The Diff is your weekly WFNY look into the amazing world of sports statistics. For a complete log of articles, click this link. Last week, I wrote about Chip Kelly’s arrival in the NFL. This week, I’m dissecting MLB playoff odds for the red-hot Indians.
It’s no surprise now to Indians fans that the Tribe is on the precipice of being an AL wild card favorite. With the help of the lowly Chicago White Sox and some favorable weekend luck, Cleveland is right in the thick of things. While they remain only one-half game back of the both wild card spots, the more jaw-dropping stat has been related to the team’s mathematical playoff odds. But why do the different sites seem to differ so much? Let’s dive into each one.
Arguably the most popular site for tracking playoff odds and one of the original innovators of this statistic, Baseball Prospectus’ algorithm also likely is the most confusing for the casual fan. It begins and ends with Nate Silver’s Player Empirical Comparison and Optimization Test Algorithm, commonly known as PECOTA. Although Silver no longer writes for BP, his incredibly complicated formula lives on.
In a nutshell, PECOTA thrives on projecting future statistics based on established equivalency charts for various leagues, weighted averages and example career paths for similarly skilled players. Note the final qualifier there, as it is an important difference. The system is controversial for some, but has been terrifyingly omniscient at times as exemplified by the 2007 Chicago White Sox and the 2008 Tampa Bay Rays.
To steal a term from Economics 102 and something that Nate Silver flaunted in his recent book, The Signal and the Noise, BP’s playoff odds formula is based on a Bayesian approach in using PECOTA. This considers the current true probability of an occurrence to be a factor related to the initially assumed probability and factors that have changed since that point in time. If you want a good introduction and some basic probability formulas, head to Wikipedia.
Thus, in order to evaluate the Indians’ current odds, we must know where they started. Silver’s PECOTA system projected the Indians for 80 wins before the season started and the Michael Bourn acquisition, per Let’s Go Tribe. In that context, although the Indians are now 82-69 with only 11 games remaining in the season, their “true” winning percentage is some formula based on the 80-ish projection and what we’ve seen so far.
In general, BP’s projections might seem lower than the other sites because of its more pessimistic consideration of the Indians pre-2013. While that might not seem fair and a bit too BCS-like, it’s how the system works. Day-to-day changes might not be as volatile here too, since the initial season projections still remain as a significant factor throughout.
(Baseball Prospectus’ odds are the slowest to load each night. The stats were updated at 8 a.m.)
Projected end-of-season record: 88.4-73.6
Playoff odds: 54.3%
A rival site to Baseball Prospectus, FanGraphs is even more user-friendly in the 2013 world. Just three weeks ago, the site announced a new partnership with Cool Standings (more on them later) and shared some more information about its playoff projections formulas.
Specifically, the difference between FanGraphs and Baseball Prospectus lies within the projection systems they use as baselines. Over at Baseball Prospectus, it’s PECOTA. For FanGraphs, it is ZiPS, the brainchild of Dan Szymborski from Baseball Think Factory fame.
Whereas PECOTA more heavily relies upon example career paths of similarly skilled players, ZiPS focuses upon weighted averages. Obviously, then, the system is a bit easier to understand. It averages the last three of four seasons for a player – giving more weight to the more recent year – and adds in some extra regression based on uncontrollable factors for the player at hand, like fielding.
ZiPS still does incorporate career projections of similar players and aging trends, but does not share those historical comparisons in a heavily emphasized way as PECOTA does. If you really want to get down and dirty with comparing these two and other projection systems, go give a click and read Tom Tango’s analysis.
In the playoff odds world, FanGraphs also follows a similar Bayesian approach as Baseball Prospectus with regression toward a “true” winning percentage. They viewed the Indians perhaps a bit more favorably before season. Also, FanGraphs now more heavily incorporates season-to-date performance, which again provides a boost for the over-achieving Tribe.
Projected end-of-season record: 88.6-73.4
Playoff odds: 65.9%
Moving from the Bayesian model to the what-have-you-done-for-me-recently model, we have Cool Standings. This website is built purely around playoff odds, which it calculates for all of the major North American sports at any point in time. It has the quickest updates and the easiest to understand system.
This site, instead of relying upon any preconceived notions of a team’s true strength, just uses the team’s run differential in the current season. For a slightly more detailed explanation, you can view their website description.
Run differential is a topic of discussion in most professional sports as a more accurate indicator of a team’s ability than record alone. It encapsulates a bit of regression and dives a little deeper, especially with a large enough sample. Bill James innovated the use of this statistic with his Pythagorean expectation, related to the geometric identity.
Cool Standings uses a variation of James’ model, with the only secretive formula being what exact exponent they use in their calculations. Using run differential, they then project out the rest of the season for every team. Their day-to-day records are also the most comprehensive throughout the season. Another cool feature is to view the top comebacks and collapses in their database.
ESPN.com says that it uses the projections from Cool Standings formula, but, for the life of me, I can’t figure out why they differ ever so slightly. At the least, it’s notable that ESPN’s odds aren’t really unique and we can give credit to the original site anyway for this model and these projections.
Projected end-of-season record: 88.8-73.2
Playoff odds: 68.5%
As a final method, Cool Standings and FanGraphs also offer the option to look at “Dumb Mode”. This shows a stark contrast to any actual forecasting. Instead, this method just assumes every team has equal chances at winning every single game – pure 50/50 coin tosses. While this isn’t really that practical or relevant, it can be a slightly nifty tool to see how the pure toss-up math compares to the actual forecasting.
Projected end-of-season record: 87.5-74.5
Playoff odds: 49.4%
Overall, even with the pessimism of the early-season projections, the current outlook is pretty favorable for the Indians over these final 11 games. Let’s take a look at what the four formulas say about the remaining contenders for the two wild card spots.
With Baseball Prospectus’ stats just being updated recently and incorporating more of their pre-2013 projections, the Indians are emerging as a slightly more likely wild card team than the Rangers. most likely teams. Note how this contrasts with the Dumb Mode projections, whereas the Indians would actually be clearly both Tampa Bay and Texas. This is where Cleveland’s good play so far in 2013 and very favorable leftover schedule both come into significant play.
Another interesting tidbit: There remain about 20% odds that Baltimore, New York or Kansas City will sneak their way into a playoff spot. That’s not insignificant by any means. Those teams are still factors. Baltimore would be the clear most likely team among that group, if it weren’t for their difficult closing schedule.
In the end, it’s neat to note how the projections relatively all line up around one magic number: 89 wins. If the Indians can thus finish 7-4 over their remaining 11 games, they should be lined up for an improbable playoff berth. And that’s all that matters is just to get in.