DMB-Blog — ACLS

Revisiting Game 7 of the 2003 ALCS

A few months ago, we were invited by the New England Sports Network (NESN) to help with the first installment of a new TV series called What IF that explores some of the high-profile turning points in recent sports history.

Because NESN is based in Boston and carries all of the Red Sox games, it should come as no surprise that they chose to launch this series with an examination of Grady Little's decision to leave Pedro Martinez in the game in the 8th inning of game seven of the 2003 ALCS.

NESN asked Diamond Mind to create a scenario in which Little signaled for Alan Embree instead of leaving Pedro in the game with one out, a runner on first, and the score 5-3 for Boston. Our task was to simulate the end of that game a meaningful number of times, identify the most common outcome, choose one of the simulation runs that matched the most common outcome, and provide NESN with a pitch-by-pitch and play-by-play account of that game.

NESN took those results and constructed a realistic game broadcast that aired on July 20, 2006. They showed the real game until the point where Grady went to the mound, then presented the results from our simulations by splicing together matching footage from previous games between these teams. The Red Sox TV announcers provided "live" commentary for the new ending, by which I mean they had no advance knowledge of the outcome while describing the action being displayed on their monitors.

The heat of the moment

Before we talk about the simulations, let's go back in time to October, 2003.

Like most Red Sox fans, I was surprised when Pedro started the 8th inning. In the 7th, Martinez had given up a run on a solo homer by Jason Giambi and struggled against the bottom of the Yankees batting order before striking out Alfonso Soriano to escape the inning.

When Pedro walked off the mound at the end of the 7th, he behaved very much like a man who had done his job and was ready to hand the ball to the bullpen. He gestured to the sky as he does when leaving a game, accepted handshakes and hugs from teammates in the dugout, and, according to several accounts, was heading for the clubhouse when Grady surprised everyone by telling him he might need another inning.

Three batters into the 8th, one run was in, a runner was on base, and the potential tying run was coming to the plate with one out.

When Grady went to the mound, I was thinking, "It's about time."

When Grady turned around and headed back to the dugout, I was stunned, and I was angry. I slammed my fist on the coffee table, jumped to my feet, and screamed at the TV.

I turned to my wife and said, "I can't believe he's leaving him in. It's obvious that Pedro's done. You watch. This is going to end badly."

When a friend called on the phone a moment later, I said, "Grady just got himself fired."

As a Red Sox fan who had lived through 1967 and 1972 and 1975 and 1978 and 1986, I was convinced this was the beginning of the end.

And it was. I watched Hideki Matsui lash a double down the right-field line to put the tying runs in scoring position. I watched Jorge Posada hit a bloop double that tied the score. And I watched Aaron Boone hit the series-winning homer in the bottom of the 11th inning.

The next morning, I was so depressed that I told my Diamond Mind colleagues that I never wanted to watch another baseball game as long as I lived. I even went so far as to replace the "Diamond Mind Baseball" sign on our front door with a hastily-designed logo for a new football simulation company. I simply couldn't imagine spending any more of my life doing baseball work, not after being so close to beating the Yankees and watching it slip away like that.

I was every bit as devastated as anyone who had followed the Red Sox for a long time. And that's why I was surprised when I looked back at that game a few weeks ago and came away with a somewhat less one-sided view of those events.

Setting up the simulations

If you're already familiar with the Diamond Mind Baseball simulation software, you can skim through this section. For everyone else, I'll take a moment to describe the simulation software and the process of setting up a scenario like this one.

Player performance in our simulations is determined by Diamond Mind's player ratings, which are based on detailed analysis of real-life pitch-by-pitch and play-by-play data and cover all important baseball skills, including hitting, pitching, defensive range and error rates, running, throwing, bunting, and so on.

All hitters and pitchers are rated for such things as their control of the strike zone, swing frequencies, and the results of those swings, including swinging strikes, foul balls, and balls in play.

When the simulation software plays a game, the dynamics of the batter-pitcher confrontation change as the pitch sequence moves through various counts. Batters have the edge when they're ahead in the count and are at a disadvantage when they are behind in the count.

