October 8, 2004

Playoff Probabilities

What Are Your Team's Chances to Win It All?

by Clay Davenport

The successor to the in-season odds report is here! The post-season probabilities chart is simply the result of adapting the Monte Carlo simulation used before to the vagaries of post-season play. The key changes that come in are:

The chances for a team to win the game are based on the team's actual season winning percentage, not the Pythagorean or any combination thereof. As seen in an earlier report, while the Pythagorean record started out as a better predictor of future performance, the actual record got better and better as the season went on, so I think it is reasonable to give actual winning percentage its due here.

I have tried to make a rough simulation of the effects of different starting pitchers for each team, by using the average of the team's winning percentage and their winning percentage in games started by this pitcher. For instance, the Yankees have an overall winning percentage of .623. In games started by Jon Lieber this season, they went 19-9, a .679 percentage; in my scheme, they'll get a start value of .651 when Lieber is scheduled. Pitchers with less than 25 starts--I'm talking about Brandon Backe, and Houston's 7-2 record in his starts--had their records moved back towards .500.

As I said, the rotation simulation is rough. I set the program up to run a strict four-man rotation, with the only exception being that they were allowed to skip ahead to a better pitcher if they clinched a series early. That adjustment primarily worked to the advantage of the Twins (25-10 in Johan Santana's starts) and Red Sox (26-7 behind Curt Schilling), getting more games for their respective aces. I'll step in and "correct" the rotation as necessary, but I'll admit that I have no idea how to set it up to handle three-day rests for pitchers who've never worked on that schedule.

As with the regular-season odds report, home field advantage is set at +/- .020 percentage points for each team. The program is aware of the seedings within each league, and sets the appropriate home and away balances for all three rounds of the post-season.

The report itself will look something like this, taken before any post-season games were played:

---------- October 5 ---------------------
     Win DS   Win CS   Win WS    Yesterday
NYY  58.8754  35.6452  19.7384
ANA  38.9045  15.1569   6.1227
MIN  41.1246  18.9131   8.2707
BOS  61.0955  30.2848  16.0831
STL  67.8376  40.9159  22.8490
ATL  53.4882  25.7787  12.6125
LA   32.1624  13.5364   5.3458
HOU  46.5118  19.7690   8.9778

"Win DS" is the percentage chance the team has of winning the Division Series, the opening best-of-five round. In a perfectly even universe, each team would have a 50% chance.

"Win CS" is the percentage chance the team would win the League Championship Series, a best-of-seven series. Our fair universe would give each team a 25% chance of getting this far.

"Win WS" is the chance that they'll win the World Series. Every team that made the playoffs would have a theoretical 12.5% chance, but as we can see a couple of teams are starting with less than half that chance.

"Yesterday" will tell you what the team did yesterday, win, lose, or sit idle, and their current situation in the series, to help track why changes show up in the percentages. The intent is to keep adding each day to the report, so that you can look back at how each team's chances progressed.

You can see the full report, updated daily for the rest of the post-season, here.