Manager Wins Above Expectancy

If a manager was able to manage close games better than “average” over a long period of time, he would rank quite well in this scale. If a manager won a lot of blowout games and lost more close ones than “average” he would do poorly. To me, this is a question of managing style and philosophy, rather than how good a manager actually is.

Let’s take a hypothetical example to illustrate what I mean. It’s the 7th inning and the visiting team is down 1 run. There’s 0 outs and a man on 1st and 2nd. Small-ball manager might be included to bunt and hope one of the next two players get a hit, going up by one with the idea of winning the game in the 8th and 9th with his relievers. Big-run manager decides to let his players swing away, with the idea that a 3 run HR or similar feats will lead to a big inning, which will make the home team unlikely to be able to score enough to win in general. Let’s also suppose that both are successful in their respective strategies 50% of the time (that is given 100 such games, both will win 50 and lost 50).

Small-ball manager will have far more 1 run wins and far fewer 5 run wins simply because of the style in which he used the outs and hitters he had, and Big-run manager would have many more big wins and far fewer 1 win wins of the 50 games each won. But both won 50 of the 100 games available to them, so each were equally successful in the job a manager actually has: winning ballgames.

My point is that both your systems seem to penalize the Big-run manager, because the big wins he is able to manage through his style will suggest the team should have actually won more games than 50, while the Small-ball manager will be awarded because due to his limiting the totally number of runs his team scores your systems is underrating how good his teams actually were.

Take this simple illustration and multiply it by the thousands of situations in which different philosophies might yield different inputs for your measure of “good vs. bad” managers, and you begin to see that any such simple system is going to only give a very very misleading at best picture of how good a manager actually is. My concern is that simply because you can use available data to rank managers, you will suppose the ranking actually means something in terms of how good the managers are at winning baseball games without considering the correlative aspects of the data you are comparing.

I’ve done some research in this area, so that’s probably why I like your article so much. I did not do as extensive a research as you did regarding pythagorean because I didn’ t know how to grab all that data and automate it. I collected them by hand and found basically what you found, I only looked at recent managers, and found simlar names to your list, Torres, LaSorda, Bochy, Alou, and I had the same surprising fact that Lasorda was among the worse in the data set I collected. One name, however, that was on my list back then that you don’t have is Dusty Baker. He must have done pretty poorly after leaving the Giants.

I forgot how I did this (I did this about 8-10 years ago), but I did (or tried to do) the statistical testing to see how significantly away from the mean of 0, and none were at the 90% level, but a good number of them were in the 60-70% level, and that’s still pretty good.

However, one flaw that bugged me and hence why I never followed up on this research is that, for example, Joe Torres was below Pythagorean by a lot when he was managing horrible teams, but then he magically became a good manager when he was managing the Yankees. This lines up with your observation about managing good vs. bad teams. So perhaps Pythagorean is not the best way to measure a manager’s worth, there might be something about the Pythagorean that is under valuing losing teams and overvaluing winning teams.

Another flaw with this system, which is major for your purposes, is that perhaps the manager is doing something that helps the team score more runs and/or prevent more runs allowed. That is not measured at all by the Pythaorean methodology (as the team would win/lose more, in line with the formula), and that, I think, is what kills it as usefulness in measuring a manager’s worth. I think that this measures part of the picture, and thus is worth looking at, but it is not the whole picture and thus not definitive.

I’ve been meaning to post my research on 1-run win/loss record at Fangraphs, but lost my notes, but maybe you can do the more extensive research back into history necessary to check validity, since you seem to have access to automated data analysis. I was just checking Bruce Bochy’s record in 1-run games since both he and Gibson were compiling many more wins in that category in 2011, as the old rubric there is that any manager who wins a lot, will regress to the mean and likely swing back the next season. Since it was relatively easy using BB-Ref, I pulled down team’s record in 1-run games in the NL for a number of seasons, looking at how teams did, and he kept on showing up among the leaders, so I kept on compiling until I covered his whole career, finding that he kept on showing up among the leaders in most games above .500 in 1-run games.

I forgot how many games he was above .500 (my memory recalls something in the 60’s plus I recall him averaging roughly 4 more wins each season he managed), but assuming a .500 mean, he was statisically significant at the 90% level to be above .500 in 1-run games.

Also, I compiled all the leaders by season for his career, IDed the manager for each team, and Bochy was among the leaders each season by far the most of any NL manager during his career, compiling more 8 seasons or more than any other manager, I think he had 40% of all managerial seasons like that, despite only managing roughly 7-8 percent of the managerial seasons available in the NL during his career. The only managers to consistently show up among the leaders were Bobby Cox (early in Bochy’s career, he was done in the 2000’s; apparently Bochy and he are friends, maybe even close friends), Dusty Baker (he was like a Bochy lite, roughly half of what Bochy did), I think LaRussa showed up a bit, but not as much one would think as he’s been hyped as a manager. The only manager with an extended streak of high above .500 seasons was, if I recall right, Bobby Valentine (it was the Mets manager early in Bochy’s career as manager), who had a string of +10 seasons, unmatched by anyone during Bochy’s career, but then he stopped managing in the majors.

Given that your WAR methodology places Bochy in the Top 4 among managers in history, Bobby Cox too, that suggests to me that you might have something there with that methodology in analyzing managers. I agree with the other commenter that there are a lot of factors that play into where a manager contributes value, many very small. But ultimately, unless it contributes to wins, the factors don’t really matter, right? It has to matter at the win/loss level in order for it to matter for measuring this.

My only caveat on the WAR methodology is that I assume you are using defensive valuations of WAR in there, and we don’t really have a system yet that is THE method. There are too many players who look good in one system and bad in another. And given the lack of data going back historically, any analysis will be limited in that way. But I like the names on your list and it is matched up, as you noted, by many same names via the Pythagorean method, as well as my 1-run win/loss record sampling of the NL over the past almost 20 years, so I think that gives your list strong validity.

Looking at your list, I saw that Frank Robinson was up there on one list, and I viewed his gruff style to be similar to Billy Martin in that they get results in their first year but then wear out their welcome.

I was also curious how Ted Williams did in his short stint as manager. I remember a quote from a player noting how Williams love to work on hititng with everyone, but then that gave short shrift to everything else. But looking at the stats showed that a lot of hitters improved greatly under his wing as manager. Maybe look only at the change in offensive WAR for his teams in the season before and his first season? That’s something that can’t really be measure either by WAR in your methodology, because the pitching could go up or down no matter what Williams did, maybe there were trades or an injury or poor performance just due to randomness, whatever, plus, as noted, the defensive WAR varies widely sometimes for even the same player, season to season, though generally consistent.

This is a very cool jumping off point. I’ve been recently thinking about trying to do something similar to the WAR analysis, only using historical Marcels (and I guess TZ for D) weighted by actual playing time. Seeing if there are managers that have players that consistently beat their projected WAR. We’d need huge sampe sizes, but hey we’re looking on the career level….

Btw, I found this article that may interest some..

being on the Pythagorean list. It’s called Mariano Rivera.

Any other managers on here with long-term dominant closers (or the reverse?)