When a pitch results in a batted ball, the outcome of the play is determined by several factors. Hitters are rated for their tendency to produce various types of batted balls, their tendency to pull the ball or go the other way, and their ability to produce different types of hits and outs. Pitchers have similar ratings for their ability to prevent hitters from doing all of these things. Depending on the situation and the nature of the batted ball, park effects, weather, the quality of the defense, fatigue and tactical choices (such as putting runners in motion or pitching around a hitter) can also have an impact on the outcome.

For this situation, we set up the two teams with their 25-man rosters for that series, entered the pitch-count data for game seven and several days leading up to that game so the simulation could realistically take fatigue into account when generating game events, set up the situation as it was when Grady went to the mound, and saved this scenario so it could be played out many times.

So ... what happened?

We simulated the ending of this game 100 times. In each case, Alan Embree was brought into the game to face Hideki Matsui with a 5-3 lead and a runner on first.

Boston held on to win 82% of the time. This should not come as a big surprise. Historically, teams holding a two-run lead in the bottom of the 8th with a runner on first and one out go on to win the game 81% of the time. Every game is different, of course, so that 81% figure doesn't necessarily indicate the most likely result of a particular game involving specific teams with their own strengths and weaknesses, but it's a strong indication that the team with the lead will go on to win roughly four out of every five times when this situation arises.

Fifteen of the simulated games went to extra innings, with New York winning ten and Boston winning five of those games.

The most common final score was 5-3 for Boston, which occurred 41 times in 100 games. The next most common score was 5-4 Boston, which occurred 13 times. The third-most common outcome -- a 6-5 win for the Yankees that matched the final score of the real-life game -- occurred 10 times.

Aaron Boone did not homer in any of these 100 simulated games, though he did enter the game in a number of them.

OK, what does this mean?

This does NOT mean that Grady Little is solely responsible for losing the game for the Red Sox.

Yes, it's true that Grady chose to leave Pedro in the real-life game and that the Sox lost that game. But that doesn't mean that a Boston loss was the inevitable result of that decision.

It's also true that our simulations show that Boston would most likely have won the game had Little chosen to go to Embree. But that doesn't mean a Boston win was a certainty. In our simulations, the Yankees still came back to win 18% of the time.

In other words, there's more to it than just "Grady lost the game."

Top five reasons we can't blame Grady

This headline is a direct rip-off of the title of the ESPN series, but it fits, so I'll use it.

After I had finished researching the teams, watching the video of the 8th inning several times, and thinking about the decisions Grady was facing at the time, I began to think there might be a little more to this story.

As the designer of a computer baseball game, I'm occasionally asked by a customer to look into a decision made by the game's computer manager. When that happens, I try to put myself in the shoes of the manager and ask what I would have done in that situation. To answer that question, I look at the game situation, identify the best options available, and evaluate those options by studying the stats and ratings of the players involved.

After setting up the 8th inning scenario for this game, I found myself staring at the screen and going through that type of analysis, asking myself whether one could make a case for leaving Pedro in the game.

And you know what? I was able to come up with some arguments in Grady's defense. They may or may not be enough to win a favorable verdict in a trial before the court of public opinion, but they were enough to justify mounting a defense rather than simply pleading guilty and moving on.

So here are my top five reasons why you can't blame Grady for this loss. (I'll subject these arguments to cross-examination later.)

1. The quality of Pedro's pitches to Matsui and Posada.

When I watched the game live in 2003, I was virtually certain that Pedro would fail to escape that inning with the lead. I knew that he had struggled after 105 pitches that season, and he was already at 115 when Grady went to the mound. I knew he was under pressure in the previous inning and had already given up two hits in the 8th. And, I'm sure I felt that way because ... well, these things always go against the Red Sox. Always have, always will. Or so it seemed at the time.

But then I watched the video of that inning for the first time in 30 months. And even though Pedro was running out of gas, he appeared to be throwing well enough to get the job done.

Against Matsui, Pedro nipped the outside corner with an excellent fastball that moved down and away for strike one, then got ahead 0-2 on a curve ball on the inside corner. The 0-2 pitch was not a great pitch. It was a 93-mph fastball that started on the inside corner before tailing and catching too much of the plate, and Matsui pulled it down the line for a double. It was fair by no more than a foot, but it was a double nonetheless.

Against Posada, Pedro fell behind on an inside fastball, then evened the count on a curve on the outside corner. His next pitch was a curve that started outside and stayed outside, but Pedro evened things up again when Posada swung and missed at another curve on the outside corner. The 2-2 pitch, the one Posada blooped into shallow center, was a 95-mph fastball on the inside part of the plate.

