April 19, 2007
Caught Stealing to Lose the Game
"It was a good time to run and it happened to be a good pitch for the catcher to throw on…It's just part of the game. You want to get a guy in scoring position. Sometimes you're aggressive and it comes out against you. We're an aggressive team, and we like to play our game, within ourselves. It's not always going to work in your favor."
"The best thing one can do when it's raining is to let it rain."
From the red-hot Alex Rodriguez and Felix Hernandez and the not-so-hot starts of the Nationals and Alex Gordon, to the debut of Daisuke Matsuzaka, and despite everyone's new favorite topics of the weather and the schedule, I know all of our readers are glad to finally be basking in the simple fact of a season's start. Today we'll discuss a couple interesting tidbits from the first two weeks along with another view of baseball's version of natural selection.
Outs on the Bases
The season is only two weeks old, but if you've been a bit distracted by the weather, you may have missed a couple of interesting baserunning events that deserve a little comment.
First, as if the Cubs haven't had enough problems, on Saturday Alfonso Soriano was picked off second base by Bronson Arroyo. That marked the third time in this young season that Soriano has been picked off. Going back to 1980, those who have been picked off six or more times in a season are:
Year Player Team PO 1986 Vince Coleman SLN 9 1980 Rickey Henderson OAK 9 1992 Kenny Lofton CLE 9 1981 Rickey Henderson OAK 8 1987 Billy Hatcher HOU 7 1980 Joe Morgan HOU 7 1987 Harold Reynolds SEA 7 1984 Al Wiggins SDN 7 1983 Tim Raines MON 6 2002 Dave Roberts LAN 6 1982 Steve Sax LAN 6 1992 Luis Polonia CAL 6 1980 Bob Molinaro CHA 6 1980 Omar Moreno PIT 6 1984 Dion James MIL 6 1990 Rickey Henderson OAK 6 1985 Mike Davis OAK 6 1989 Vince Coleman SLN 6 1980 Dave Collins CIN 6 1986 John Cangelosi CHA 6
So, in less than a dozen games, Soriano is already halfway to breaking into the top 20. I know we'll all be watching to see how much progress he makes on this, either in cutting down on these sort of mishaps, or making a run at the title.
Secondly, as mentioned by John Perrotto on Sunday, on April 10th Erick Aybar of the Angels was thrown out stealing with two outs in the top of the ninth to end a 7-6 loss to the Indians. That was the second time caught stealing this season for Aybar, who was also gunned down in the bottom of the ninth on April 5th in a 4-3 loss to the A's.
Thinking this must be very rare event, I took a quick look at the Retrosheet play-by-play data from 1970 through 2006. Surprisingly, getting thrown out to end a game occurred 84 times during that span (excluding 1999) including once in 2006 when on May 17th pinch-runner Willie Harris was thrown out trying to steal second with Trot Nixon at the plate in a 4-3 loss to the Orioles. Aybar now joins Albert Hall, Derrell Thomas, Glen Barker, Rex Hudler, Garry Maddox, Larry Lintz, Luis Polonia, and Willie Harris as the only players to be thrown out twice to end games in a single season since 1970. At least Jose Cardenal (1974), Rolando Roomes (1988), Maddox (1978), and Daryl Boston (1989) went out with dignity by making the last out on an attempted steal of home. Nor for that matter does it compare to Babe Ruth's famous caught stealing going for second base to end the 1926 World Series with his team down 3-2 and Bob Meusel at the plate.
The question all of this raises, however, is whether either of Aybar's attempted steals made sense. In the game on April 5th, using the 2006 Angels run environment, Win Expectancy (WX) would tell us the Angels had a 9.5 percent chance of winning after Aybar singled in his pinch-hitting appearance with two outs in the ninth. Had he successfully stolen second, he would have raised the Angels' WX to 14.6 percent. On its face, then, the break-even percentage (that percentage above which it makes strategic sense to attempt a steal) would be 65.1 percent. Given that Reggie Willits was at the plate facing A's closer Huston Street, the break-even was probably a little lower than that, since Street was not likely to give up an extra-base hit to the likes of Willits (although he certainly may have drawn a walk, moving the tying run into scoring position).
