|
|
Thoughts from Diamond Mind
Welcome to the Diamond Mind weblog. For a while, we've been looking for the best way to publish small pieces of baseball commentary and research, items that may not warrant a full article (such as the ones we've been writing for ESPN.com since 1998) or items that would be outdated by the time our next email newsletter is due to be issued. A weblog seems perfectly suited to this purpose, and we hope you enjoy what you read here. To provide feedback, email us at blog@diamond-mind.com.
We won't be able to reply to those messages individually, but we'll read them all,
and we may respond by posting a reply in the weblog itself.
|
|
Blog Archives Weblog Home
|
Wednesday, August 20, 2003
More fun with bunting OK, now it's the top of the 8th and the A's have just pulled within one run of the Red Sox after a trio of singles. Nobody out, runners on first and second. Sound familiar? Once again, the manager -- Oakland's Ken Macha this time -- tries to sacrifice the runners over. Or, to be more accurate, he asks Ramon Hernandez to do just that. Hernandez can't get the job done, pulling the bat back to take strike two and then bunting the next pitch foul for a strikeout. So the A's gave up an out and left the runners in place. Then Scott Hatteberg lines a single to right to score the tying run before Bill Mueller throws away a grounder to let the go-ahead run cross the plate. Same game, two bunts with runners on first and second and nobody out. One bunt is perfect and the team comes away with nothing to show for it. The other bunt attempt is about as bad as can be, and both runners come around to score anyway. Go figure. To be fair, the story of this game isn't the bunt attempts. Oakland is ahead 7-6 mainly because Boston has left fourteen (!) runners on base through seven innings. posted by Tom at 10:09 PM The downside of the sac bunt In yesterday's post, I noted that managers often opt for the sacrifice bunt with a runner on first and second and nobody out, and that this is good strategy if you're playing for one run but not if you're going for two or more. A few minutes ago, the Red Sox got into this position with righty Mike Neu on the mound, lefty Johnny Damon at bat, and the Sox holding a 6-4 lead. Even though Damon had the platoon advantage, Grady Little called for the bunt, Damon executed it perfectly, and Little had two runners in scoring position with one out. It didn't work out this time. Bill Mueller rapped a ground ball up the middle that looked like a base hit, but Neu stuck out his foot and blocked it, in the style of a hockey goalie making a kick save, and the ball popped right up to Neu's glove. The runner on third was a sitting duck, and there were two out. Then Nomar Garciaparra popped out on the first pitch, and the A's were out of the inning with no damage. posted by Tom at 9:08 PM 2003 Tigers versus 1962 Mets In case you haven't already seen it, I want to let you know about a fun little fantasy series we played the other day. Scott Ridge and Jayson Stark of ESPN.com came up with the idea of pitting the 2003 Detroit Tigers and the 1962 New York Mets in a best-of-seven series. They called to see if we'd be interested in putting it together, and we were delighted to help out. As usual, Jayson did a terrific job writing up the series in an article that appeared on ESPN.com earlier today. ESPN is running Jayson's story along with the game results, boxscores, and game logs. You can also find composite batting and pitching totals and several other reports on our web site. We decided to play the games with Diamond Mind Baseball version 9 because we wanted to take advantage of two new features. The first is the play-by-play game log. The second is the ability to generate a league web site -- a collection of linked HTML pages, including DMB reports and boxscore/scoresheet/gamelog files -- with a single command. Version 9 is still under development, but the game engine and these two new features are ready for prime time, so we weren't worried about running into any major bugs. To the best of our knowledge, we didn't run into any minor bugs, either. For those of you who've been waiting for a version 9 update, our next email newsletter, which will be sent out in a few days, will have the latest news on features and dates. (If you'd like to receive the newsletter or review back issues, you can do that here.) posted by Tom at 5:04 PM Tuesday, August 19, 2003 Mailbag: sacrifice bunting Several readers wrote to ask about (or state an opinion about) bunts in situations other than the one with a runner on first and nobody out. In my post yesterday, I should have listed this as another loose end that needs to be tied up before publishing the article I mentioned. It makes a lot of sense to extend the multi-outcome work to include other tactical situations, and I am planning to do that. In the meantime, I can tell you that the traditional two-outcome analysis shows that bunting a runner from second to third with nobody out is a viable option if you're playing for exactly one run. You don't need the bunt singles and errors to justify this tactic. This is also true with runners at first and second and nobody out, but again, only if you're playing for exactly one run. My guess is that most managers use this tactic when they're playing for two runs, but the two-outcome model shows that your chances of scoring two or more runs decrease if you sacrifice. With singles and errors added to the model, it's quite possible that those chances would increase, so this is a situation worth studying with the multi-outcome model. posted by Tom at 2:08 PM Monday, August 18, 2003 