The Art of Forecasting Baseball Stats: BSports vs. ESPN & MLB.com

Failure is a big part of baseball.  A career three hundred hitter fails in roughly seventy percent of their at bats and even the most disciplined hitter strikes out on occasion.  MLB players, fans, and front offices are always turning to the past to forecast the future and on many occasions past performance is not indicative of future results.

Predicting the future is impossible but with advanced analytics and sophisticated projection models, Bloomberg Sports, MLB.com and ESPN attempted to project the performance of nearly every MLB Player for the 2012 season. Let’s see who developed the best projections across several fantasy baseball statistical categories.

BSports vs. ESPN & MLB.com

In order to determine the best 2012 MLB player projections we compared all players with at least 400 at bats.  This was done to avoid players who were influenced by outside factors such as injuries.  The data set was thereby narrowed to 172 players (MLB.com did not have projections for Norichika Aoki so we excluded him from the analysis even though he had greater than 400 ABs).

Next, we compared the actual 2012 statistics of the 171 MLB players that had 400 or more plate appearances to the 2012 projected stats for Bloomberg Sports, MLB.com and ESPN to determine which projection set was most accurate.  We calculated the R² value for each statistical category by comparing the 2012 actual results with the 2012 projections for each organization.  The winner in each category has the highest R² value.  Bloomberg Sports decisively was more accurate than ESPN and MLB.com in every statistical category with the exception of runs.

As you can see in the chart below, in the categories that account for individual performance (HR, SB, and AVG) BSports clearly outperformed the other companies.  In the categories that include team situations (R, RBI) BSports finished 3rd and 1st respectively with runs being the most difficult statistic to project having the lowest R² value of any statistic.  Clearly, BSports had the more accurate projections when comparing the players who had more than 400 at bats.

Below are some of the best examples of how precise the Bloomberg Sports 2012 Projections were and how they nearly mirror the actual 2012 statistics for these players.

Predicting the future is impossible; however, it is possible to accurately forecast future performance and results based upon a sophisticated calculations and insight. Does Bloomberg Sports have a crystal ball or do they understand how to forecast baseball player’s offensive performance better than anyone else in the business?  For an answer we reached out to Bloomberg Sports developer Craig Glaser to find out what makes Bloomberg Sports unique at developing projections.

Glaser describes the main difference between Bloomberg Sports projections and the competition’s as:

“It comes down to math/science vs. opinion…  Opinion only enters into one very specific area of our projections, playing time. Even with playing time we don’t rely on our in-house analysts to come up with our projections – we go to the outside world and ask hundreds of knowledgeable baseball fans in an effort to take advantage of the wisdom of the crowd’s phenomenon.  Beyond that our projections are entirely based on the evidence which the players have provided about their true talent level.  The core concept of our projections is regression – we break down a player’s results into a number of different, independent skills – see which of those skills are more predictive than others, and then regress each skill by the appropriate amount.  We have a number of other adjustments involved (aging, ballpark, role, etc.) in addition to regressions and all of it is based on significant research about how a player’s past performance has best predicted his future performance.”

In 2012, the Bloomberg Sports projections proved to be amongst the most accurate in the industry.