THE BOOK--Playing The Percentages In Baseball

I thought it was a good article by Walsh also.  I would love to see, if it has not been done already, if there is a skill element to driving in runs over and above (or under and below) that which is expected given a player’s overall profile.

This gets us back to the “clutch” debate, where clutch in this case is defined as changing your approach (or perhaps simply doing better than expected) depending on the bases/outs state in order to maximize your driving in runs performance.  Of course that still does not mean that you are doing something good.  A player could conceivably alter his approach such that he drives in more runs than expected given his overall profile, yet substantially reduces his value added, also given his overall approach.  An extreme example would be a player who never walks with runners on base but not the bases loaded.  That would obviously be a terrible strategy, but one that would not hurt (and proably help) his driving in runs ability.

One thing I have been curious of lately.  Is it possible that the value of the (non-intentional) walk for a power hitter is less than that of a non-power hitter.  You would think so.  You would think that a non-power hitter would rarely be walked, for example, with no outs and no one on (when it has maximum value, I think) and that a power hitter would mostly be walked with runners on and bases empty, or at least with runners on, period.  Has anyone ever looked at the average value of each event (walk, single, etc.) for different types of hitters (power/non-power, free-swinger/disciplined, etc.)?  If my “theory” is correct, we have to start using different linear weight values for different classes of hitters in order to compute their overall value as compared to another player in a different class.

Well, at that point, you’re a step away from what we should be using: BZR (UZR for batters). Of course, with BZR, there’s the need to account for a player’s speed in computing the probabilities, but that’s do-able.

That’s not really what I meant.  BZR, if you mean, applying a run value to a batted ball, given the batted balls’ characteristics, is not what I mean.  In any case, if that is what YOU mean, it does not work, not to mention the fact that in the long run it gives you no more information than simply using a batter’s actuall events (s,d, outs, etc.).  The reason it does not work is, for example, that all “hard hit ground balls” are not created equal.  The bigger, stronger players will hit harder “hard hit ground balls,” etc.

I think this is what MGL is talking about:

You start with this table:
http://www.tangotiger.net/RE9902event.html

(Or the Markov version, Table 50 in The Book.)

Figure out the LWTS for each player’s events, by base/state.  Average the LWTS by player-event.

So, say you have Bonds, with 100 walks with bases empty 2 outs, and 50 walks with man on 2b, 0 outs.  That gives us:
100 x .13 + 50 x .48 = 37 runs on 150 walks, or .25 runs per walk

We know the overall run value of a walk is .33 runs, so we can see that a Bonds walk (in this illustration) has much less run value.

MGL’s question is if certain profile of players, the Thomes, etc, has a run value of a walk that is different than others.  That is, does Thome/pitcher change their approach sufficiently that the impact of a random Thome walk have different impact than a random Ichiro walk?

Yes, that is what I mean.  And you don’t want to just take the lwts value of each event for each player and use that to forecast player value.  It is a little like using win added value or whatever it is called.  First we need to know if a player’s lwts value for each event is luck or “skill,” the skill being, as Tango states, a change in approach by the batter/pitcher.  As I initially said, you would think that the walks for power hitters would be worth less than those of non-power hitters, etc., but I have never heard of anyone doing the research to see if my theory is true and to what extent.  It probably would not make a whole lot of difference, but it might make a significant one for some players.  It is a little like using lwts batting order to figure player value and for forecasting player value.

Gotcha. I’ll take your word on BZR, though in that case, that raises some serious questions about PZR to me as well. I think you’re right about the value of a walk, but I would assume that it would be drowned out by the slight increases in the value of singles and doubles, which as James Click showed (here: http://baseballanalysts.com/archives/2005/10/extra_base_hits.php) have slightly more “driving-in” value for bigger hitters.

MGL, I don’t know if it would be based on a certain family of hitters.  I don’t think you can necessarily treat Thome, Hafner, Giambi as the same family, in terms of batting approach.  But, it’d be interesting to see.

I would definitely suspect that Ichiro and Barry do have a different batter/pitcher approach (whether they approach it differently, or the pitcher approaches them differently).  I think it’s a very individual thing.