Sacrifice bunting revisited In his article in yesterday's New York Times, Jack Curry surveys current attitudes toward the sacrifice bunt. In the article, a few players and managers are quoted as saying that the bunt has its place, that it puts pressure on the defense, and it's important that players be able to execute a sacrifice when the situation calls for it. The primary spokesman for the anti-bunt camp was GM J. P. Ricciardi, whose Toronto Blue Jays have laid down a grand total of 5 successful sacrifices this year, reflecting Ricciardi's beliefs that it doesn't make sense to play for one run and that most players aren't especially good at bunting even when you ask them to work at it. While he was working on this story, someone told Mr. Curry that I had made a presentation on this subject at the 2002 SABR convention in Boston. Unfortunately, he and I played phone tag for a couple of days and didn't connect until the story was all but done. But he did briefly mention my work in his article, so I'm bound to get some questions about it. I've been meaning to turn that presentation into an article for our web site, but I just haven't found the time. If the full-blown article was ready for publication, I'd run it now. But it's not, so this seems like the right time to write up a shorter version. I became interested in this subject because the baseball analysis community has been dumping on the sac bunt for a long, long time. For example, John Thorn and Pete Palmer made the following claims about the sacrifice bunt in their classic book The Hidden Game of Baseball: "The potential run value is always lower after a successful sacrifice."Neverthless, last year's AL batters laid down 499 successful sacrifice bunts, almost all of them by position players. Were they all playing for managers who need a crash course in strategy? Or is there something wrong with the analysis that led to the conclusion that the bunt is such a bad idea? Those who say that the sacrifice bunt is obsolete typically use the following argument: 1. By studying the play-by-play data for many seasons, we can see that an average of X runs is scored when you have a runner on first with nobody out and an average of Y runs is scored when you have a runner on second with one out. 2. Y is always less than X, so even if you successfully sacrifice a runner to second, you're giving up runs. The value of the base gained is less than the cost of the out. As you know, most sacrifice bunts are used for the purpose of scoring one or two runs, not setting up a big inning, so the average number of runs is a less appropriate measure than the probability of scoring at least one or two. Turns out, however, that if you base X and Y on those probabilities instead, Y is still less than X most of the time. But there are two potential problems with this line of thinking. First, the values of X and Y are computed by examining all innings, regardless of the park, the pitcher, the hitters who were due up, and any other tactics used by the manager later in those innings. In other words, they're averages, and averages don't always apply to specific game situations. Second, this type of analysis assumes that there are only two possible outcomes -- one type of success and one type of failure. What if there were several types of successes and several types of failures, and what if those other outcomes happen often enough to be meaningful? Would that make the bunt appear to be much more attractive? To shed some light on question number one, I developed a tool that I called lineup-dependent expected runs analysis. It's a bit of software that allows me to compute the expected number of runs and the probability of scoring at least N runs for any combination of era, park, pitcher, and batting order. With this tool, I can go looking for situations where the values for X and Y are different enough to make the bunt look like a viable strategy. Maybe it makes more sense in a pitching dominated era like the 1960s. Maybe it's better than swinging away against Pedro Martinez in his prime. Perhaps a really anemic offense should be bunting some of the time. I was surprised to find that this part of the project did NOT produce any evidence to support the frequency with which the sacrifice bunt is used in today's game. I was able to find a lot of scenarios in which X and Y moved closer together, but they didn't cross over unless the batter was a weak-hitting pitcher. The conclusion: even if you adjust for the quality of the pitcher and the quality of the hitters, the two-outcome model still says that today's managers are asking hitters to bunt far too often. The second part of the study proved to be much more interesting. The simple two-outcome model assumes that (a) a successful sacrifice moves up a runner (or runners) while the batter is thrown out at first, and (b) a failed sacrifice leaves the runners in place while giving up an out. But everyone knows that bunts can result in more than just two outcomes. The batter might reach on a single or an error. He might foul off one or two attempts and then swing away. He might bunt into a double play. We have a program that analyzes real-life play-by-play data to compute the average number of runs and the N-run probabilities in all baserunner situations. That's the sort of program that people use to do the two-outcome analysis, and I've used it that way on many occasions. In fact, for much of the last fifteen years, I've used those numbers to join the anti-bunt chorus. For this project, I extended that program to evaluate a few special cases. Looking only at innings where the first batter reached first base, I computed