Baseball Articles | Index
Predicting the Final Standings in the 1996 AL
Written by Tom Tippett
This is the time of year when your favorite beat writer or SportsCenter host starts talking about how team such-and-such is favored in the wild card race because their team has a much easier schedule than the other contenders. But you won't see them try to quantify that statement in any meaningful way. Is the easier schedule worth one game in the standings? Three? Five? How much does it really matter?
Over the past few years, we've developed a way of predicting how the final standings will turn out if teams play at the level they have established up to that point in the season. There are two important elements to the method. First, each team's projected winning percentage is calculated based on the strength of their remaining schedule. Second, strength of schedule is based on each opponent's expected winning percentage given the number of runs they have scored and allowed to this point in the season.
Let's look at the second part first. Bill James developed the idea (called the Pythagorean Theory) that you could accurately predict wins and losses by looking at runs scored and runs allowed. Specifically, the projected winning percentage is the square of RunsFor divided by the sum of the squares of RunsFor and RunsAgainst. In any full season, most teams will finish within one or two wins of their predicted record. There will always be a few teams that win 6-8 more or fewer games due to a particularly good or bad record in close games. We'll let you argue about how much of this is due to luck and how much is due to the character of the team and its players. For our purposes, it's sufficient to point out that this formula is remarkably accurate.
For games through August 25, 1996, most teams are within a game or two of their predicted record, as expected. Three of the exceptions are especially interesting:
Our method calculates each team's expected winning percentage for the remainder of the season by looking at their own expected performance and the expected performance of their opponents, weighted by the number of games against each opponent. (It would be good to build in the home field advantage, too, but we haven't done that yet.) The chance of winning each game is the average of the team's expected winning percentage and their opponents expected losing percentage. For example, suppose the Orioles are playing Oakland. The O's expected winning percentage is .525 and the A's is .472. We give the O's a 52.65% chance to win this game, since they should win 52.5% of the time and the A's should lose 52.8% of the time.
Baltimore has the easiest remaining schedule. They're finished with Cleveland and Texas. They have 21 games remaining with sub-.500 teams, including 7 with the Tigers. Their remaining opponents have a combined winning percentage of only .464, weighted by the number of games against each team. The Yankees are next at .473. Even though the Yankees aren't as good as they've looked, this break in the schedule should be enough to keep them on top. The toughest schedule belongs to California. Not only do they play half their remaining games against division leaders and all but five of the rest against wild-card contenders, but their record to date is quite inflated (see third bullet above).
Here are the projected final standings based on information through the games of August 25th:
W L Pct GBL New York 91 71 .562 - Baltimore 86 76 .531 5 (tied with Chicago for Wild Card) Boston 82 80 .506 9 Toronto 76 86 .469 15 Detroit 60 102 .370 31 W L Pct GBL Cleveland 95 67 .586 - Chicago 86 76 .531 9 (tied with Baltimore for Wild Card) Minnesota 81 81 .500 14 Milwaukee 77 85 .475 18 Kansas City 74 88 .457 21 W L Pct GBL Texas 92 70 .568 - Seattle 83 79 .512 9 Oakland 77 85 .475 15 California 75 87 .463 17
These look a lot like the current standings, you say? Yes, they do. That's partly because 80% of the season is already in the books. Even though the Yankees and Angels have played over their heads, those bonus wins can't be taken away. But it also points out that strength of schedule doesn't mean all that much if teams continue to perform to expectations. (That's a very big IF, so we'll come back to it.) This method suggests that the Orioles will win 53.1% of their games, because their run differential indicates that they should win 52.5% of the time against average opposition and their remaining opponents should lose 53.6% of the time. So their "easy" schedule translates into about one more win than if they were playing opponents of average strength.
The big question, of course, is whether teams will perform as they have done this year. Some teams are markedly different from their early-season makeup due to injuries (either new ones or players returning from earlier ones), trades, or minor-league callups. Teams that have clinched their division or are out of the race will start playing their September callups more often, presumably hurting their performance to some degree. In any 30-game stretch, some teams will be especially hot or cold, and we don't think it's possible to predict streaks with great accuracy.
You can be fairly sure that the real-life results will be different from what we've predicted, but we thought it might be interesting to show you these results anyway.
September 4, 1996 Update
Things didn't change a whole lot during the first week after we posted the predicted final standings that you see above. Our discussion will focus on the wild card teams, since we're ready to concede the division titles to the teams currently in front.
The model now predicts that the White Sox will win the wild card by a game over Baltimore. Chicago went 4-2 last week as they faced two of the weaker teams on their schedule (MIL, TOR). Baltimore had a decent week, going 4-3 against Oakland and Seattle. Baltimore's remaining schedule remains the easiest by a large margin, with 17 of their remaining 26 games against Detroit, California and Toronto. So, if the O's schedule is so easy, why does the model still pick the White Sox? Mostly because the Sox have outscored their opponents by 96 runs compared to only 35 by Baltimore. The model says the better team (Chicago) will overcome the stronger schedule and hold on for that last playoff spot.
