### Sunday, October 10, 2004

## More Inane Brad Lidge Usage Arguments

After my attempt at satire on Thursday failed to garner any positive feedback, I'll go back to dry, boring analysis today. I've thrown around some ideas about Jose Lima's magnificent performance last night, but I'm too terrified about jinxing the Dodgers to seriously analyze that game. How about some more Phil Garner talk instead?

In the seventh inning with a 5-5 tie, Phil Garner put Mike Qualls into the game. Garner went with a double switch since Craig Biggio had been retired to end the sixth, inserting Jason Lane at left field and into the lineup at #9, putting the pitcher's spot at #1. In the eighth, after Dan Miceli walked Johnny Estrada, Garner put Brad Lidge into the game with one out and the #6 spot due up in the bottom of the eighth. I don't know what Joe Morgan said at the time since someone was vacuuming, but he later said he'd argued that Garner should have used a double switch here. Lidge retired both batters he faced and sure enough Lidge was pulled for a pinch hitter in the bottom of the inning with two outs and runners at the corners. The pinch-hitter, Orlando Palmeiro, grounded out on a close play to the right side. In the ninth, Russ Springer struck out Eddie Perez and Charles Thomas before yielding a single and a stolen base to Rafael Furcal and the game-winning hit by J.D. Drew. The Astros came up short when Jeff Kent grounded into a double play to end the ninth.

Throughout the final portion of the game, Morgan continually criticized Garner's initial double switch and then his decision not to double switch when Lidge entered the game. I wanted to throw something at the TV. The original double switch made a good deal of sense since Craig Biggio has been a pretty terrible left fielder this season whil Jason Lane is a substantially better. In fact, Lane is arguably the better offensive player at this point, with a .275 EqA vs. Biggio's .270. Now, Lane only had 156 plate appearances so that's not overwhelming evidence, but this move looks pretty good since the odds that the Astros would need a key at bat from the #9 hitter before the 9th were better than the odds that they'd need one from the #1 hitter.

The hypothetical second double switch is quite obviously a bad idea. With 6-7-8 due up, the odds of needing a big at bat from the #1 hitter would be pretty low, so there's not a huge amount of marginal value to be added. It obviously makes no sense to make a switch that puts Lidge up to hit earlier, so the 6-9 hitters can't be removed. That means that Morgan is arguing that Garner should have taken out one from Carlos Beltran, Jeff Bagwell, Lance Berkman, and Jeff Kent. Uh, 'fraid not, Joe. Not only are those the Astros best hitters by a long shot, but they're also all good fielders. It makes no sense for the Astros to play for the possibility of a marginally better eighth inning at the expense of a big ninth inning when their best hitters will be due up.

There are decisions Garner made that are bit more ripe for questioning. Why Dan Miceli was brought in in the seventh is a question I don't know how to answer; had Qualls finished the seventh and Lidge started the eighth, Miceli could have pitched the ninth and arguably is a better person for the job than Springer. But that's pretty marginal, and I think Garner has a better idea about the differences among his pitchers than I do. The other issue is that Garner perhaps should have let Lidge hit for himself in the eighth. This would no doubt be extraordinarily controversial, and Garner would be in deep trouble if he'd done it and the Astros lost. Let's use win expectancy to examine the possibilities.

The Astros had runners on first and third and two out in the eighth. If the inning ends without a run scoring, the Astros win expectancy goes to .506. If one run scores before the inning ends, it goes up to .857. If they score two runs, it goes to .938. Let's assert, for the sake of argument, that Lidge could be expected to retire all three batters in the ninth and perhaps be effective in a possible tenth inning. Had Lidge hit for himself, been retired, and breezed through the ninth the Astros win expectancy entering the bottom of the ninth would be .649. Had the Astros failed to then score in the ninth and Lidge shut down Atlanta in the tenth, the Astros' would have a .662 win expectancy.

With Eddie Perez, Charles Thomas, and Rafael Furcal coming up, let's say that Russ Springer has a 50% chance of not allowing a run since his fair run average using the new Baseball Prospectus reliever reports is a whopping 5.49 (= .61 R/IP) and he's done that against competition that closely resembles the collective line of that trio. We'll give him a 40% chance of giving up one run, 6% chance of giving up 2, and 3% chance of yielding 3, which average out to .61 R/IP. I know that's fairly arbitrary, but seems reasonably close. So if the Astros fail to score in the eighth, they have a 50% chance of entering the bottom of the ninth with the aforementioned .649 win expectancy. If we use my guesses at Springer's expected runs allowed to weight the Astros' win expectancy, we get a .398 win expectancy. That's a little low, however, due to the Astros having their best hitters up. Those hitters average a .303 EqA, so we could roughly expect them to do 1.166 times better than average using the crude (EqA/lgEqA) method, raising their win expectancy to .464. We can bring that down a bit to, say, .440 to adjust for the fact that John Smoltz is better than the average 9th inning reliever.

