At BJOL this morning, James brought up a new paradigm -- the issue of "Natural" and "Conditional" effects.
.....
The current use of closers is not as irrational as models may make it look because these are human beings. I have managed a lot of different types of workers over the years. In general, people are a lot more effective when they know what is going to be expected of them. A closer who can get into a routine, pace himself, watch the score and get himself physically and mentally ready as necessary may be a more effective pitcher. Trying to identify the highest leverage use puts him in a situation where he has to prepared to go in at any time in the game, it assumes that the manager knows when to best use him, it risks warming him up a couple of times and not bringing him in... So while not theoretically ideal, in real life bullpen roles may help pitchers to be more effective when they are used and takes away what could become a very complicated calculus from a manager. And that may be worth the trade off. Plus, of course, for the reliever, a certain status from being a closer.
Asked by: raincheck
Answered: 12/23/2013
Well. . .I think you are sort of on the corner of the truth there. I don't doubt that what you are saying is true; it's a little imprecise. Combining this issue and the issue we were also discussing, about why runs scored are low in post-season. .. . 1) We sometimes tend to assume, in our field, that you can manage real players the way you would manage an ABPA or Strat-o-Matic team, 2) This assumption is very useful, in that it flies in the face of a lot of conventional wisdom which is actually nonsense, but 3) This assumption is not absolutely valid, either.
...
Conventional wisdom ties together all manner of performance units into imaginary packages. Conventional wisdom, for example, holds that hitting is contagious, when in reality it is not. Conventional wisdom assumes that if a hitter has 8 hits in his last 11 at bats, he is red hot (and thus more likely to get a hit in his next at bat), whereas if he is 0-for-15, he is "cold", and thus less likely to get a hit in his next at bat. Conventional wisdom believes that if you put Good Hitter B in the lineup behind Good Hitter A, that will make Good Hitter A a better hitter, when in reality it will not. Conventional wisdom insists that if you put a fast base runner on base, the next hitter will see more fastballs and thus will hit better, when in reality he will not. These are major assumptions that we hear constantly, but there are a thousand more like them that we hear occasionally or infrequently.
Clutch hitting is a conditional effect--a major one, obviously. If a pitcher has to run the bases, the announcer will question whether this will cause him to lose effectiveness the next inning, although studies have shown no such effect. Generally speaking, whenever the sportswriter or announcer tries to tie together one event in a baseball game with another, he is usually spouting nonsense, and one should generally ignore him. Tim McCarver would insist to the end of his career that if a leadoff man drew a walk, it was dramatically more likely that he would score (and dramatically more likely that there would be a big inning) than if the leadoff man hit a single. Sometimes he would put a number with this supposed difference. In reality, there is no difference whatsoever; a leadoff walk has exactly the same effect as a leadoff single. Another popular one is the idea that after Team A scores a run, it is really important for them to have a "shut down inning", and prevent the other team from "answering" the run.
In our area we tend to ignore this kind of blather, and assume that each player has a skill set and each combination of players has a set of probable outcomes, which are the same without regard to who is on deck, whether there is a fast runner on base, whether the runner on first reached by hit or error, etc.
Players are players; they do what they do. This is generally true, but it is not ABSOLUTELY true.
.....
I like to distinguish between NATURAL effects, which are those effects which would be replicated in a simulation (or table game) of ordinary sophistication, and CONDITIONAL effects, which are those effects which are dependent on conditions of the game, and would NOT be replicated in a simulation or table game of ordinary sophistication.
Most conditional effects that baseball men talk about are complete nonsense and should be ignored--but not all of them are. The Home Field advantage, for example, is a conditional effect, albeit such a common one that we accept it without complaint.
In discussing why runs scored decrease in post season, I explained a couple of days ago that batters don't hit well in cold weather. That's a conditional effect. Simulations rarely adjust for the temperature at game time, so we wouldn't replicate that in a model. Predictably, I got pushback from a dozen or more readers, saying "No, no; it's just that quality pitchers pitch a disproportionate number of innings in post season play." That would be a NATURAL effect.
I expected that reaction, because people in our field always tend to believe that everything can be explained as a natural effect--but it isn't. It's a conditional effect, based on the weather.