Here's the updated projection for the wild card teams, along with the change from the previous week:
W L Pct GBL Change Chicago 87 75 .537 - +1 Baltimore 86 76 .531 1 0 Seattle 84 78 .519 3 +1 Boston 82 80 .506 5 0 Minnesota 80 82 .494 7 -1
September 11, 1996 Update
Once again, things didn't change very much in the past week. The Yankees are now projected to win two fewer games than in the August 25 version, and every other team is on track to finish within one game of their August 25 projections.
The Yankees won only 3 of 6 from Oakland and Toronto, so their projected win total has slipped to 89 games. Baltimore took advantage of their easiest week (California and Detroit) to go 4-2 and increase their projection to 87 wins, tying them for the projected wild card lead with Chicago, who matched Baltimore's 4-2 record against stronger opponents (Detroit and Boston). Boston gained one projected win by splitting six games with good opponents, Seattle and the White Sox.
The Yankees are projected to win the East by two games over Baltimore, making this the projected finish in the wild card race:
W L Pct GBL Change Chicago 87 75 .537 - +1 Baltimore 87 75 .537 - +1 Seattle 84 78 .519 3 +1 Boston 83 79 .512 4 +1 Minnesota 81 81 .500 6 0
The last column shows the change in win projection that has occurred in the past two weeks. So far, anyway, it looks as if the contending teams have all picked up the pace a little bit.
Baltimore no longer has the easiest remaining schedule, now that six of their games with California and Detroit are in the books. With 11 of 20 remaining games against Detroit and Toronto, it's still the weakest of the wild-card contenders. But it's no longer easier than the Yankees schedule, who have seven games with Toronto and Detroit and a crucial home-and-home series with Boston.
September 18, 1996 Update
For the first time since we started three weeks ago, teams began to depart from their trend lines in the past week. The league's weak sisters failed miserably to live up to their role as spoilers and the White Sox squandered an opportunity to win key games against contending teams.
The Orioles were the big winners, winning their final four meetings with the Tigers and taking a crucial series with the White Sox for a 6-1 record. Although the O's no longer have a cupcake schedule, they took full advantage of their games against Detroit with six wins in seven tries. Chicago dropped four of six against Boston and Baltimore to put themselves in a very deep hole with only twelve games to play. If they don't make the playoffs, this week will be the reason. The Red Sox had a normal week (3-3 at home against Milwaukee and Chicago) and lost ground to the surging Orioles. The Yanks kept pace with Baltimore as they beat up on Detroit and Toronto for a 5-1 week. Minnesota, which wasn't really in the race anymore, put the final nails in their coffins with a 2-4 week. Seattle kept their hopes alive by taking four of six from Minnesota and KC.
The Yankees are still projected to win the East by two games over Baltimore, making this the projected finish in the wild card race (the last column shows the change in projection from 8/25):
W L Pct GBL Change Baltimore 89 73 .549 - +3 Chicago 86 76 .531 3 0 Seattle 85 77 .525 4 +2 Boston 83 79 .512 6 +1 Minnesota 80 82 .494 9 -1
As we mentioned, Baltimore no longer has an easy schedule, but with seven games against Toronto, it's not a killer either. With a three-game bulge in the wild-card race, they're not likely to give up their lead. And with Baltimore, Boston and New York playing most of their remaining games head-to-head, it goes without saying that this projection system means very little at this stage. What matters now is who wins these crucial games.
As far as these projections are concerned, the most interesting event of the past week was the combined 0-13 record posted by Detroit and California. In the first draft of this article three weeks ago, we wrote that the Angels were the team most likely to collapse down the stretch, since they had been winning more games than they should all year, had just installed a new manager who then went into the hospital, faced a tough schedule the rest of the way, and had nothing to play for. But we pulled this comment before publication because it was pure speculation. By the way, we didn't feel the same way about the Tigers, who had been showing some signs of life (their team ERA has come down by more than a half run per game in recent weeks). Do weaker teams consistently fall short of projections down the stretch? We don't know, and we don't have time to study this right now. Maybe we'll come back to this next year.
September 25, 1996 Update
In the past week, the Yankees and Mariners were the big winners. New York won 5 of 7, a win over Toronto and a pair of victories against each of their division rivals, Baltimore and Boston. Seattle went from wild-card contender to a serious threat for the West division crown after sweeping Texas and taking two of three from Oakland. Cleveland continues to mop things up after clinching the Central against the White Sox early in the week.
The Yankees are now projected to win the East by four games over Baltimore. They'll have to work for it, with five remaining games with the arch-enemy Red Sox and a pair with the pesky Brewers. If Baltimore can hang close through mid-week, they could make things interesting with their final four-game set in Toronto.
