THE BOOK--Playing The Percentages In Baseball

Managers are clearly more interested the using the closer in the ‘9th’ inning, instead of maximizing his expected value by using him more in the 8th (or sometimes even the 7th, I guess). Instead of saying that they do this to blindly adhere to the save rule, maybe it’s the other way around. Maybe they believe that the regularity in the role, in the real world, more than compensates for the slightly suboptimal usage--and that the save rule simply happens to coincide with that orientation. The 3 run lead 9th inning save helps to get the closer a proper number of total inning workload, while not changing his role.

Doesn’t there almost have to be a better explanation of why the save rule usage has become so universally adopted, than just saying that managers can’t think independently at all, and are only covering there arses? Maybe it’s just an accident that the save rule happens to coincide with the optimal pattern of reliever usage in terms of leverage vs role in modern baseball.

I think I’ve heard enough from insiders about how managers think, that their number #1 rule is to cover their a$$e$.

There’s simply no way that his team and his media would stand for one second a situation where they remove a better reliever for a worse one entering the 9th inning.

This happens in corporate America, where they high powered negotiators come in and work out the deals, while their minions are left to dot i’s and cross t’s to finish the contract.

Apparently, have a 3% failure rate in dotting i’s is too great to handle, even if it increases the chances of getting more contracts signed in the future.

4%, not 3%.

With the understanding that the high powered negotiator would have a 2% failure rate in dotting i’s.

When I was toying around with the idea of writing a book (which I still am), one of the chapters that I started writing was “The four most powerful words in baseball (and life): It’s not my fault.”

As a field, Sabermetrics isn’t a secret any more.  Yet, we still get below average hitters hitting in the 2 hole, third base coaches who are far too conservative, and the use of batting average to justify all sorts of stupid moves.

I think things are a bit more complicated with the role of closer than people in the sabermetric community understand. As has been pointed out ad nauseam, it makes the most sense to use your best reliever at the most critical junctures in the game, not just in the ninth inning, and some quite mediocre pitchers can often effectively fulfill the closer role, Joe Borowski being a case in point.

One reason that managers use their best reliever in the three-outs-in-the-ninth role is to cover their rear ends, but there’s more to it than that. In the first place, there are the player expectations. If the manager already has a closer on his team, that player will likely resist any change in the way he is used. That’s because the number of saves he has in a season will play a large part in determining the monetary value of his future contracts. If the manager doesn’t have a pitcher who thinks of himself as a closer and uses his relievers in a more effective manner, the saves will probably get spread around, and there’s sure to be some guy who will be upset because he’ll think that he’s proven that he can close, and he’ll want the opportunity to rack up the saves and get the big money.

Larry Bird and others associated with the NBA have said that there are players on every team who don’t want the ball when the game is on the line. I suspect that there are pitchers like that as well. If a pitcher gives up a lead in the sixth inning, he can tell himself that he didn’t really cost his team the game, that they could have come back in the last three inning. If he gives up the lead in the ninth inning, the onus is going to be on him, both in his own eyes and in those of his team mates. Some guys just aren’t going to want to be in that situation.

My point is that there are a lot of reasons why the current closer role persists, and a manager’s fear of failure is only a part of it.

I think we understand very well, Chuck.  We agree that some pitchers are not comfortable in the 9th inning.  I think it should be easy to find TWO pitchers who are comfortable in the 9th, don’t you think?  I’m not asking for 5 or 6, just 2.

In any case, what happened in the 1970s and 80s?  Was agent white added to mother’s milk so that the post-Eck babies have some altered mindset from which they can’t get out of?

Chuck, you’re right on a number of counts.  There would be players who would be angered because in a lot of minds, saves = good relief pitcher = $$$.  You’re also correct that people place far too much psychological value on the events of the ninth inning.  (This was chapter 2 of my book, “The Myth of the Modern Closer.") The fact that this is a mythology shared by managers, fans, the establishment, teams, general managers, and the pitchers themselves doesn’t make it any less inefficient.  Changing a culture just takes someone willing to put his neck on the line.  Where have the brave men gone?

Pizza, if you ever do write that book, put me down for a preorder…
-j

Someone from an MLB front office is willing to make the pitch to their minor leagues:

With a plan that is clear, has very few rules and will end up using the setup man and the closer just as often for just about as long, I wouldn’t be surprised if we could get some interest for this in our minor leagues.

I say that the plan has to be simple. Certainly, there is a tradeoff between the complexity of the rules that you use and the degree of leverage you can get out of your top guy. In the real world though, simplicity is going to be a big factor. Heck, look at how simple the rules seem to be to bring in a closer.

The usage needs to be the same. For now, anyway. The arguments that this will be changing the frequency of his usage, his total innings, or whatever should not (IMO) be addressed during this first step. All we want to be able to say is, Given the same usage (frequency, duration) and only switching around when you bring in the pitcher labelled as the closer, the following simple set of rules would increase your win expectancy by x.

If we had those simple rules and the x, I could make the argument. I don’t have time to do that, but was wondering if someone from your site would be interested in coming up with that. Pizzacutter did some good work, but what I am talking about is really diving into it and being sure how often and for how long the setup man would pitch during the year and the closer would and how many wins one could expect.

What do you think? Could we post a request where we just keep it as “A member of the front office of a Major League team asked me blah blah blah”.

To get some simple rules that could squeeze (a couple?) more wins out during a year without changing overall usage would be great. We will worry about the culture. Thanks.