This, also, is a conditional effect: Closers are more effective when used in one role than when used in another. I think we can begin to demonstrate that this is true by asking this question: In the last 40 years, has there been any starting pitcher who was as effective, batter for batter, as Koji Uehara was in 2013? There hasn't been. He had a 1.09 ERA, a component ERA lower than that, gave up less than one hit for two innings pitched, held batters to a batting average of .130 or .131 or something, and had a strikeout to walk ratio (not including IBB) of 14 to 1. I doubt that any starting pitcher in the history of baseball has ever been that effective. Well, says the naturalist, but that's 70-75 innings. No pitcher has been that effective in a SEASON, but starting pitchers have had runs or 70-75 innings in which they were extremely dominant. OK, but. . .what about Craig Kimbrel? If you look at Kimbrel, Papelbon, Mariano Rivera. . .these pitchers have reached levels of effectiveness, sustained over a period of years, beyond that of any starting pitcher, ever. I believe that a) no starting pitcher in the history of baseball has been as effective, batter for batter, as Jonathon Papelbon, and b) Papelbon is the LEAST effective of the three that I mentioned. Table games either a) ignore this effect altogether, and creates "codes" for these pitchers which are the same regardless of how they are used, or b) put this effect into the game as a limitation or defect; that is, pitchers LOSE effectiveness if they face more than X number of hitters. But it isn't a limitation or defect; it is a positive ADVANTAGE. Pitchers GAIN effectiveness when used in that way.
These pitchers (and others. . .ie, Eckersley in the early 1990s, Dibble). ..these pitchers reach that level of effectiveness BECAUSE they are used in this particular way. It is a conditional effect. Trying to be careful about what I am saying.
I am not saying that the use of closers in today's game is absolutely optimal, and that there could not possibly be any better way to do it. I am saying that there is a conditional advantage in the current usage pattern which the people who criticize that usage are not recognizing and not dealing with. We CAN'T recognize it or deal with it, because we really don't know how to do that. We can't predict who will have that advantage or when they will have it; we can't say with confidence how large it is, or when EXACTLY it would be lost if pitchers were used in a different way. Because we can't put parameters on it (with confidence), we are inclined to deny that it is exist. But I, at least, am 100% convinced that it does exist.
.....
This is an issue that Dr. D has yelled about for 10, 20 years: that sabermetricians are fine with believing that something exists if it is simple enough to measure.
I'm not sure why Bill uses the terms "natural" and "conditional" -- after all, everything he lists under "natural" is also "conditional," that is, occurring BECAUSE something else occurred.
But on the other hand, his remark about "We could replicate it in a table top game," that one is very useful in terms of understanding the problem. ERA's go down in the playoffs; well, we could send ERA's down in Strat-O-Matic by allowing CC Sabathia and Mariano Rivera to pitch half the innings.
Rather than calling "increased pitcher innings" a Natural effect and "reliever effectiveness in a set role" a Conditional effect, I would call them Simple effects and Complex effects. Or "Strat-O-Matic'able" effect.
Or just "effect simple enough for us to detect by measurement" and "effect NOT simple enough for us to detect by measurement," but of course, this is precisely the situation that provokes such an emotional response among the sabermetricians. They hate the idea that such a thing is possible, and react by scoffing and jeering. James, above, attempts to analyze the issue coolly.
.....
I think Bill himself, in his zeal for facts and research, sometimes misses the idea that something CAN exist even though you're not MEASURING it.
True, he misses this idea less than anybody else does. The entire point of his mini-essay, above, is to get sabermetricians to open their minds about the existence of "conditional" effects -- effects that would never exist in Strat-O-Matic, but which do exist in the major leagues.
Bill himself argued that catcher impact on pitching could exist even though CERA was indicating that no such effect existed. He did this by proving that even if you stipulated a +1.00 ERA advantage for a catcher, that the saber models would not detect this actual skill.
I'm not a fan of "lineup protection" logic, but I do hear pitchers -- all the time -- SAY that they DO pitch Robinson Cano differently if they don't respect the hitter behind him. That is something real: the pitchers are changing their pitches. If our models are not detecting the results, that doesn't mean something real isn't occurring.
The pitcher just flat told us that he's going to behave differently. "But, pitcher," we say, "My little stat here is not showing you as behaving any differently." Overrulled, counselor.
.....
An example of "conditional effects" that pertain to the 2014 Mariners: Jack Zduriencik is convinced that Eric Wedge did a spit-poor job of managing his bullpen in 2013.
Dr. D also believes that the bullpen wayyyyy underperformed. Charlie Furbush, for example, walked 4+ men per game and ran a 3.76 ERA. At times he got frozen out. I wonder whether the 2014 bullpen already has the seeds of success, with or without Grant Ball Four.
.....
At any rate, the light bulb is obvious. Just because an aspect of the game would never show up in Strat-O-Matic, that doesn't mean that we get to dogmatically overrule the people who play the game and tell us that it does exist.
Happy holidays,
Dr D