The question then becomes whether or not Aybar had a reasonable expectation of a success rate of over 65 percent in this situation. Going back to 2005, Jason Kendall has caught 25 percent of attempted base stealers, although he was better in 2006 (topping 30 percent) than in 2005. In running on Street, opposing baserunners have stolen 11 bases in 13 attempts in his career. Considering just those two factors, one might think it was indeed reasonable or Aybar to run.
However, although Aybar is fast, he is not a good percentage basestealer getting caught 96 times in his minor league career against 179 successful steals--that's a rate of just 64.7 percent. Based on just that information, it would seem that at best this was marginal decision by manager Mike Scioscia. In fact, Scioscia had some additional information that informed his decision, as reported on MLB.com:
With that combination--the time we had for the pitcher [coming to the plate] and throwing time [for the catcher]--I thought we had a better than 75 percent chance of making it. Erick got a decent jump. It took a perfect throw to get him, and they got it. If it was a 50-50 proposition, obviously we're not going to do it.
Based on this added information, Scioscia had indeed calculated the odds beforehand, and found the risk acceptable. What he didn't know was that on the other bench A's manager Bob Geren was also thinking:
We were talking on the bench, saying, 'Do you think they'll take a shot right here?' And we all agreed, 'Yeah, they might.' That's what makes them good.
As reported on MLB.com, the bench then signaled Kendall to relay to Street to use a slide step and call for a fastball. Both were done, and Kendall fired a strike to second to nab Aybar. The result is an excellent example of the game within a game that baseball fans come out to the park to see.
In the April 10th game, Aybar came in to pinch run after pinch-hitter Casey Kotchman had singled with two outs. In this situation the Angels had a 7.7 percent chance of winning, and if Aybar had been successful it would have risen to 11.4 percent, putting the break-even rate at 67.5 percent. This time, Aybar was facing the battery of Joe Borowski and Kelly Shoppach. Opposing runners had been successful 30 of 37 and 25 of 37 times against the two respectively in their careers. Again, it would seem like the chances were pretty good as far as a decision to run. What tends to put this attempt in a worse light, however, is that Howie Kendrick, 4 for 4 on the day, was at the plate. In this instance Aybar was running on his own, and Scioscia again attributed the caught stealing to "a perfect throw" delivered by Shoppach.
But even though Aybar may not be the best candidate for stealing second in these situations, both examples do reveal the fact that although managers may be derided for their choice when it doesn't work, the break-even percentages are not necessarily higher than in other situations. The result is that in a larger number of cases than might be at first be guessed at, it can be the right play, as evidenced in part by the frequency at which it is attempted.
Selection Pressure Redux
Regular readers will recall that two weeks ago, after chronicling the decline of Angel Berroa's once-promising career, we looked more closely at the 96 position players since 1901 who endured three consecutive years of offensive decline during their prime years, yet nevertheless retained their starting jobs. As it turned out, Berroa was one of only four players whose first-year park-adjusted offensive production was actually below league average. The other three were Berroa's top comparable, Rafael Ramirez, Mets shortstop Bud Harrelson, and speedy center fielder Brian L. Hunter. In part because Harrelson and Ramirez were better defenders (Berroa made approximately 85 fewer plays than would have been expected in his four years as a starter), we concluded that it would be unlikely if Berroa was ever again given a starting role.
But that sad story aside, we used the decline of Berroa to make a larger point regarding selection pressure in the highly competitive world of major league baseball. The primary reason that Berroa and the other three players retained their jobs as long as they did despite below-average production was because in playing demanding defensive positions, there was less pressure on them to hit. This is analogous to a characteristic of an organism that is less important in its struggle to survive and reproduce, and is therefore less acted upon by natural selection. As a result, that feature can vary more freely without impacting the overall success of that organism's species. However, if the importance of that characteristic changes due to new conditions and actually becomes detrimental, then the pressure is on for the organism to adpat or face a diminished future.
Another way of illustrating this point is to take a second look at all 96 of the instances of decline posted on my blog, and aggregate them by position. The following graph shows the park- and league-adjusted OPS in year one of the declines by position, followed by a table that includes the averages of the subsequent three years.
