THE BOOK--Playing The Percentages In Baseball

Monday, August 25, 2008

Five of the ten best teams play in the AL east?

By Tangotiger, 10:05 AM

As AsrosFan is telling us:

I was just glancing at Andrew Dolphin’s ratings of MLB teams. I decided to go with the Predictive ranking.

Predictive Both schedule strength and rating vs. schedule strength are determined considering both the wins and losses and the score differentials. This rating contains none of the biases in the standard rating, but does weight recent games slightly more than past games since those are a better indication of the team’s current strength. This rating is the best of the first three for seeing how good teams are, and thus is the best for predicting future results.

I could have used Improved RPI, but I wanted one that included run differential. Here are the rankings of all 30 teams: http://www.dolphinsim.com/ratings/mlb/index_pred.html

The first thing we notice is that the AL is ranked as having 10 out the top 11 teams. Mr. Dolphin’s analysis shows the AL as being much stronger, and thus teams with worse records than NL teams can actually be better due to strength of schedule. Indeed, looking at it from the expected losses standpoint, which Mr. Dolphin says “can be used to rank schedules”, the top 14 SOS are in the AL, meaning all the AL teams. And it’s not really close.

That would be astonishing enough, but this is what really caught my eye. Down further on the page, the rankings are split up by division. If we scroll to the AL East, we find these rankings: Boston: 1 Tampa Bay: 3 Toronto: 7 New York Yankees: 9 Baltimore: 10 . All five teams in the AL East are ranked in the top 10, that is, the top third of baseball. Baltimore, which at the time of the latest update, had a 61-66 record and RS/RA of 653/653, is ranked two spots ahead of Milwaukee, which had a 75-55 record and a RS/RA of 616/555.

To which I responded to the above and other questions in that thread:

When it comes to manipulating numbers in an intelligent manner, I’ll put Andy up against anyone.

I also have no insight into what Andy did, other than he does something similar for many sports. His site is blocked at the office, but from what I remember, it’s a standard “strength of schedule” adjustment, something more important in college ball than the pros.

The AL/NL difference between Orioles and Brewers makes sense. I use a standard +.025 wins per game for AL and -.025 wins for NL. That is, I consider the average AL team to be a true .525 team and the average NL team to be a true .475 team.

If the actual AL win% is .505 (because AL plays themselves so much more than they play NL), then you need to add +.020 wins per game for each AL team. And similarly, subtract .020 to each NL team. If you have some 125 games, that means you add about 2.5 wins or 25 runs differential to the AL team and subtract 25 runs to the NL team. So, a 50-run difference between an AL and NL team in NL team’s favor is really making the two teams equal.

Now, Andy did say he weights the more recent game a bit heavier (as he should, and that’s what the day-to-day Marcel does), then if the Orioles have been playing far better recently, then they’ll rank a bit higher than they normally would if you look at their season overall.

All that to say is that Andy’s got method behind his madness and deserves no outright dismissal. At the very least, it requires questioning even before skepticism.

• Sabermetrics • Forecasting • Talent_Distribution

#1 john (see all posts) 2008/08/25 (Mon) @ 11:10

I use a standard +.025 wins per game for AL and -.025 wins for NL. That is, I consider the average AL team to be a true .525 team and the average NL team to be a true .475 team.

I know this has probably been answered but how did we determine that the average AL team is a true .525 and NL a .475?

#2 Tangotiger (see all posts) 2008/08/25 (Mon) @ 11:40

You can read MGL’s 3-part series on the matter that was at Hardball Times. Checkout their archives, and post a link back here for others…

#3 john (see all posts) 2008/08/25 (Mon) @ 11:52

http://www.hardballtimes.com/main/article/is-the-al-really-superior/

There’s a July 5th, 2006 article that MGL wrote. Has the balanced changed any since then?

#4 AED (see all posts) 2008/08/25 (Mon) @ 11:52

I’m not entirely sure what is surprising about all of this.

Regarding the AL and NL relative strengths, the AL went 149-103 in this year’s interleague play, and is 576-432 over the past four seasons. In the quick-and-dirty sense, this would indicate that the average NL team should expect to go 69-93 against a 162-game schedule of AL teams. Specifically, if 75-55 Milwaukee played all of its games against AL opponents, it would be more like 66-64, not coincidentally putting it right in line if you compare standard ranking vs. record.

It’s definitely the case that the predictive ranking is more refined, but I’d contend that the rankings are in line with what one should expect.

Side point—more recent games are not weighted more heavily in baseball.

#5 Tangotiger (see all posts) 2008/08/25 (Mon) @ 15:23

http://www.hardballtimes.com/main/authors/mgl/2006/

#6 brent (see all posts) 2008/08/26 (Tue) @ 00:44

This is another reason that baseball needs to come up with a better system of divisions and playoff spots.

#7 MGL (see all posts) 2008/08/26 (Tue) @ 04:00

While the methodology for estimating team “true talent” from their w/l record or their RS/RA (which is better of course) plus their SOS may be sound, it is such a blunt instrument that I would give it little attention.

Of course if you know nothing else, it will have to do. But even the casual fan can do better. You don’t have to be a sabermetrician or a genius to know that for whatever reason, CLE and SD are much better than their records indicate, that OAK is not nearly the same team that it was at the beginnning of the season, that MIN and CWS are not that good despite their record, that ALA has played way over their heads, that MIL is a much better team with Sabathia, that ARI is much better with Haren and Dunn, that HOU is not a very good team despite their recent success. Etc. Etc. How much attention should we pay to a statistic or a measure that a casual fan can trump? I say very little.

The reason why this method is such an imprecise instrument to say the least is two-fold of course:

One is the disconnect between a team’s true strength, which is a function of each player’s true talent prorated by their playing time and simply added together, and their w/l record (or pythag w/l record), which is driven by true talent plus the random fluctuation associated with sample error, the latter being quite large of course.

Two is the constantly changing true talent of a team associated with injuries and personnel changes.

Here is a challenge to someone with a lot of time on their hands:

Look over the last 20 years of post-season play. Figure out the true talent of each team going into the post-season based on the following:

Do a Marcel (going into the post-season, which is basically the next year’s projection)) for all the players who had significant playing time in the post-season. Add them all up, prorating the projections by their post-season playing time.

Now rank each team according to that true talent estimate. Now see of often the best team, according to this true talent estimate, wins each series or game. Compare that to how often each team with the best season w/l record wins each series or game. (You will have to do a league adjustment for the WS of course).

I guarantee that the former will crush the latter!

#8 tangotiger (see all posts) 2008/08/26 (Tue) @ 07:50

Brent: totally agree.

I always suggest the Olympics format, whereby when you get down to the final 4, you make AL1 play NL2, and NL1 play AL2.

This way, if the Rays/Sox are the two best teams in the league, they at least have the chance to play each other in the World Series.

However, in baseball, you have to contend with “tradition”, which is defined in baseball as “I hate any idea that changes the current status quo, even if the current status quo is only 10 years old”.

It would take the NBA, NHL, and NFL to make the change first so that we can see what a great idea it is, for MLB to finally adopt it.

Or, as I noted about the NHL’s propensity to change divisions every 5 or 10 years with little fanfare and ZERO complaints: create a system of changing divisions every 4 or 8 years, ala World Cup.

Page 1 of 1 pages

<< Back to main

THE BOOK--Playing The Percentages In Baseball

Monday, August 25, 2008

Five of the ten best teams play in the AL east?

Latest...