Out west, Texas is projected to hold off the Mariners with two games to spare, despite their recent troubles. Strength of schedule isn't much of an issue here, as both teams finish with the Angels and Athletics. If it stays close, however, Seattle has one game to make up with Cleveland, which should help Texas hang on.
Here are the current projections for the wild-card teams, with Minnesota dropping out after a 2-5 week, and only Seattle having a new projected win total compared with a week ago:
W L Pct GBL Change from 9/18 Baltimore 89 73 .549 - 0 Seattle 87 75 .537 2 +2 Chicago 86 76 .531 3 0 Boston 83 79 .512 6 0
The odds are very much against the Red Sox, who have the most ground to make up and the toughest remaining schedule. They must sweep their two-game home series with Baltimore and hope that the Yankees clinch early and take it easy over the final weekend. All three of the other contenders have easy schedules for the final week, so the race will be won by the team that can avoid the upsets. (By the way, these projections are based on stats through Sunday, September 22, but I'm writing this on Tuesday. Baltimore and Seattle lost their Monday games while Boston took an extra-inning thriller from the Yankees. The ray of hope for the Red Sox is just a tad brighter today.)
Last week, because the Tigers and Angels were in the midst of long losing streaks, I wondered whether the weaker teams consistently fall short of projections down the stretch. This may be true, but the NL cellar-dwellars are providing some counter-examples. The Pirates ripped off an eleven-game winning streak over the last two weeks, and the Phillies and Giants have both played .500 ball lately.
October 3, 1996 Update
Now that the season has ended, let's take a look at how well our model predicted the real outcome of the race over the past five weeks. Here are the actual final standings (see projections above) with the difference in victories shown in the last column:
W L Pct GBL Predicted wins New York 92 70 .568 - 91 +1 Baltimore 88 74 .543 4 86 +2 Boston 85 77 .525 7 82 +3 Toronto 74 88 .457 18 76 -2 Detroit 53 109 .327 39 60 -7 W L Pct GBL Cleveland 99 62 .615 - 95 +4 Chicago 85 77 .525 14.5 86 -1 Milwaukee 80 82 .494 19.5 77 +3 Minnesota 78 84 .481 21.5 81 -3 Kansas City 75 86 .466 24 74 +1 W L Pct GBL Texas 90 72 .556 - 92 -2 Seattle 85 76 .528 4.5 83 +2 Oakland 78 84 .481 12 77 +1 California 70 91 .435 19.5 75 -5
Our impression is that the biggest failure of the model was the inability to predict that Detroit and California would go into tailspins during the last five weeks of the season. The Tigers were 6-26 and the Angels were 9-22. We thought the Angels were the most likely team to underperform, but we definitely didn't expect this from the Tigers.
Baltimore finished two games ahead of its projection. Was it because they traded for Eddie Murray, Pete Incaviglia and Todd Zeile? Maybe. Zeile's performance with the O's (.239, 5 HR in 117 AB) was below average for a guy playing a hitter's position. Incaviglia posted some good numbers in 33 atbats. It's possible these three guys made the difference. But if you take out their games against the Tigers, they did exactly what they were projected to do without these guys. So maybe it didn't make any difference after all.
To a large degree, the extra wins picked up by New York, Boston, and Milwaukee can also be traced back to the Tigers. The Red Sox also picked up some meaningless wins against New York the final weekend.
The biggest disappointment, in our minds, was the mediocre performance of the White Sox. Although they finished only one game below their prediction, they were able to go only 15-15 the rest of the way against a collection of slightly-below-average opponents. One would hope a playoff contender would do a little better than this. For the second year in a row, Cleveland continued to win even after clinching the division early.
Out West, Seattle surprised Texas with a four-game sweep at home, but the race went according to form the rest of the way. The Mariners didn't have enough pitching to make up more ground. We think their improved health accounts for the two extra wins they picked up. Their projected finish was based on their performance through August 25, which included the period when Ken Griffey and Edgar Martinez missed 22 and 23 games respectively. If we could have assumed that those two guys would be in the lineup every day, the projection would have been higher.
We hope you've enjoyed following along with this little exercise. As we said at the beginning, we didn't really expect the model to predict the final standings. We know that trades, injuries, streaks, slumps, late-season callups and luck would all play a part in producing some results that we didn't expect. And it wouldn't be a whole lot of fun to watch the pennant race if we already knew how it would turn out. Our goal was to give you another way to look at the races. We hope our projections, along with our week-by-week analysis, gave you a little something to think about.
Pythagorean Projections -- End of Season Update
During our August 25 write-up (above), we also mentioned the Pythagorean Theory, which predicts wins and losses from runs scored and runs allowed. All but three AL teams finished with a win total that was within three games of their Pythagorean projections. The three who didn't:
Copyright © 1996. Diamond Mind, Inc. All rights reserved.