THE BOOK--Playing The Percentages In Baseball

Tango, I certainly agree with you when it comes to assessing skill and projecting performances.  But what about looking backward to see who had the best season?  Not saying who actually was a better pitcher, but who gave their team a better chance to win.  Acknowledging that almost all of that difference is attributable to luck as far as we can discern, would you take that into consideration?

I would think it would amount to no more than a tiebreaker, going on gut alone.

Actually, I just wrote a piece about this a few weeks ago and it’s on Pine Tar Press

http://www.pinetarpress.com/?p=9299

Tango’s intuition is correct in that pitching “inconsistency” is better and hitting consistency is better.  My estimate is that an “inconsistent” pitcher ends up beating his Pythag by 1% as compared to average and a “consistent” pitcher is about 1% worse than his Pythag.

On the hitting side, I have some initial analysis (not yet confident in it fully) that one can have a more consistent offense by having the right balance of OBP and SLG,

If I plot run scoring variability on the Y-Axis and the ratio of OBP/SLG on the X-Axis for teams over the last 10 years, I get a faint “U-Curve” leading me to think that the bottom of that curve is a “sweet-spot” of balancing your lineup between OBP and SLG.

Yeah, I did one of those “balanced hitting teams” once a long time ago at BaseballBoards.com (RIP).

It was something like having one type of team will win .520, and another type of team will win .480, even though both teams would have the same runs per game.  So, that’s +/- 3 wins, based on “synergy” or lack thereof.

If you spread that out to each player, that would mean an extra 0.3 wins or so per player, for “synergy”.

There’s more to leverage from synergy than from consistency.

***

As for backward looking: sure, that’s going to have a pretty substantial impact.  Scoring 22 runs in one game doesn’t have much impact for each run.

With regards to the comments above, additional quick analysis suggests that consistency is a lot like “clutch” in that it may explain why things happened in the past, but if Team A had a “consistent” pitching staff in 2011, it is not particularly predictive that they will also be consistent in 2012.

I disagree somewhat. I believe that you want consistency (i.e. less variance) when you are good and inconsistency (i.e. more variance) when you are bad. If my pitcher options, in a league average of 4.5, is 0-10-0-10 or 5-5-5-5, I’d choose 0-10-0-10. If my options are 0-6-0-6 or 3-3-3-3, I’d choose 3-3-3-3. I may be missing something.

It all depends on what you run level you define as consistency.  A consistent pitcher who lets up either 2 or 3 runs each time out would have a slightly better record (.717) than an inconsistent one who let up either 0 or 5 (.696).  Of course you would have the advantage of being able to yank the inconsistent pitcher when he was having an off day.

If you allow exactly 3 runs (in 9 innings), against a team that averages allowing 4.5 runs, you lose 27% of the time, you tie 15% of the time, and you win 58% of the time.  If you split the ties in extra innings (with some other pitcher), then you win 65.5% of the time.

If you allow exactly 6 runs, you lose 68%, tie 10%, and win 22%.  That gives you 27% wins.

If you allow 0, you win 96%, tie 4%, and lose 0%, for an overall 98% wins.

So, 0-6-0-6 gives you 62.5% wins.  Therefore you are correct.  Thanks for straightening me out.

Of course if you have a pitcher that averages only letting up either 2.5 or 3 runs a game you aren’t going to quibble that much on how he does it.

The assistant GM in Will’s article seems to take it further than which one is better for your team in the long run.  He is applying something beyond expected value.  He is using a mental formula along the lines of

Utility = Expected - (RiskAversion X Variance)

The assistant GM has a risk aversion factor.

I understand the math of the consistency issue, as calculated by Pythagorus (sometimes it calculates as better or worse), but I think that is small potatoes compared to the real world. If you told a manager at the beginning of the season that every player on the team (pitcher or hitter, star or scrub) would play as consistently as possible within his overall production, I assume he would take that in a heartbeat. At least I would.

And that is because, knowing exactly how every player will perform, I (an avg manager) can leverage them much more optimally, relative to the typical manager who has no idea how a guy will perform on a day to day basis around his “uncertainty” range.

Of course, I have no concrete evidence for this, and so it is only a theory or expectation. Maybe it only would make a manager feel more comfortable, without any actual W/L benefit at the end. But it would take some strong argument to move me off this position.

I think Monte/#6 makes a good point. It depends on your offense. If you’re the Yankees, you want the 3-4-3-4 pitcher, because you’re scoring 5.5 runs a game. If you’re the Giants, you’re better off with the 1-6-1-6.

I did a regression study of constistency. Here is the link

http://cybermetric.blogspot.com/2010/04/how-much-does-team-consistency-matter.html

Team winning pct was the dependent variable and runs, runs allowed and a consistency variable for both scoring and run prevention were the independent variables. The consistency variables did matter, thought not a huge amount and they tended to matter in they way people are mentioning here. Being more consistent in scoring runs per game is good as is being less consistent in allowing runs per game.

I also did a study on the value of a balanced lineup. It seemed like being more balanced in OBP helped teams score more but being less balanced in SLG helped teams score more

http://cyrilmorong.com/BalanceBTN.htm