The one-line summary is: keep the same number of games for your ace reliever, keep the same number of innings for your ace reliever, but change WHEN he gets to come into the game.

The only change I’d make is to stop the 3-run lead, bases empty ninth situation.  Don’t bring him in then.  Do bring him in the 8th inning of a 1-run lead.

However, the “same number of innings and games” is a bit of a bother, since this will naturally force us to remove the ace reliever a few times after the 8th, and preclude him from pitching the 9th.  Even in the heyday of Goose, Rollie, and Sutter, that never happened.  Those guys were brought in early, but they still finished the game.

So, I don’t think it’s a fair test to ask to simply change the WHEN without also changing the HOW LONG.

As detailed in The Book, the best ace reliever usage pattern was by Troy Percival, 1999-2002.  He got close to the maximum LI that I’d look for.  I’d say a study of how he was used would be ideal.

Alternatively, Trevor Hoffman has above 2.0 LI numbers every year since 2002 (average of 2.10):
http://www.fangraphs.com/statss.aspx?playerid=1035&position=P

I didn’t check the other relievers, but that’s probably the highest of the Fangraphs era.  However, the Padres bullpen LI averaged around 1.1 since 2002, so there’s a good chance that Hoffman was able to rack up the high LI because the Padres were involved in alot of close games to begin with.

For aspiring sabermetricians, I’d recommend poring over the Relievers chapter in The Book, reading Pizza’s article linked at the top, going through the Fangraphs great links, and reading Rob Wood’s excellent article here:
http://www.baseballthinkfactory.org/files/primate_studies/discussion/robw_ood_2003-02-04_0/

Someone from a real MLB front office said I did good work?  I’m blushing all different shades of red.

From what I found, trading the ninth inning, 3-run leads for eighth inning, 1-run leads is a net workload increase of 5 innings for the closer on an average team (perhaps he just wouldn’t pitch as many garbage time/mop-up innings...) The biggest problem that I ran into was that it was hard to model the potential effects of fatigue within the game.  Closers don’t often go 2 innings, so I couldn’t really study what might result (note to managers: could you?  It’s for research purposes...) I assumed that closers were just as effective in their second inning as their first, which might be flawed.  I’d need to look a little deeper on that.

And as to the “cultural” aspects, this is going to be a constant fight.  In my “real job”, I’m familiar with a branch of psychology known as health psychology.  In it, we discuss how to get people on board with new practices that will benefit their health (stop smoking campaigns, using sunscreen, etc.) and some of the barriers that are in the way there.  This un-named official is someone who would be an “early adapter.” It’s something that’s needed if a policy or intervention is to be adapted widely, or even within one system.

Well, Mr./Ms. Early Adapter, if you’re reading this, I accept your challenge.  I’ll dig a little bit on this.

Pizza says:

I assumed that closers were just as effective in their second inning as their first, which might be flawed.  I’d need to look a little deeper on that.

See The Book, Table 93, p.230, where I conclude:

The Book Says:
Relief pitchers should be brought in the eighth inning of close games to face at least one batter. Facing one or two batters in the eighth inning results in no change in performance in the subsequent ninth inning.

***

Also note that at the bottom of p223, top of 224, I lay out a reasonable case where the LI for your ace should be 2.3.

Interesting that Troy Percival and Hoffman have the highest leverage ratings.  They were/are used almost exclusively for 9th inning save situations, including the dreaded 3 run, nobody on base save.

Just taking a quick look at Hoffman, 2006.  I count 64 games (out of 65) where he pitched with bases empty, 0 outs.  This is how it broke down:

8th: 1 (blowout game)
9th: 58
XI: 5

So, I think Rally is right on that.

Focusing on the 58, the average LI that he entered that inning was 2.05.  Included in that is 15 games with a 3-run lead, 1 with 4-run lead, and 2 blowouts.  That’s 18 games that I would not have wanted him to pitch.  The average LI of these 18 games is 0.79. 

(The LI for 3-run leads are 0.8 in the top of the inning and 1.0 in the bottom.  These are not terrible choices, but simply not good choices.)

We can see that Hoffman simply didn’t have many blowout games at all, unlike your typical closers.

In the highLI games in the 9th (40 times), his average LI was 2.61.

In all, what saved Hoffman in 2006 was that he was used in only 4 blowout games.  This is how he broke down:
blowouts: 4
average leverage: 15
high leverage (min 1.5): 45

What did I propose in The Book?  On pp 223-224:

If we can give our ace 45 high-leverage games (at a leverage level of 2.8), 10 normal-leverage games (leverage of 1.0), and 5 low-leverage games (leverage of 0.2), this comes out to a leverage of 2.3. This is our target. If a manger can better schedule his relievers, the ace’s leverage will go from 1.9 to 2.3. This is a 20% gain! It’s like turning someone who can impact his team by winning four more games than an average reliever into winning five more games instead. A 1-win gain is substantial, when dealing with optimization.

Now, that’s what I call pretty cool.  I said to give your ace 45 high-leverage, which is the exact count that Hoffman got.  I said 5 blowouts, and Hoffman got 4.  I said 10 others in average leverage, and he got 15.

So, this is a great job of not having wasted outings for Hoffman.  While he was used a bit too much in the average leverage, there’s not much I can complain about here.

It’s not so much stopping the 3-run 9th inning situations, but *waiting* for those situations, to the point that you end up having to waste him in blowout games if he doesn’t pitch in an otherwise high-leverage game (in the 8th).