THE BOOK--Playing The Percentages In Baseball

Wednesday, March 31, 2010

How much does playing in a strong division hurt a team?

By , 01:29 AM

You hear all the time about what a tough time teams like TOR have (it used to be TB until they somehow, miraculously, got good) playing in a division against teams like Boston and the Yankees, perennial powerhouses, at least lately, due in part to their large payrolls.

There are two aspects to this: One, simply is that it is tough to win a division or even the wild card when one or more teams in your division are very good. That is true regardless of how good your team is. Obviously, the worse the other teams in your division are, the better your chances of making the post-season, regardless of your own strength. Secondly, because of the imbalanced schedule (the fact that you play many more games against teams in your division than outside your division), your average opponent is tougher, thus “artificially” reducing your own number of wins. Of course, if, for example, two other teams in your division were really tough, and two other teams were really bad, only the first thing would come into play, and I am not even sure about that - certainly it would at the extremes (if two teams were 110 win teams, two teams were 40 win teams, and you were an 85 or 90 win team, you would have almost no chance of winning your division, even though your average opponent is about average).

Anyway, I am only going to address the second thing - how the strength of your average opponent affects your win/loss record. The way I am going to do that is to look at each player’s average opponent and then add everything up per team. For example, if player A on TOR faced an average opponent that was 5 runs above average per 162 games, Player B’s opponents were 2 runs above average, etc. and we added everything up after prorating by each player’s playing time, we would get the average of the whole team’s opponents in runs/wins above average. There is no particular reason why I did it player by player. I just happen to do that because it is one of the ways in which I adjust for context when I do my player projections so I already have the data organized that way.

The way I determine, BTW, the strength of an opponent is to use a projection for that opponent, based on their last 4 years’ performance (just like a regular projection). That projection, without an age adjustment for the next year (which would be the final step in an actual player projection) is a proxy for that player’s true talent over that year. So, for example, if in game one of the season, batter A from TOR faces pitchers who are projected to be 10 runs better than average per 162 games (basically a RA of .06 - 10/162 - better than average), he gets “credit” for 4 or 5 PA facing an opponent who was +10 (per 162) better than average.

Anyway, for 2009, here are the team totals. Keep in mind that even if everyone played the same number of games against every opponent in their league (and say there were no IL games), a bad team is going to face an above average overall opponent and a good team, a below-average overall opponent. So we expect the Yankees to face below-average opponents right off the bat, and KC, above-average ones. How much? Well, your team divided by one minus the number of teams in your league (again, assuming no IL games). So, if the Yankees were a true 95-win team in 2009, they would be a true +140 run team, and each team they played against would be a true -11 run or -1 win team. It goes without saying that if a team is a true +11 win team, those 11 wins have to be deducted from all other teams - the rest of the teams in the league are all true 80 win teams. So, a true 94 win team (against a neutral opponent) will actually win 95 games (on the average)!

Also keep in mind that the overall strength of your pitching opponents has partly to do with the vagueries of each team’s pitching rotation and not just on the strength of your opponents’ overall team pitching. You might just happen to face a team’s best or worst (and everything in between) pitchers over the course of a season.

For these numbers, plus means that the opposition was good (bad for the listed teams) and minus means that they were bad (good for the listed team).

Team/Opponents batting/Opponents pitching/Total Opponents

ARI -10.5, 4, -6.5

So Arizona faced players who were 10.5 runs worse in hitting per 162 games than the league average, and 4 runs better in pitching, for a total of 6.5 runs worse. So, their “strength of schedule” was good for them to the tune of .65 wins.

Here are the teams that benefited the most from their opponents:

CHN -11.8, -5.6, -17.4
MIN -9.6, -7.7, -17.3
SLN -9.3, -5.3, -14.6
CLE -11.2, -.5, -11.7
ANA -7, -3.5, -10.5

Everyone in the NL central other than SLN and CHN (their true talent that is) was bad! And I guess it is good to play in the AL Central, especially if you are MIN and CLE. Remember that we are using true talent estimates to come up with these opponent adjustments, and true talent estimators thought that CLE had a good team last year! Oh, and now we know why ANA won so many games last year - easy schedule!

And here are the teams that were hurt the most:

WAS 13.3, 8.3, 21.6
BAL 17.2, 3.3, 20.5
TOR 18.2, .6, 18.8
TBA 9.7, 7.4, 17.1
NYN 16.2, 1.7, 17.9

As you can see, it is no picnic to have had to play the Red Sox and Yankees! And because WAS and the Mets were so bad last year, everyone else in the division was good.

And the rest:

ATL 5.1, -4.1, 1.0
CIN 1.9, 1.2, 3.1
COL -11.8, 3.3, -8.5
FLO 6.3, .8, 7.1
HOU -1.9, 1.1, -.8
LAN -9.3, 2.8, -6.5
MIL -6.8, 3.6, -3.2
PHI 2.6, -3.3, -.7
PIT -4.3, -1.3, -5.6
SDN -1.5, 7.4, 5.9
SFN -2.3, -2.6, -4.9

BOS 11.3, -2.6, 8.7
CHA -3.5, -2.3, -5.8
DET -2, 2.7, .7
KCA -1.9, -2.4, -4.3
NYA 4.4, -5.5, -1.1
OAK -1.3, 1.1, -.2
SEA 1.5, -5.4, -3.9
TEX -2.9, 3.4, .5

• Sabermetrics • Forecasting

#1 JB H (see all posts) 2010/03/31 (Wed) @ 08:50

These results think an NL team moving to the AL would lose about half a win. That’s far below where everyone else seems to have it.

