THE BOOK--Playing The Percentages In Baseball

Izzy had an LI of 2.34 to lead the league.

***

If I take the top 150 relievers by PA, and sort by LI, the top 30 averaged 1.88.  Followed by the setup guys at 1.42, then 1.07, 0.85, 0.62.  The rule of thumb seems to be that each reliever gets 30% more leverage than the guy below him.

The rest of the scrubs averaged 0.79. The average LI for relievers was 1.03.

What is interesting here is that the guys that are not in the top 5 get more leverage opportunity than the guy in the 5th slot.  That is, managers have decided to give their 5th best reliever unconditional mopup duty, because he doesn’t trust them, but in the new blood, the 6th through 15th reliever, he might have a pearl, and gives those guys more of a shot.

The least trusted relievers, with at least…

... 10 games: Agustin Montero (LI of 0.18)
... 20 games: Bruce Chen (0.24)
... 30 games: William Eyre (0.31)
... 40 games: William Eyre (0.31)
... 50 games: Hector Carrasco (0.54)
... 60 games: Will Ohman (0.57)
... 70 games: Will Ohman (0.57)
... 80 games: Geoff Geary (0.88)
... 90 games: Salomon Torres (1.57) *

* He’s the only reliever with 90 games, so he’s also the most-trusted at that game level.  And his LI was 1.57, which is somewhere between setup and closer roles.  He also had a WPA of +2.9 wins.  So, he certainly doesn’t belong on this list.

You have to be really careful about drawing any conclusions about the manager for one simple reason.  I’ll try and explain it.

Say I have a guy who is a great reliever with a great projection (that’s what makes for a great reliever - a great projections, since that means that he was historically great).  Let’s also say that he does not have injury that we know of.

Let’s also say that this great reliever had a bad or terrible season, as far as we know by luck alone.  In 60 innings or so, this will happen to a lot of great relievers, by luck alone, although off the top of my head, I don’t know the SD of ERA in 60 innings.

If the manager is a good one, IMO, he will give this pitcher all high leverage innings all season long and at the end of the season, he is going to look like an idiot (the manager that is), because he would have a pitcher with a bad ERA with a high average leverage.  And just the reverse for a bad (historically) reliever who had a lucky year.

So honestly, I really don’t like these kinds of constructs.  While on the average, the pitchers with good years are indeed good pitchers who should have high leverages and vice versa for the bad pitchers, by no means should this be the case for every pitcher.  A team with a great manager who understands the “gap” between a pitcher actual performance to date all season and his projection is going to constantly be putting good relievers with bad performances to date in high leverage situations, and bad relievers with good performances to data in low leverage situations.

IOW, I’d love to see these same numbers, but rather than leverage index matched up with THIS year’s performance (ERA), I’d like to see it matched up with the last 3 years weighted, either including or not including this year (including or not including this year is a little tricky).  That way we can really discern the smart from the stupid managers.

Again, this is especially true for relievers with few innings.  Dumb managers will take historically bad relievers who have had 20 or 30 innings of good performance and start putting them in high leverage situations.  They will also take great relievers like Cordero, formerly of Texas, who are having a bad season, and start putting them in lower leverage situations.  Smart ones of course will pay little attention to short-term performance except how that might change their true talent, via injury or whatever…

There’s no question that you have to select your pitcher before his performance, not after. 

***

Possible injuries aside, let’s look at Izzy. His WPA in the 4 seasons coming into 2006 was +11 wins, which is great!  (For comparison purposes, Mo and Wagner were +14.) So, he should have been slotted to be the ace, and get the LI of 2.0, which he did.  And, by the time the numbers told us that this is a different Izzy, the season is already almost over.  For example from 2003-2005, he had 11 HR on 129 hits, while this year he had 10 on 47, putting him at 2.7 SD from his base (it was 2.4 through July).  He had 163 K in 2003-2005, with 58 NIBB, giving him a K per K+BB rate of .74. compared to the low this year of .60, or 4.2 SD from the mean.  So, pretty clear that he’s changed.

A fair analysis of Larussa then should see a shift in Izzy’s usage late in the season, perhaps not to the extent that Izzy now sucks, but that at least he’s no better than a setup guy, if we now look at all his data from 2002-2006 (with more weight on this year’s data).

In his last month of play, Izzy was -2.0 wins in WPA.  This doesn’t happen unless: the manager puts him in high LI games, and the reliever blows it time-after-time.

Larussa rolled the dice and he lost.  Was it just bad luck, that he couldn’t have predicted that, who knows.  But, I’ll blame the dice holder anyway, fairly or not.

***

As for the Angels, like I said, I don’t blame the Angels, because of their talent base.  That still leaves us with Carrasco, an average reliever, performing well, who doesn’t deserve to be used in mopup duty.

***

Finally, Ryan Dempster.  The high BB Dempster is not my idea of a closer.  He had great 2005 followed by horrible 2006 according to WPA.  His 2005 and 2006 performances are fairly consistent on the surface, but WPA has him going from +3.7 wins to -3.1 wins, an enormous 6.8 win shift.  The biggest culprit is that he allowed only 22% of runners to score in 2005 (Pedro-territory) to allowing 37% to score this year!

With about 120 runners on base, 1 SD = .042, making him go from about -2 SD to +2 SD year-over-year.

