Feng: Tigers, Indians in dead heat for AL Central title
Will the Tigers win the AL Central this season? The team has many question marks entering the season. Will Justin Verlander will return to his Cy Young form of a few years ago? Can Miguel Cabrera shake off the injuries that plagued his 2014 season and hit like the Most Valuable Player again?
Everyone has their opinions. Some diehard fans watch 162 games a year and might even make a wager on a game or two. Other quant types like myself crunch numbers to get the unbiased insight from data.
Here, we'll combine these various predictions into one prediction for 2015 win totals. These types of ensemble predictions are usually more accurate than any one prediction by itself. Nate Silver of FiveThirtyEight uses this ensemble method for his March Madness predictions. In addition, I started combining various college football predictions this past fall and got a firsthand look at how multiple predictions compensated for the mistakes of each other.
The first ingredient in our recipe is the betting markets in Las Vegas. These win totals represent the collective wisdom of people sure enough of their opinion of the Tigers that they're willing to put money on whether they win more or less than 84.5 games this season.
The ensemble of preseason baseball predictors also includes six different win total predictions from analytics. This group represents a diversity of methods, from the player-based PECOTA model of Baseball Prospectus to the calculations of Joe Peta that consider the clustering of hits in scoring runs (I'll have much more to say about this in future columns). This list gives these six predictors.
Since the markets provide a much different prediction than analytics, I've weighted this win total four times more than an analytics prediction. With an equal weight for each of the numbers based predictions, the market makes up 40% of the ensemble predictor.
What does this ensemble prediction say about the AL Central?
First, let's get the bad news out first. The model predicts more wins for Cleveland than Detroit this season.
However, there's also good news for Tigers fans. The ensemble model has Detroit right behind Cleveland. With a prediction of 83.9 and 83.6 for Cleveland and Detroit, respectively, it's a dead heat between these two teams for the AL Central.
I used the market predictions from Bovada, which had both Detroit and Cleveland at 84.5 wins. However, the markets favored Cleveland to finish under 84.5 wins while it made no such distinction for Detroit. Another online market had Cleveland as low as 82.5 wins as recently as Sunday.
In the six predictors from analytics, Cleveland finished ahead of Detroit in 3 of them. However, Detroit was ahead in two others while the two teams had the same prediction in the last model.
Here's the take-home message on the AL Central: The ensemble model predicts a tight race between the Tigers and Indians. The markets ever so slightly like Detroit more, while the number crunchers give a small edge to Cleveland.
The ensemble model also gives a certainty for each team. With seven different predictors, each team has a spread in win totals that can be measured through the variance in these numbers. With Detroit's question marks, I'm surprised the Tigers have an average uncertainty in their win total. Cleveland has the seventh-smallest uncertainty.
What about the other AL Central teams? The ensemble model predicts 80.3 wins for Kansas City. No one is impressed with their World Series appearance last season as the model has the Royals as a below average MLB team. Chicago is right behind Kansas City with 79.5 wins, while Minnesota brings up the rear at 72.3 wins.
The ensemble model has its good features, but it's not any type of magical formula. None of the predictors knows whether Joe Nathan will end the season on the mound in Detroit or in a rocking chair. You have every right to adjust this prediction based on watching games.
Complete list of win totals from ensemble model:
1. Washington, 93.43.
2. Los Angeles Dodgers, 92.68.
3. St. Louis, 88.55.
4. Los Angeles Angels, 87.61.
5. Seattle, 87.23.
6. Boston, 86.16.
7. Pittsburgh, 84.17.
8. Cleveland, 83.94.
9. Detroit, 83.64.
10. Chicago Cubs, 83.44.
11. San Diego, 83.28.
12. San Francisco, 83.17.
13. Oakland, 82.90.
14. Toronto, 82.29.
15. Baltimore, 81.92.
16. New York Yankees, 81.35.
17. New York Mets, 81.25.
18. Tampa Bay, 80.68.
19. Miami, 80.67.
20. Kansas City, 80.34.
21. Chicago White Sox, 79.46.
22. Milwaukee, 78.85.
23. Cincinnati, 77.17.
24. Houston, 75.86.
25. Texas, 75.76.
26. Colorado, 72.50.
27. Atlanta, 72.30.
28. Minnesota, 72.27.
29. Arizona, 71.96.
30. Philadelphia, 68.48.
Ed Feng still has bad memories of the 1993 World Series as a Phillies fan, so he became a Tigers fan upon moving to Ann Arbor. He applied his Ph.D. research in chemical engineering to sports analytics and runs the site The Power Rank. His column will appear regularly at detroitnews.com