THE BOOK--Playing The Percentages In Baseball

I can’t believe that people are not smart enough to realise that Atlanta must have been lucky at home and unlucky away. I means last year Atlanta was better away. (Home: 40-41, Away 46-35). Are we to believe that some of the players who were good last year away are now not able to play on the road? I would love to live in a world were people think and try to learn before accepting opinion given by journalists.

The sad part is that Dayne Perry used to be one the regulars at BP.  Surely he knows this is poppycock…

Link to Dayn’s article:
http://msn.foxsports.com/mlb/story/Teams-battle-for-home-field-advantage-in-baseball-playoffs-091410

It would have been interesting to see if perhaps rookies do better at home than on the road, so Heyward, for example, would get a bigger advantage.

And, isn’t the home win% in the post-season smaller than regular season?  If I had to guess, it would be .520 or even less.

Anyone want to provide that info?

I think that HFA is larger in the post-season (especially the WS), but I am not sure either.

And I am pretty convinced that HFA has increased significantly (at least .55 now) since the end of PED’s (mostly greenies), especially in day games.

Of course if you look at post-season games, I am not sure there is a large enough sample size to be very certain as to the true HFA.  Do we even have 1000 post season games in the last 50 years?  If yes, and we find a HFA of .52, the 2 sigma confidence interval is .49 to .55.  That doesn’t tell us much of anything, does it?

For the World Series 1922-2008 (excluding 1922-23 and 1943-45), the home team was 268-200 (.573), higher than the regular season W%.  Of course that’s WS and not all playoffs, but I agree with MGL--it doesn’t tell us much.  Using the binomial distribution, there’s a 7.1% chance that a .540 team (roughly the regular season home W%) could go 268-200.

@#1.  You do realize it is a different team at a number of key positions.

If it looks like a duck, walks like a duck and quacks like a duck, it probably is a duck, even if a journalist says so and you can not prove it statistically at the 95% level.

Looking at their H-A record, they should want HFA. Badly.  That’s a huge split.  Of course, Chipper Jones had the biggest split, so maybe they won’t be as good as home without him, and it won’t matter.

If you accurately know which teams have a “true” much larger or much smaller HFA than average, or what the public thinks, then you have no excuse for not being a millionaire.

If that sentence wasn’t clear, I am saying put your money where your mouth is.  Bet this theory into a fortune.  My guess is it won’t happen (for a reason).  Atleast not for the author of the piece linked to.

#9, correct of course.

Someone posted a study (THT, BP?) where they ran a regression to predict HFA. They found the significant variables and then used the regression equation to estimate each team’s (or park’s) “true” HFA.  Every team was slightly different (with uncertainty of course, such that in reality most teams could have been functionally equivalent), with a few teams being significantly different (including COL and BOS and the Twins in the Metrodome, all for obvious reasons of course).  In fact, every team but those 3 were within .01 of each other.

ATL, BTW, had a slightly LOWER estimated HFA than average.

Bottom line is that the idea of a team’s actual H/R record in one season telling us something about their future true HFA is nonsense, and one would not need any analysis to know that.

The reason the last part of the sentence is true is this, and it is a very important point in general:

If you see lots of variation among teams (or players) in splits in only, say, 135 (or any small number) opportunities, in order for those variations to have any significant predictive value, the variance (standard deviation) in true talent for that split would have to be enormous.

So, you could answer the question at hand, by asking yourself this:

Do you think that the spread of true talent among MLB teams in terms of HFA is very, very large?  IOW, do you think some teams have a true 60% HFA or more and some teams have none or even a slight (or more) disadvantage?

If the answer to that is, “No,” or “Of course not, don’t be silly,” then it is not possible for a small sample of H/R data to have much predictive value.

FWIW, Perry could have went to B-REF and looked up the Braves Home/Road Splits.

Hitting
Home: tOPS+ of 105
Away: tOPS+ of 104

Pitching: FIP (with the formula that I used, obviously)
Home: 3.5 FIP
(56*13+(222+22-19)*3-600*2)/686=0.2959
0.2959+3.2= 3.5

Away:3.8 FIP
58*13+(223+22-35)*3-503*2)/621.3=0.6084
0.6084+3.2= 3.8

Not a huge difference. Considering that the park over the past few years has been less neutral on HR(from ESPN).

2005:  0.880
2006:  0.929
2007:  0.955
2008:  1.014
2009:  0.861
2010:  1.043
Total: 0.947

Seasonal figures, for reference sake:

1901-09 .533
1910-19 .540
1920-29 .543
1930-39 .553
1940-49 .544
1950-59 .539
1960-69 .540
1970-79 .538
1980-89 .541
1990-99 .535
2000-09 .542

(http://www.baseballprospectus.com/article.php?articleid=9372)

The long answer is that every team probably has a slightly (ever so slightly) different true HFA, depending upon the characteristics of their home park and how their team is able to take advantage of that park...

I used to think it was all about the characteristics of the park, but after looking at some other sports (soccer in particular, where HFA can be huge, even though the differences in fields is rather minor compared to baseball’s bizarre field configurations), I’d guess that travel has as much to do with HFA as anything.  It would be interesting to look at HFA on the first game of a home stand compared to HFA on the last game of a home stand.  Or HFA on the first game of a home stand compared to the first game of the second series in a home stand.