Pedro threw eight pitches to these two hitters. Four of them were, in my view, good-to-great pitches, one tailing fastball on the outside corner and three curves that nipped the edge of the plate. The two worst pitches were far enough off the plate to be unhittable. The other two, the ones that were put in play, are debatable.

Matsui hit a belt-high 93-mph fastball on the inner third, and he hit it very hard. I'm sure Pedro wanted that pitch to be lower and further inside, so it cannot be considered a great pitch or even a very good one. But every baseball fan has seen plenty of 93-mph fastballs that produce swings-and-misses, fouls back to the screen, or weak contact. A solid extra-base hit was not the only possible outcome of that pitch, or even the most likely one.

It's hard to be sure from the video, but it looked like the pitch Posada hit was better than the one to Matsui. It appeared to catch less of the plate, and it was clocked at 95. Much of the time, that pitch is fouled off, breaks the bat, or produces a weak popup on the infield. This time, it found no-man's land in shallow center field.

Pedro threw a series of mostly good-to-excellent pitches. Not every 93+ fastball in the strike zone is going to be crushed or blooped for a hit. And we cannot be sure that Embree would have produced a sequence of pitches that was measurably better. So maybe we ought to cut Grady some slack and tip our hats to Matsui and Posada instead of pointing the finger at the manager.

2. The matchups favored Pedro

If you look at the matchups, it's far from clear that going straight to Embree was the only reasonable option.

If Pedro stays in, he faces the left-handed Matsui and the switch-hitting Posada, who would be batting from the left side. Against Pedro in 2003, left-handed batters hit .238 with an on-base percentage of .292 and a slugging percentage of .364. In other words, the league's lefty hitters collectively produced at the level of Neifi Perez and Pokey Reese when facing Pedro that year.

Embree is an oddity in that he's a left-handed pitcher who often has more success against right-handed batters. Lefties hit Embree a little harder than they hit Pedro that year: .263 average, .277 on-base, and .424 slugging. Against righties, however, Embree yielded only a .221 average and a .317 slugging percentage, though he walked enough guys to push their on-base average up to .314.

Matsui was a better hitter against right-handed pitching. Against lefties, he batted .287 with a .335 on-base and .379 slugging. Against righties, he also batted .287, but with a higher on-base (.360) and slugging (.460).

Posada was very good from both sides of the plate but better against lefties. Versus righties, he batted .276 with an on-base percentage of .405 and a slugging percentage of .510. Against lefties, his on-base was about the same but his batting average rose to .295 and his slugging to .541.

Based solely on their regular-season numbers, it's not obvious that Embree is a much better choice against Matsui. Pedro was better than Embree against lefty hitters, but Matsui was better against righty pitchers. Combine these two effects and you get an edge for Embree, but not a big one.

Against Posada, it's a similar story. From the perspective of the pitchers, you'd rather have Embree face him. Posada would bat left against Pedro and right against Embree, and Pedro was a hair less effective against lefty hitters than Embree was against righty hitters. On the other hand, Posada was better against lefty pitchers that year. Combine these two effects, and you have a small edge for Pedro.

What about the batter-pitcher matchups? The numbers clearly favor Pedro. In 2003, including the ALCS, Matsui was 2 for 16 against Pedro. From 2001 to 2003, Posada batted only .243 with an astounding 22 strikeouts in 37 atbats against Pedro. On the other hand, against Embree, Matsui was 3 for 9 in 2003 and Posada 3 for 11 with a homer from 2001 to 2003.

3. A tired Pedro is still better than a rested Embree

Pedro is one of the greatest pitchers of all time. He may not have been at his best in that situation, given the effort he had exerted to that point in the game, but even if he's only at 90% of his normal, dominant level, that still might be better than what Embree can offer.

4. Cleveland, 1999

Baseball history is full of examples of an ace being called upon to go beyond his normal limits and coming through for his team. In the 1968 World Series, Mickey Lolich tossed three complete-game victories in eight days, the last in game seven on two days rest. In 1990, Jack Morris pitched 10 shutout innings in game seven, on three days rest, no less. In 2001, Curt Schilling and Randy Johnson logged a ton of innings in the post-season, with Johnson coming out of the bullpen in game seven after starting game six the day before.

