (I think it was Phillips).  Taking TWO pitches!  You can figure out what the BE point of the steal would have to be in order to take one or two pitches.  Obviously it would have to be higher than the usual 70% or so.

The other question is since attempting a steal when the batter puts the ball in play is probably a GOOD thing, why would you want your batter to take a pitch anyway.  With 2 outs, you might, but with 0 or 1 out, I would think you definitely would like your batter running AND the batter swinging if he gets a good pitch (he bats like he normally does).

Anyway, these are the types of things that players, managers, coaches, etc., need to resolve.  They are fairly complicated issues (and require some math) and players, managers, et. al are not capable of doing the proper analysis in order to resolve them.

All the experience and “baseball smarts” in the world are not going to give you the correct answer to these questions.

It is like blackjack again.  No matter how long you have been playing (or working in the casino industry), you are not going to be able to “figure out” basic strategy.  Only a mathematical analysis (or computer sim) is going to be able to do that.

When I have been on teams that have signs for steal, hit and run, and run and hit then steal meant runner goes batter does not swing, hit and run - runner goes and batter must swing, run and hit - runner goes and the batter has the option of hitting.  When I have been on teams that only had signs for steal and hit and run, managers sometimes defined the hit and run as always swing sometimes as optional swing.  The problem for sabermetricians is that we can not identify what type of play is on from what we see of the play.  A run and hit where the batter swings looks exactly like a hit and run.  A run and hit where the batter does not swing looks exactly like a straight steal.  The managers best strategy is probably to have all of the three plays be run and hits with only a few exceptions for runners of superior base stealing ability.

Research that I have done on the IBB indicates that the actual cost in extra runs for the actual lineup situation in which it is employed is only .10 runs instead of the .16 runs that putting a man on first in an average situation would cost.  Using the actual hitting abilities of the man being IBB and the following batters I have calculated that the chance of scoring runs by letting that batter hit in that situation is .07 runs greater than normal so the actual loss in run value for the intentional walk is about .03 runs.  However, because the situation often occurs when the win value of a single run is high, the actual win value of an intentional walk is slightly better than break even.  Hence, some managers are using it well and others not so well.

Sounds about right with the IBB Peter.

Table 11 in The Book shows that the win probability when the IBB was issued was .724 for the batting team.  The win probability post-IBB was .734.  This assumes all batters are average, meaning we are really only measuring the impact of the walk in the particular game state (predominantly with first base open, and at least one out).

However, if it’s Bonds being walked, then it’s a good tradeoff.  That is, if an average batter has his team starting at .724, then Bonds specifically batting means the batting team has a much better chance to win, say at .739.  If you walk him, and you have an average batter at plate, then the .734 applies.

So, what you really need to measure (for the most part) is the gap between the batter and the guy on deck.  And, you need to apply the LI to that situation.  If when an IBB occurs the LI is 2.2 (implying a runs per win of 5.0, and high-leverage), and the gap between the two batters is .05 runs, then this is a break-even play.  (i.e., .05/5=.01 wins)

(The figure reported in The Book is 17.8 runs per win, nowhere close to 5.0, meaning it’s a low-leverage play.  Hard to believe, but likely explained by all the NL IBB with the pitcher on deck.)

Just guessing, but at least a quarter of the IBB are poor plays, causing at least a .01 win change, using the qualities of the specific players involved.  Likely another quarter are in the .005 to .01 range, meaning not a good play, but quasi-defensible.  Continuing my guessing, probably 40% are in the -.005 to .005 range, meaning more or less break-even.  And a smidge, maybe 10-15% are in the -.005 and lower, meaning a good IBB.

The change in win probability of IBBing a batter in a specific situation can be calculated directly by subtracting the win probability that existed before the IBB from the win probability at the end of the half inning.  The only thing you have to estimate is the win probability that would have existed at the end of the half inning if you had let the batter hit. Then you can calculate the expected change in win probability of letting the batter hit.  Comparing the two values will tell you if it was a good decision or not.

Peter, I don’t have a particular problem with what you are doing, but I don’t think it’s necessary to do the change in win expectancy to the end of the half-inning.  pre and post batter is sufficient.

The key is, as you said, to estimate the win prob if he was being pitched to.

No, doing the win probability to the end of the half inning is essential.  What you want is the runs that actually occurred with the lineup that actually existed. If you use the win probability post batter you are using the estimate of the runs that would occur during the remainder of the inning for an average lineup and what the win probability would be with an average lineup, not the actual one.  The difference is substantial.  As I pointed out in a previous post an IBB results in an actual increase in run value for the rest of the inning of .10.  The estimate of expected run value change with an average lineup is .16.