the average number of runs for all innings in which the offensive team: - had the second batter swing away - had the second batter attempt a bunt This is where things got interesting. In recent AL seasons, about 12% of all bunt attempts produced either a bunt single or a bunt error. In other words, a non-trivial number of successful bunts gained at least two bases WITHOUT giving up an out. These outcomes had very high run values and increased the probability of scoring at least one run by a meaningful amount. If bunt double plays were quite common, they could have nullified the gains from singles and errors. But they're not. They happened only about 1% of the time. (By the way, I used AL seasons so I could focus mainly on bunting by position players.) And foul bunts didn't have a major negative effect, either. Yes, the batter ended up behind in the count, and that affected his production, but not by enough to offset the added value of the singles and errors. The bottom line is that the sacrifice bunt looks better in the multi-outcome model than in the two-outcome model. In fact, it looks good enough to be a legitimate part of a manager's toolkit when a position player is at the plate and the team is playing for one run. Not all position players, mind you. I'm not suggesting that a great hitter -- or even an average hitter -- be asked to bunt a lot of the time in these situations. Just that the folks who've been saying that the bunt should NEVER be used are overstating the case. One of the reasons I've yet to finish my article is that there are some loose ends that need to be tied up, these two among them: - I'd like to apply the multi-outcome part of the analysis to a larger sample of seasons and see if it holds up. I looked at three recent AL seasons, and things might look different in other seasons. - the multi-outcome analysis is subject to the same concern I mentioned back at the beginning. It averages all innings in which the leadoff hitter reaches first. Perhaps there are combinations of park, pitcher, and lineup where the bunt is less attractive and others where it's even more attractive. If we could combine the lineup-dependent expected runs analysis with a multi-outcome model, we might be able to pinpoint those situations. Well, this was supposed to be the short version of the paper. I don't know how to make it any shorter without leaving out some important stuff, but I don't want to make it any longer, either. The full article, if and when it appears, will explain the methodology in detail, provide all the supporting numbers, and present a large number of examples. In the meantime, however, I'll tell you what I believe: 1. If you're playing for one run, the singles and errors help make the bunt a viable option even with a position player at the plate. That puts pressure on the defense because it has to defend more than one threat. The offense is well served by keeping that weapon in its arsenal, even if it doesn't deploy it very often, because it keeps the defense honest. 2. This doesn't mean managers should go bunt crazy. With a runner on first and nobody out, bunts are attempted by position players about 10% of the time, and I'm not suggesting that the rate should be higher. Perhaps it should be a little lower, but it shouldn't be zero, either. 3. Big-league managers are better at what they do than many of us give them credit for. I believe they've always factored the singles and errors into their thinking. 4. This doesn't mean big-league managers deserve a free ride. Some of the bunts they call simply cannot be justified. For example, I'm just amazed that a very good hitter like Roberto Alomar has 139 sacrifice bunts in his career. A small fraction of those bunts may have been justified, but a guy with a lifetime OBP of .373 can get the runner over in so many other ways that it doesn't make sense to give him up so often. posted by Tom at 2:06 PM Rolen for Polanco In the August 11 issue of Sports Illustrated, Tom Verducci lists his choices for the five biggest trade-deadline steals of the past ten years. Number four on that list was the 2002 deal that sent Scott Rolen and Doug Nickle to St. Louis, with the Phillies getting Placido Polanco, Bud Smith, and Mike Timlin. Verducci's comment on that deal: "The Phillies came out of a messy divorce from Rolen with a fraction of his value." I don't understand why this deal is on the list. Rolen didn't want to be in Philadelphia, so the Phillies had to get something in return before losing him to free agency. And, in Polanco, they got a lot in return. (Timlin's a good player, too, but he's in Boston now.) Don't get me wrong. I think Rolen is a terrific player, the best player in this deal, and trades often turn out to favor the team that acquired the best player. He's only 28 years old, a stellar defensive player at the hot corner, and a very good hitter (.387 OBP, .539 SPC in 2003). I like him a lot. But Polanco is a terrific defensive player who can play second, third and even short. With David Bell slumping, he's proven to be their answer at third base. And he's producing at the plate, too, to the tune of a .370 on-base percentage and a .473 slugging average. He's a year younger and is making more than $4m less than Rolen this year. Rolen has shown that he's the real deal, and it's possible that Polanco is just having a career year, so it's too early to tell which team will be better off in the end. But this could turn out to be a win-win deal. Even if it doesn't, it's way too early to tell, so it has no business being on a list that includes Derek Lowe and Jason Varitek for Heathcliff Slocumb. posted by Tom at 11:14 AM |
 |