And who can forget Pedro's performance in the 1999 division series? Battling a muscle strain incurred in game one of the series, Pedro surprised everyone by volunteering to pitch in relief in game five. All he did was find a way to no-hit the powerful Indians lineup for six innings and save the game and the series for the Red Sox.

If you're Grady Little, and you know what this guy has done in the past, isn't it entirely reasonable to think that Pedro would find a way to retire Matsui and Posada or at least get past them with minimal damage so you can bring in a fresh Embree to face Jason Giambi?

5. We don't know what we don't know

As much fun as it is to second-guess managerial decisions, it's important to recognize that we don't always have all of the facts. Sometimes the manager knows something we don't, something that has a legitimate and major bearing on the decision at hand, something that would also change our minds if only we had that information, too.

Now, you might ask, "If that was true, if a manager felt he could save his skin by disclosing this information after the game, don't you think he would do that?"

Not necessarily. If it meant throwing a player under the bus, maybe he would take the arrow for his player because he feels it's the right thing to do. Or perhaps he wanted to demonstrate to future employers and players that he can be trusted in that way.

Cross-examination

OK, we've made a case for the defense, but do these arguments really hold up?

Pedro threw some very-good-to-excellent pitches to Matsui and Posada, any one of which could have produced a double play ball that would have changed the outcome. But he had already worked very hard in this game and didn't appear to be at his best -- his velocity was still up there and his curve ball was still biting, but his location wasn't as precise as usual.

And it seemed clear to most observers that Pedro thought he was out of the game after the seventh. It's tough to ask someone to gear up for a game like that, perform at a high level for two hours, reach back for a little extra to work out of a jam in the seventh, come back down from that high in the belief that his work is done, and regain that level of focus and intensity ten minutes later.

Rebuttal: I've heard this argument a few times, but it looked to me that Pedro was as intense and focused as ever in the 8th inning, so even if was asking a lot of him, it's hard to argue that Grady was wrong about Pedro's ability to perform.

It's true that Pedro had dominated Matsui and Posada to that point in their careers, but what little success those hitters had against Martinez had come recently, in this ALCS. Perhaps they were starting to figure him out, making the matchup data less indicative of what would happen next.

Was a tired Pedro still a better pitcher than a more-rested Embree? That's debatable. Pedro was at 115 pitches when Grady came to the mound, and while he had shown the ability to go into the 120s on certain occasions, he was noticeably less effective after 105 pitches.

And it's not as if that fact wasn't on everyone's minds at the time. Two weeks earlier, in the first game of the Oakland series, Pedro was at 97 pitches when he started the 7th inning, and he proceeded to load the bases on a hit and two walks. He managed to get out of that jam without allowing a run, but he needed 33 more pitches to do it. Five days later, in game five of the Oakland series, Pedro's pitch count was at 96 going into the 8th, and he promptly gave up a run on two hits before being lifted.

Embree wasn't fully rested, either, having pitched 1-2/3 innings the night before and 2/3 the night before that. But he needed only 25 pitches to get through those two outings, so he should have been able to face another two or three hitters in this game. And Embree had been pitching very well in the playoffs. Edge to Embree, in my view.

In addition, there's the batter-learning argument. Dave Smith of Retrosheet has demonstrated if a starting pitcher remains in the game long enough to face the same hitters more than once, the hitters put up better stats in their second atbat than in their first, and they continue to improve when they face the same pitcher later in the game. Embree would have negated the batter-learning effect, while sticking with Pedro played into it.

Finally, it might be a stretch to equate Pedro with Lolich, Morris, Schilling, and Johnson. Just because others have done this before, and just because Pedro himself was able to do something special at his peak in 1999, it doesn't mean he was physically capable of doing it in 2003. (There's no question about his mental and emotional ability to handle that situation.)

Reaching a verdict

For me, the most interesting part of this exercise is that I'm more open to the idea that leaving Pedro in the game was a viable option. Without a sense of impending doom clouding my vision, I can look back at that inning and see that Pedro threw an awful lot of good pitches. I can point to some objective evidence to suggest that Pedro had a very good chance to retire either or both of the hitters he faced.