If travel was the main factor in HFA, I could imagine some ways that it tends to wash out in the aggregate season statistics.  In baseball, teams are constantly traveling, so when you play the first game of a home stand, you might have had just as long of a flight as the opponent, or maybe an even longer flight.  Then, on the second and third games of a home stand, your opponent might not get any home cooking, but at least they would not have had to take a plane to make it to the game.  In sports with games played less frequently (like, say, soccer), the home team generally gets at least a good 3-4 days at home running up to the game, maybe up to a week or more.

Of course, in the playoffs, both teams essentially travel equally.  (Except perhaps for the first game, you sometimes get a situation like the Twins coming to New York having taken a late flight the night before while the Yankees slept in their own beds.  That’s pretty rare, though.) So ultimately, I’d be surprised if HFA was a major factor in the playoffs, but I don’t necessarily suspect that field/park configurations and roster construction necessarily play a big part in that.

It’s just a hypothesis, of course, but I think that perhaps we get too distracted by the differences in field configurations and don’t concentrate on the other factors which may be important.

There has been some good research on HFA in baseball.  IIRC, travel was NOT a big consideration.  Actually familiarity of the park and the “quirkiness” of the park (both related of course) are probably the larger factors (other than just being home, sleeping in one’s own bed, etc.).  Given that HFA advantage is small in baseball, for ALL teams (other than perhaps the Rockies, for obvious reasons), and given that a series has to go to the last game in order for the extra home game to be played, HFA is definitely not much of a factor in the post-season…

IIRC, the biggest components to the HFA is walks, strikeouts, and triples.  This suggests that familiarity with sightlines (K, BB), and quirkiness of the fences (triples).

Looking at my binomial home field adjustments for 2009 (this is five year rolling average):

UIBB : .007
SO : -.009
HR: .004
XBH: .005
3B: .020

UIBB should be UIBB/(PA-IBB-HBP), SO should be SO/(PA-IBB-HBP-UIBB), and so on for HR. XBH is (2B+3B)/(H-HR-1B_INF). 3B is 3B/(2B+3B).

Good stuff Colin!  Love it when everyone rolls up their sleeves.

In order to be able to compare them fairly, you need to then rebaseline then against the same denominator (say PA).

So, since walks are about 8% of all PA, then we need to multiply the SO numbers by 92%, bringing it down to -0.008.

The denominator of HR is 75% of all PA, so the HR impact is .003 per PA.

The denominator of XBH is about 20% of all PA, so the XBH is now only .001 per PA.

The denominator of triples is 5% of all PA, so the triples is now only .001 per PA.

Here are the adjusted numbers per PA:
-.008 K
+.007 NIBB
+.003 HR
+.001 XBH (includes triples)
+.001 3B

You can also multiply each of those by the run values to get the impact those things have.  Once you do that, you see HR take the lead:

+.0042 HR
+.0021 NIBB
-.0018 K
+.0011 3B
+.0009 XBH (includes 3B)

You can alternatively divide the original numbers by the average frequency of each event.  So, if you have +.003 HR per PA, and the average is .030 HR per PA, then the impact is 10% of PA.
20% 3B
10% HR
9% NIBB
5% SO
2% XBH (includes 3B)

Lots of ways to think about it.

This suggests that familiarity with sightlines (K, BB)

Or a difference in the umpiring.  And perhaps, on the margins, that the home pitcher comes in warm and throws harder in the top of the first.

From Eric Seidman in a recent BP article (emphasis mine):

“The Phillies were virtually in, and the Braves weren’t far off. In spite of the high odds for both teams, home-field advantage is an important animal, especially given how poorly the Braves have played on the road.”

Grrrr!

A. Home field advantage is NOT particularly important in the post-season, mainly because of the structure of each series (the last game is the extra home game).

B.  Neither the Braves nor any other team (other than the Rockies) is going to (significantly) benefit more so than any other team in the post-season.

Eric should know better!

As usual, if Eric or anyone else actually believes that the Braves’ (or any other team’s) 2010 h/r record has any significant predictive value, there is a bet waiting to happen.

I hear crickets as usual…

More silliness from the same article (sorry Eric, but surely you know that):

“This Phillies team should beat Beachy and Minor, even if they tend to struggle against pitchers they have never seen before.”

So if tomorrow (or the next day) the Phillies face a pitcher they have never seen before, they are likely to do worse than any other team would do against pitchers they have never faced?

Eric is a Phillies fan, and I think that fans of every team feel their team struggles against pitchers they haven’t faced before.  I think people tend to expect a team, especially a good offense, to really rip a new pitcher and knock him out of the box after 3 innings.  If they pitch 5 IP and allow 3 R, it isn’t remembered as a good offensive showing, even though it is 5.4 R/9.  I think fans underestimate the quality of a replacement level pitcher.

Yes, one would hope that BP writers wouldn’t fall into that trap, if that’s indeed what Eric was thinking.

Sure, you would expect that kind of thinking from a fan.  But a good sabermetric writer like Eric, and in a BP article?  No…