So how much value does pinch-hitting for Lidge add? Let's readjust the win expectancy of leaving Lidge in the game to account for the possibility he might give up a run, using his 1.34 Fair Run Average as our guide. If Lidge enters the ninth after failing to plate a runner in the eighth, the Astros can expect to win .588 times. If Lidge manages to score the runner from third, that becomes .946 or .952 if we adjust for the Astros' batters in the ninth. If we assign a 5% chance to Lidge scoring the runner, the weighted win expectancy for Lidge becomes .606.

So how much value does Orlando Palmeiro add? Well, Palmeiro's not a great hitter. His line this year is .241/.344/.346, which seems about right for a 35-year-old with a career .277/.356/.348 line. Since he's facing Smoltz, let's say Palmeiro's got a 22% chance of driving in the runner from third and a 10% chance of drawing a walk, with Carlos Beltran then having a 40% chance of driving in the run, so we'll assign Palmeiro a 16% chance of causing exactly one run to score and a 10% chance of causing 2 runs to eventually score in the inning. I'm fairly certain Springer would still pitch the ninth with the Astros ahead; the other option would have been Dan Wheeler, whose numbers with the Astros are better but who had a pretty poor season with the Mets, so let's use Springer's numbers again. With a one run lead, the Astros can expect .772 wins with Springer pitching. With a two run lead, that's .944. If we weight those based on Palmeiro's expected contribution and brng back that .440 from above for if Palmeiro fails to score the runner, our final win expectancy once Garner brings Palmeiro in becomes .544.

All in all, I would argue that the Astros had a slightly better chance of winning had Garner let Lidge hit for himself. Obviously, that would have been the most controversial decision of Garner's managerial career had he made it, and had it not paid off he perhaps could have lost his job despite engineering (or at least overseeing) an astounding turnaround after being hired. But if Garner wanted to give his team the best chance of winning, there's a good chance he would have been better off letting Lidge embarass himself at the plate because Lidge is just that good on the mound and because Palmeiro is just that mediocre.

In the seventh inning with a 5-5 tie, Phil Garner put Mike Qualls into the game. Garner went with a double switch since Craig Biggio had been retired to end the sixth, inserting Jason Lane at left field and into the lineup at #9, putting the pitcher's spot at #1. In the eighth, after Dan Miceli walked Johnny Estrada, Garner put Brad Lidge into the game with one out and the #6 spot due up in the bottom of the eighth. I don't know what Joe Morgan said at the time since someone was vacuuming, but he later said he'd argued that Garner should have used a double switch here. Lidge retired both batters he faced and sure enough Lidge was pulled for a pinch hitter in the bottom of the inning with two outs and runners at the corners. The pinch-hitter, Orlando Palmeiro, grounded out on a close play to the right side. In the ninth, Russ Springer struck out Eddie Perez and Charles Thomas before yielding a single and a stolen base to Rafael Furcal and the game-winning hit by J.D. Drew. The Astros came up short when Jeff Kent grounded into a double play to end the ninth.

Throughout the final portion of the game, Morgan continually criticized Garner's initial double switch and then his decision not to double switch when Lidge entered the game. I wanted to throw something at the TV. The original double switch made a good deal of sense since Craig Biggio has been a pretty terrible left fielder this season whil Jason Lane is a substantially better. In fact, Lane is arguably the better offensive player at this point, with a .275 EqA vs. Biggio's .270. Now, Lane only had 156 plate appearances so that's not overwhelming evidence, but this move looks pretty good since the odds that the Astros would need a key at bat from the #9 hitter before the 9th were better than the odds that they'd need one from the #1 hitter.

The hypothetical second double switch is quite obviously a bad idea. With 6-7-8 due up, the odds of needing a big at bat from the #1 hitter would be pretty low, so there's not a huge amount of marginal value to be added. It obviously makes no sense to make a switch that puts Lidge up to hit earlier, so the 6-9 hitters can't be removed. That means that Morgan is arguing that Garner should have taken out one from Carlos Beltran, Jeff Bagwell, Lance Berkman, and Jeff Kent. Uh, 'fraid not, Joe. Not only are those the Astros best hitters by a long shot, but they're also all good fielders. It makes no sense for the Astros to play for the possibility of a marginally better eighth inning at the expense of a big ninth inning when their best hitters will be due up.

