THE BOOK--Playing The Percentages In Baseball

I look at the cumulative data every year.  Since the humidor, there is no discernible effect.  However, the only way to do it is to compare players’ road stats when they play for COL to the same players’ road stats when they played for another team.  So you have very small samples to work with.

I have not included the 08 data yet.  I’ll do it and report back.

My intuition is that such an effect would be due to the movement of pitches changing in Coors versus other fields.  So I’d think the humidor wouldn’t impact it at all - the air is still sparse and pitches still break differently in Coors.  Unless, balls move differently in Coors due to their size or composition, as opposed to the air in which they’re thrown.

I guess what I’m saying is… it’s interesting to know that there’s a Coors hangover effect and that it’s disappeared with the increased use of the humidor.  But do we have any clue why that is?

Also… is there an easy way to parse data so we can look at Holliday’s performance in the third road series on a road trip?  Hopefully by then, the Coors hangover will have gone away and we can get a better guess at his stats than taking a look at his total road splits.

I’m an idiot on that last paragraph.  Didn’t read the second half of your link… sorry about that.

It is interesting to read (or re-read) the BP article on a possible hangover effect. They concluded that there was not likely one, however, their sample sizes are small and it is hard to get a handle on an effect when you break the data down into so many buckets.  At the least, they should have aggregated the stats into 2 buckets and shown, for example, “less than 5 games into a road trip,” and “more than 4 days,” or something like that.

In any case, they were looking to see if the Rockies players got acclimated to playing on the road (and also if they got acclimated to playing at home).  IOW, did they do better as a roadstand or a homestand progressed.

I was always looking at something related but different.  I was simply assuming that their players (pitchers and hitters) would do worse on the road as Rockies players than they would (on the road) as non-Rockies players.  I never looked at how or whether this effect might change as a road trip went on.  In fact, I did not think that there was likely any difference in the “road effect” as a road trip went on, and BP, with limited data, saw no evidence for that either.  I have not looked at it exclusively however.

My effect is simply a “Rockies road effect,” not a “hangover effect,” which implies something transient which will wear off.

However, I do sometimes call it a hangover effect.

Anyway, I’ll have some numbers later on…

Pinto offers some contrary data:
http://www.baseballmusings.com/archives/030781.php

First of all, BA??!!!!  Second of all, shouldn’t the guy that writes the article at least average up the BA for each column.  I don’t feel like doing that myself.  These “studies” where someone looks at a bunch of individual players are ridiculous.  Just give us the aggregate numbers.  If you want to give us individual player numbers, that’s fine, but meaningless.  As I said, I’ll have my numbers later on tonight.  I’ll have all the components as well as wOBA. Finally, the issue is not about “training at a high altitude and then playing at a low altitude.  This is not a marathon.  The issue is that the Rockies players may be used to the baseball acting in a certain way at home when it is pitched.

I agree with MGL’s impression of Pinto using BA.  In any case, I don’t think David was writing an article so much as just presenting data.

***

I also support this simple use of wOBA:
.7 BB, HBP
.9 1B, RBOE
1.3 2B, 3B
2.0 HR

Exclude IBB, and bunts.

Nice and simple.  I do the 2B/3B thing because alot of what is a 3B is really a double+speed.  Not to mention that there are so few triples to begin with, that they are annoying. 

Plus I get to use one-decimal places for everything.  I never liked the idea of the precision associated with 1.24 for the double.

To reiterate what I do to compute the “road effect” of the Rockies’ players or any team is simply this:

Look at all players’ road stats when they play for a certain team, like the Rockies.  Look at the road stats, not including Coors Field, for any of those players if they play for another team in the same timer period.  Look at the difference and weight by the less of the two PA.

So if player X played for the Rockies some time between 93 and 01 (non-humidor period), I compiled his road stats.  Say his road wOBA was .300.  If that same player played for another team in the NL only also some time between 93 and 01, I would compile his road stats, not including those in Coors Field (so that the final two sets of parks are from roughly the same stadiums - see comment below).  Say it was a wOBA of .320.

So for that player we have a difference of 20 points weighted by whichever sample (road stats while a Rockie or road stats while on another team) has the smaller number of PA.

Get a weighted average of all the differences for all players who played for the Rockies and another team during that time period.

For Rockies 93 to 01, we get a Rockies road wOBA of .309 and a non-Rockies road wOBA of .324.  There is a 1.05 ratio and a 15 point difference.

That is in 9857 PA (adding up all the “lesser of the two PA").  The standard error for the difference between 2 independent samples (don’t know if they are independent - I think they are) of sample size 9857 is 7 points.  So the difference is significant at the .05 level (2 SD).

There seems to be a road effect.

From 02-08, the humidor years, it is .308/.289, a difference of 19 points for 5922 PA. With a standard error of 9 points, that is also significant at the .05 level.  The humidor has not seemed to get rid of the road effect.

Interestingly, I did the same thing for Arizona players, who also play in a somewhat extreme environment - hot, dry, and at a fairly high altitude.  They were .302/.319, a difference of 17 points in favor of when they played for ARI, from 98-08, also significant at better than the .05 level.

I also looked at TB and OAK, two sea-level teams, also for 98-08.

They were, TB, .309/.316, in other words, they did better on the road as a TB player, and OAK, .338/.341, also better as an OAK player.

I’ll try and look at some other teams.  Of course, if I look at all the teams in only 6 or 8 thousand PA, I am likely to get some anomalies by chance alone.

Another issue is the unbalanced schedule.  For example, when playing for the Rockies, your road games are often in the other NL west parks.  When not playing for the Rox, your road stats will be in a random park in the NL, with a slight bias against parks in the NL west.

I probably should park adjust all the stats when compiling the numbers.  I don’t know if it will make that much difference.  For example, if you are a Rockies’ player, you play lots of road games in a pitcher’s park in SD, neutral and formerly pitcher’s park in LA, a pitcher’s park in SF, and a hitter’s park in ARI.  Maybe a little bit of a combined pitcher’s park, but not all your games are against NL West teams of course.

One more thing.  Interestingly, if we look at the individual component differentials for Rockies’ players, we find that on the road when playing for the Rockies you hit more HR per PA than when playing for another team.  For all the other components, you hit fewer, and you strike out A LOT more on the road when playing for the Rockies.  My guess is that Rockies’ players get used to trying to hit the HR at home, and even on the road, they tend to swing for the fences.

Using similar logic, if you play for OAK, since it is hard to hit a HR in the Colosium (I think), maybe on the road, they will hit a lot fewer HR (per PA) then when on the road but playing for another team.  Let’s see.

Nope, that did not fly. While playing for OAK, players hit .032 HR per PA on the road.  While playing for another team, they hit .031.  Of course, the standard error for the difference in only 12389 PA is too large to really know for sure what the “true” difference in HR rate is…