Luck?  Sure, maybe.  But, the dice were in Dusty’s hands, and I’ll blame him.

And we certainly can’t go by a pitcher’s absolute talent level.  We (the manager) have to go by his talent level as compared to the other guys in the pen.  So even if Issy becomes a true “setup level” reliever at some point in the season, his optimal LI would depend on the next best guy in the pen.  If there is no one better, again, projection-wise, then Issy should still get the highest LI situations.

By far and away, the 2 biggest mistakes that managers make are: 1) judging relievers by their season-to-date performances, which is extremely dangerous because of the sample sizes they are dealing with, and 2) being too regimented about who pitches when, which has been talked about and analyzed ad nauseum of course.

With that kind of bullpen (Cards), I guess the biggest mismanagement really falls to the GM.  (Though the team WPA for hitters, starters, and relievers were also close to league average.)

Nonetheless, the next highest LI in the bullpen was 1.20 (Izzy was 2.34).  That is a substantial drop.  The team LI was 1.02.  And like I said, in a typical team, each reliever has 30% more leverage than the next guy.  Izzy had 95% more leverage than the next guy!  Braden Looper is no great shakes either, and they did have the rookie, but the spread should have been much tighter. 

Otherwise, I’m in agreement.

If managers would simply bring in the best reliever available when the game is on the line and their worst relievers when the game is not, considering the handedness of the batters (in terms of which relievers are the “best” or “worst") coming up, pretty much everything would take care of itself.  It is really not that hard.  The two hurdles to that, even if a manager understands that concept, is that he does not consider his best and worst relievers to be the those with the best and worst “projections” according to some ERC formula or something like that, AND they are too regimented on how they are “supposed” to use their pen.

For example, I think that a manager basically realizes that a 3-run lead in the 9th is NOT a situation where the game is on the line, but somehow he can’t make the connect to the fact that he doesn’t have to use his closer.  Ditto for a one-run lead bases loaded, two outs in the 7th.  He realizes that the “game is on the line” but he can’t make the connect to the fact that it might be a good time to use his closer, even for a couple of outs.

For example, I think that a manager basically realizes that a 3-run lead in the 9th is NOT a situation where the game is on the line, but somehow he can’t make the connect to the fact that he doesn’t have to use his closer.

While I agree with everything else you say in this thread, and I obviously agree that the current bullpen setup is far from optimal, I don’t think it’s that the managers “can’t make the connect”.  I believe its a conscious decision that most managers make to use their closer in as many save situations as possible. 

Why?  A sense of obligation to help the player make as much money as possible.  Saves are still a moneymaker in baseball (hello, Todd Jones).  Pitchers want to be starters or closers because that is where the money is.  Unless you pitch for the Cubs, there’s very little money to be made in the 7th or 8th inning, because there’s no equivalent to the save for the setup man that they can take to arbitration or use in free agency. 

A manager who used his best reliever in close, tight situations and then handed the easier “save opportunities” to inferior relievers would appear to the players to be taking money away from his closer.  For better or worse, guys like Trevor Hoffman could care less what their leverage index is, but if you start messing with their save opportunities, they’re going to care. 

Now, this is obviously fixable; teams simply have to stop paying for saves and pay their relievers based on a more realistic view of their performance, and this whole issue goes away.  But that’s above the manager’s head, and we can’t lay all of it at their feet.

I don’t think there’s a manger in baseball that has the authority to unilaterally go against the grain and start using his best reliever in the 6th or 7th inning of close games.  That’s going to need to be an organizational decision, where the front office, the coaching staff, and the players get on the same page and are all on board. 

They’ll need to convince the local media to not crucify them in the papers everytime a save is blown in the 9th inning by the third or fourth best reliever, and they have to convince the coaches that it’s worth their while to be patient explaining the logic to the players and the reporters. 

The manager is culpable to some extent, but I’d stop far short of using bullpen optimization as a proxy for their intelligence.  I’d imagine several major league managers understand the basic ideas of leverage index (even if in their lingo), but they don’t have the support nucleus to actually go ahead and implement it into their real-world situation.

I agree that optimal bullpen usage, especially using the closer less often in easy save situations (and more often in non-save situations) has to be an organizational decision.

I’m not sure that managers are a making conscious decisions to “pad closers’ save numbers” all the while realizing that it is not necessarily optimal in terms of their team’s wp.

I really do think that there are some managers out there that if you sat down with them and explained some of these stratgies, that they would be capable of understanding them and perhaps not even opposed to using them.

One big-league manager who I won’t name said that one of his relievers stated that he would rather NOT pitch in the 9th.  So, this mindset might be in the players head, that they consider the 9th inning more crucial than LI says it should be.

LI tells us how crucial a player should feel a situation is.  How a player perceives that situation can be far different.

I agree that by and large the LI should reflect the “pucker factor” but there are certainly going to be gaps in that perception.  3 and 4 run leads in the 9th are definitely going to be one of those gaps as players and managers are scared to death to lose a lead like that even though they probably realize in their heads how unlikely it is.

Even though a loss is a loss, it is human nature to attach more “importance” to a loss in the 9th when you have a large lead.

Which is why managers are really resistant to bring in less than their best reliever with a 3 run lead in the 9th.  As I mentioned in a previous thread, many managers will sometimes bring in their closers in the 9th with a 4 run lead and that is NOT just because they need the work.