Position Yr1 Yr2 Yr3 Yr4 1B 134.1 125.7 119.3 110.6 RF 130.5 122.6 117.5 111.1 LF 125.6 116.2 107.9 100.4 CF 121.3 117.1 110.5 103.3 3B 119.3 113.3 106.5 99.5 2B 118.5 111.5 104.9 100.6 SS 115.5 109.9 102.1 96.4 C 111.8 105.4 102.2 97.0
What you can see from both the graph and table is that the positions, when sorted in descending order by offensive production in year one, are arranged in approximately the opposite order of the "defensive spectrum" popularized by Bill James. (Nate Silver took a new approach to this particular subject in February.) By way of review, the idea is that defensive positions can be arranged on a spectrum from the least-demanding position to the most, or 1B -> LF -> RF -> CF -> 3B -> 2B -> SS, and that players generally move from right to left on this spectrum over the course of their careers. Shifts in the other direction are rare and seldom work. Catcher and designated hitter are special cases, and were therefore not included in the original formulation. With catchers, you could argue that play the most demanding defensive position, and are therefore situated at the far right, but they don't move to the left easily (excepting the rare counter-examples, like Craig Biggio, B.J. Surhoff, or Dale Murphy) and instead, they jump over the middle and usually land at first base or DH. While a few more have made the switch to a corner outfield position (Brian Downing, or Yogi Berra), it's rarely successful, with examples like Todd Hundley or Carlton Fisk to caution against that particular switch.
What we're seeing in the graph and table above is a corollary to the idea of the defensive spectrum. They testify to the fact that less-demanding positions put greater selection pressure on a player's offensive contribution, since only those players who started out at a higher offensive level were able to keep their jobs through three consecutive years of decline. Those who began at a lower plane didn't make our list of 96, as they were selected out of regular jobs before they could record subsequent declines. Another way of putting it is that the defensive "characteristic" is more highly selected for at the right end of the spectrum, and that as we move to the left, the offensive characteristic becomes paramount.
There is, however, one other interesting aspect of the graph to point out. In the modern game third base is viewed as a much less demanding defensive position than second base, so the offensive output of third baseman is expected to be correspondingly higher. Yet in the table you can see that second and third baseman are basically on the same slope, with each starting out about 19% over league average, and then declining to league average in year four. This is not what we would expect.
The reason why the two positions are so closely matched here is because our modern view of those two positions is just that--modern. For the first three to four decades of this century it was second base and not third that was considered the less demanding defensive position, and therefore the one more often populated by better offensive players, as shown in the following graph.
Just why this shift occurred has been the subject for much discussion, but you can imagine that that the prevalence of bunting and chop hitting in the first part of the century highlighted the importance of a good-fielding third baseman, and was likely a contributing factor. It's no surprise either that James himself weighed in on this question in an essay titled "Why Did the Defensive Spectrum Jump?" in his 2001 book Win Shares. In that article he noted the same trend I've shown in the graph above in the decades from 1900 through 1969 using Runs Created, and likewise shows the shift towards the current set-up occurring sometime in the 1930s.
James does not attribute the shift in the spectrum to bunting and chop hitting, but rather to the dual factors of an increasing number of double plays and decreasing number of errors. Since 1947 there have been more double plays in the majors each year (except 1963 and 1975) than errors. His reasoning was that as double plays became more common, managers were more willing to sacrifice offense in order to get double plays. At the same time, as errors became less frequent managers were willing to put lesser defensive players at third base, since the cost wasn't as high. Interestingly, he notes how this trend is detectable in the common language we use to describe the positions. In the early part of the century second baseman were referred to as "keystone men," whereas today they are more frequently referred to as "pivot men," emphasizing their ability to turn two.
Getting back to our data, the reason second and third basemen rank so closely is that heavy hitters Eddie Collins, Larry Doyle, Jimmy Williams, and Frankie Frisch--all of whom started playing before 1930--take the top five second base spots in terms of offensive production in year one, while Deadball Era third baseman Bill Coughlin, Eddie Foster, and Bobby Byrne find themselves at the bottom of the list for their position.