Do you think you’re adjusting for league quality enough?

#2 (see all posts) 2010/03/31 (Wed) @ 09:24

Would I be interpreting the chart correctly to say that the Orioles, Blue Jays, and Rays “lost” about 2 wins to strength of schedule, the Red Sox about 1, and the Yankees were neutral?

#3 (see all posts) 2010/03/31 (Wed) @ 10:16

Interesting, I’ve always thought it was incomplete to try to explain the difference in leagues simply at the league level, it seems to me if people want to calculate adjustments based on competition level, it really needs to be at the division level rather than the league level. This article seems to support my thoughts.

Quick question - I’m sure you have a good reason for using the 3 letter abbreviations you do, but they seem strange and not all that intuitive to me, like FLO instead of FLA, SFN isntead of SFG, SDN instead of....I dunno, something else, just curious where those abbreviations come from and why you use them?

#4 Tangotiger (see all posts) 2010/03/31 (Wed) @ 10:24

Before I discovered online databases, I always used SFG. SFN comes from Retrosheet. For the Expos, it was always MTL, not MON as Retro has it.

Somebody made a choice for IDs, and that’s that, basically.

***

I don’t know why you think FLA is more “natural” than FLO. Personally, I used FLO. I suppose if I were from Florida, I’d use FLA if that’s the standard there, like MTL seems far more natural than MON.

#5 Greg Rybarczyk (see all posts) 2010/03/31 (Wed) @ 10:24

Does this account for the fact that a good team naturally makes its in-division opponents records worse, and vice-versa? If you beat your in-division opponents badly all year, you make their records really bad, which then makes your “strength of schedule” look weak, because you played all these teams that lost so many games. This is tough to avoid if you want to stay within a season (which you pretty much have to, due to roster turnover)…

#6 (see all posts) 2010/03/31 (Wed) @ 10:40

Greg, I don’t think that’s a big problem. For instance, the Red Sox went 16-2 vs BAL, 9-9 vs NYY, 9-9 vs TB, and 11-7 vs TOR last season and they still come out with a +8.7 schedule.

#7 Greg Rybarczyk (see all posts) 2010/03/31 (Wed) @ 11:41

Hmmm… might not be much of an issue most of the time. Probably a bigger deal in college football…

#8 Patrick (see all posts) 2010/03/31 (Wed) @ 12:29

"Does this account for the fact that a good team naturally makes its in-division opponents records worse, and vice-versa? If you beat your in-division opponents badly all year, you make their records really bad, which then makes your “strength of schedule” look weak, because you played all these teams that lost so many games. This is tough to avoid if you want to stay within a season (which you pretty much have to, due to roster turnover)… “

Well you’re talking in terms of “records”, and unless I misunderstood, that’s not quite right. This is broken down on the basis of invididual PLAYERS and calculated on that basis.

IE, how many games a team actually won is never used in this analysis.

There’s some details - Does playing a lot against harder competition in past years affect a players true talent prediction negatively? Sure, but probably just a little bit.

#9 Xeifrank (see all posts) 2010/03/31 (Wed) @ 14:00

Why don’t you use your simulator and swap schedules between a team like the Yankees and Twins and see how this effects their expected win total before vs after.
vr, Xei

#10 MGL (see all posts) 2010/03/31 (Wed) @ 17:47

Patrick, right, the SOS is based on individual player projections, so it has almost nothing to do with past w/l records, per se.

Xeifrank, since the sim is based on player projections, the expected w/l totals should match up with my SOS numbers.

Mitch #2, yes, exactly.\

Right, the abbreviations are from retrosheet. The “rule” is for all one word cities, it is simply teh first 3 letters and for multi-word cities, it is the first letter of the first two words, followed by N or A. In 2005, ANA (Angels) was changed to ALA, although it probably should have been changed to LAA. I think at the time the name of the team was still up in the air.

The strength of each team’s schedule and hence the strength of every player’s opponents impact their projections, so when you do projections, this is an iterative process. First you have to do the projections with no SOO (strength of opponents) adjustment for each player. Then you do the SOO for each player based on the projections of their opponents. Then you have to redo the projections. You only need to do that a couple of times.

JB H, each league does not add to zero only because of slight errors in the lwts calculations for each player which is based on rounding error for the out value. There is no adjustment for the difference between the strength of the leagues. Yes, it is true that all the NL teams suffer from playing the AL teams, this is not included in my numbers. It is assumed that both leagues are equal, so you can only compare the SOS numbers of one team to another in the same league. When I do my projections, I do them league by league, such that a zero context-neutral lwts player in the AL (an average player in the AL) is better than a zero player in the NL.

#11 james (see all posts) 2010/04/01 (Thu) @ 10:27

In your initial post you pointed out the two disadvantages of a tough division i.e. a tougher schedule and harder to win or come second.

Could you look at both of these at the same time at a team level if you artificially created a “balanced schedule” with no divisions by ignoring interleague games, and assuming that each teams record against the other teams would be the same if they all played each other an equal number of times so each AL team would play each other by 162/13 =12.46 times a season and NL teams 162/15 =10.8

The top four teams should then make the playoffs.

By comparing reality with this artificial situation could show which teams are winning less games than they would if there were no divisions and also how much does the division/wildcard format penalises good teams in tough divisions and rewards average teams in weak divisions.

James

Page 1 of 1 pages

<< Back to main

THE BOOK--Playing The Percentages In Baseball

Wednesday, March 31, 2010

How much does playing in a strong division hurt a team?

Latest...