There are decisions Garner made that are bit more ripe for questioning. Why Dan Miceli was brought in in the seventh is a question I don't know how to answer; had Qualls finished the seventh and Lidge started the eighth, Miceli could have pitched the ninth and arguably is a better person for the job than Springer. But that's pretty marginal, and I think Garner has a better idea about the differences among his pitchers than I do. The other issue is that Garner perhaps should have let Lidge hit for himself in the eighth. This would no doubt be extraordinarily controversial, and Garner would be in deep trouble if he'd done it and the Astros lost. Let's use win expectancy to examine the possibilities.

The Astros had runners on first and third and two out in the eighth. If the inning ends without a run scoring, the Astros win expectancy goes to .506. If one run scores before the inning ends, it goes up to .857. If they score two runs, it goes to .938. Let's assert, for the sake of argument, that Lidge could be expected to retire all three batters in the ninth and perhaps be effective in a possible tenth inning. Had Lidge hit for himself, been retired, and breezed through the ninth the Astros win expectancy entering the bottom of the ninth would be .649. Had the Astros failed to then score in the ninth and Lidge shut down Atlanta in the tenth, the Astros' would have a .662 win expectancy.

With Eddie Perez, Charles Thomas, and Rafael Furcal coming up, let's say that Russ Springer has a 50% chance of not allowing a run since his fair run average using the new Baseball Prospectus reliever reports is a whopping 5.49 (= .61 R/IP) and he's done that against competition that closely resembles the collective line of that trio. We'll give him a 40% chance of giving up one run, 6% chance of giving up 2, and 3% chance of yielding 3, which average out to .61 R/IP. I know that's fairly arbitrary, but seems reasonably close. So if the Astros fail to score in the eighth, they have a 50% chance of entering the bottom of the ninth with the aforementioned .649 win expectancy. If we use my guesses at Springer's expected runs allowed to weight the Astros' win expectancy, we get a .398 win expectancy. That's a little low, however, due to the Astros having their best hitters up. Those hitters average a .303 EqA, so we could roughly expect them to do 1.166 times better than average using the crude (EqA/lgEqA) method, raising their win expectancy to .464. We can bring that down a bit to, say, .440 to adjust for the fact that John Smoltz is better than the average 9th inning reliever.

So how much value does pinch-hitting for Lidge add? Let's readjust the win expectancy of leaving Lidge in the game to account for the possibility he might give up a run, using his 1.34 Fair Run Average as our guide. If Lidge enters the ninth after failing to plate a runner in the eighth, the Astros can expect to win .588 times. If Lidge manages to score the runner from third, that becomes .946 or .952 if we adjust for the Astros' batters in the ninth. If we assign a 5% chance to Lidge scoring the runner, the weighted win expectancy for Lidge becomes .606.

So how much value does Orlando Palmeiro add? Well, Palmeiro's not a great hitter. His line this year is .241/.344/.346, which seems about right for a 35-year-old with a career .277/.356/.348 line. Since he's facing Smoltz, let's say Palmeiro's got a 22% chance of driving in the runner from third and a 10% chance of drawing a walk, with Carlos Beltran then having a 40% chance of driving in the run, so we'll assign Palmeiro a 16% chance of causing exactly one run to score and a 10% chance of causing 2 runs to eventually score in the inning. I'm fairly certain Springer would still pitch the ninth with the Astros ahead; the other option would have been Dan Wheeler, whose numbers with the Astros are better but who had a pretty poor season with the Mets, so let's use Springer's numbers again. With a one run lead, the Astros can expect .772 wins with Springer pitching. With a two run lead, that's .944. If we weight those based on Palmeiro's expected contribution and brng back that .440 from above for if Palmeiro fails to score the runner, our final win expectancy once Garner brings Palmeiro in becomes .544.

All in all, I would argue that the Astros had a slightly better chance of winning had Garner let Lidge hit for himself. Obviously, that would have been the most controversial decision of Garner's managerial career had he made it, and had it not paid off he perhaps could have lost his job despite engineering (or at least overseeing) an astounding turnaround after being hired. But if Garner wanted to give his team the best chance of winning, there's a good chance he would have been better off letting Lidge embarass himself at the plate because Lidge is just that good on the mound and because Palmeiro is just that mediocre.