We are probably saying the same thing. 

The win probability can include whatever parameters you want.  In my case, I would have a win probability table with Bonds at bat, followed by the exact hitters, or with Bonds on base, and then those same hitters coming up.  The effect of Bonds walking is simply comparing the win prob of Bonds at bat plus those batters, to Bonds on base, and those batters at bat.  (And of course, Bonds’s true talent estimate would be based on the base/out state he sees.)

To get real technical, you’d have to take it to end of game since the IBB will increase the chance of said batter coming to bat in future innings.

"Anyway, these are the types of things that players, managers, coaches, etc., need to resolve.  They are fairly complicated issues (and require some math) and players, managers, et. al are not capable of doing the proper analysis in order to resolve them. All the experience and “baseball smarts” in the world are not going to give you the correct answer to these questions.”

This is only partially true.  The unwritten book of baseball conventional wisdom is the distillation of many years, and many thousands of games, of real experience.  True, this knowledge has come through trial-and-error, not mathematical analysis, but I’m amazed by how often sabermetric analysis reveals that standard baseball strategy is essentially correct. 

SBAs are a great example.  The analysis in The Book finds that SBAs are made and not made when they should be, in proportions with uncanny similarity to what teams actually do.  And I’m convinced that once we figure out how to account for the benefits of SBAs when the ball is in play (or discount the CSs arising from blown H&Rs), we’ll find that teams are stealing almost exactly as much as they should be. 

Defensive talent seems to be allocated to the right places on the diamond, and the amount of reduced offense accepted at key positions seems about right given the defensive gain.  Batting orders aren’t quite optimal, but they’re pretty close. And so on.  Obviously, there’s room for tactical advances—but they tend to be very small, as The Book frequently emphasizes. 

I guess it’s a glass half-full/half-empty issue (and we know MGL is a ‘half-empty’ guy!).  But I’m generally impressed by how much the process of thousands of players and hundreds of managers seeking competitive advantage over the decades --and emulating success when they see it—has produced a set of generally sound tactical and strategic conventions.

This thread is probably dead (a nice title for a song), but Guy, because OVERALL managers gets lots of things right (at least “rate-wise"), by no means means does that mean that they understand WHY they do so, and know or understand the correct strategies and BE points, etc., such that they do the correct or optimal thing all or even most of the time.

Whereas OVERALL they may get the stolen base attempts and sac bunt attempts correct (and other things), they are quite often using incorrect strategies and making incorrect deicisons along the way.

There is a huge difference between stealing almost exactly as much as they should OVERALL and stealing optimally on a case-by-case basis.  A HUGE difference.  Same with sac bunt attempts and batting orders.

So we are not talking about “half-full, half-empty” or making semantic distinctions here.  We are talking about two entirely different things.  For example, even if the SB % were exactly correct overall for every possible game situation, that does NOT mean that many, many SBA were not executed improperly, if you will excuse the multiple negatives.

Of course it is a semantical issue what constitutes “good or bad” managing, but as Tango likes to say, the numbers speak for themsleves.  Every time a manager makes a mistake, and I am alleging that they do so almost on a daily basis, it costs X amount of wp, and at the end of the year, we can compute how many wins a particular manager or managers in general cost their team, as compared to using optimal strategies.  And of course, whether a decision or strategy is “optimal” or not includes all of the usual “intangibles.” Some things, like bullpen usage, probably lend themsleves to including more intangibles than other things, like the IBB (off the top of my head).

By the way, if your “theory” is correct, that through trial and error and the “wisdom” of experience, managers get many, if not most things, reasonably correct (which I vehemently disagree with, BTW, at any level), why would almost ALL IBB be incorrect, and why would almost ALL number 2 hole hitters be mis-placed, and why would the best hitter be put in the 3 hole rather than the 1,2, or 4 hole where they generally belong, and why would closers be used in the 9th in 3 run games but rarely in tie games or when down by a run or NEVER in the 7th inning, etc.?

One of the reasons I used the BJ analogy (and I am an expert on BJ, BTW) is that, similar to strategy in baseball, there is absolutely no way, through experience, trial and error, etc., to “figure out” perfect basic strategy in BJ without using a computer (or doing the math).  You can come close, I suppose, but you cannot “figure out” the close plays.  In baseball it is even harder, as it is way more complex and many things that are incorrect seem intuitively correct.  Not to mention the fact that many of the conventional wisdoms and strategies are derived from the manager’s desire to avoid failure, minimize risk, not look bad, etc., rather than maximize his wp.  An analogy in BJ is that because the player must play first and if he busts the hand is over and he has no chance to win, the average player likes to stand pat more often than he should.