And I think it's fair to say that Pedro was a little unlucky. If either Matsui or Posada puts one of his better pitches in play instead of swinging through them or taking them for strikes, he probably gets out of the inning. If Matsui's liner is a foot foul instead of a foot fair, he's still in a big 0-2 hole. If Posada's bloop is caught, the runners probably stay where they are, allowing Grady to bring in Embree to face Giambi, who batted only .192 against lefties that year.

But even if there was a good chance that things would work out for the Sox even with Pedro in the game, that doesn't mean Grady made the right call. The right call is the one that gives your team the best chance to win, not just a good chance to win.

And I believe Embree would have given them a better chance to win the game, though I don't think the difference is as large as some would believe. Just because the real-life game turned out to be a loss, the probability of winning didn't go from 82% with Embree to 0% with Pedro. Most likely, their winning chances were still in the 60s or 70s with Pedro pitching.

Note: We ran another 20 simulations with Pedro staying in the game, and Boston won 15 of those games. That suggests their winning chances were still as high as 75% when following Grady's course of action, but I'd caution against reading too much into that number. The results of 20 simulations are less reliable than the results of 100 simulations.

And there's one more thing. Visiting teams win about 90% of the time when they hold a three-run lead entering the bottom of the 8th, so Grady's decision to have Pedro start the inning may have reduced the team's winning chances by a significant amount even before this What IF scenario came about.

So, if I had been on the hot seat, and if I didn't have access to important information that would shed a different light on this question, I would have gone to Embree. Even before that, I would have gone to the pen to start the inning.

Closure

In October of 2003, I was furious with Grady Little and depressed about the outcome of that game.

A few months ago, when NESN first approached me about this project, I didn't like the idea. I didn't see the point of bringing back the bad memories.

And I wasn't keen on subjecting Grady to another round of criticism. He seems like a good guy who has taken more than enough abuse from Red Sox Nation. Before that fateful night, he did plenty of good things in his tenure as Red Sox manager. Besides, I have a healthy respect for anyone who is willing to make gutsy calls in front of 50,000 screaming fans and millions more in the TV audience.

Today, I'm glad I took part in this project because it helped put to rest any lingering bad feelings I had about that game.

There were two great teams in the American League in 2003 and 2004. Two teams that played each other to a standstill in 38 grueling regular season games and two unbelievable playoff series. If either club had won both of those series, it would not have been a fair reflection of how evenly matched they really were. One ALCS win for each was about right.

Even if you don't buy any of the pro-Grady arguments in this article and you're still convinced that Grady cost the team a chance to win back-to-back titles, try not to forget about all those other times when fortune favored the Sox.

Remember how close the Sox were to being swept by Oakland in the 2003 ALDS? Or how close they came to getting swept by the Yankees in 2004? That year, their survival turned on little things like a screaming line drive off Matsui's bat that went right to Trot Nixon with two men on base, a Tony Clark ground-rule double that saved a run when it barely skipped into the stands, and a running-on-fumes Keith Foulke getting a game-saving strikeout on an 80-something fastball that was more hittable than the pitches Matsui and Posada hit off Pedro a year earlier.

Lest you take these simulation results and get too carried away about the possibility of back-to-back World Series wins, keep in mind that the What IF game cuts both ways. If the Red Sox had beaten the Yankees and the Marlins to win the World Series in 2003, we might be left without some of the greatest stories in baseball history.

Perhaps the Sox do not feel the need to trade for Curt Schilling or make some of the other moves that shaped their 2004 season. Maybe the Nomar trade never happens. Maybe Grady keeps his job. Maybe Johnny Damon doesn't show up at spring training with long hair and the whole Jesus/caveman thing never happens. For whatever reason, maybe the Sox don't make the post-season in 2004, or maybe they don't face the Yankees again.

Then there's no comeback from three games down. There's no bloody sock. Derek Lowe doesn't salvage a miserable season by getting the win in all three post-season clinchers. Dave Roberts doesn't steal the most important base in franchise history. A-Rod doesn't slap the ball out of Bronson Arroyo's hand. And a few chapters of the Big Papi legend remain unwritten.

In 2003, when Grady slapped Pedro on the shoulder and headed back to the dugout, it felt like the beginning of the end. Short term, it was. But it was really just the beginning of one of the greatest sagas in the history of the game, one that finally had a happy ending for Red Sox fans.