The MLB season is in the books, and it’s time to look back and evaluate how the CGMs MLB Statistical Stratification System performed. Roughly 60 Games into the MLB regular season I attempted to create a new stratification method to replace conventional power rankings. You can read that article here to learn about the system’s methodology. Baseball is a game where analytics can tell a story, but there fails to be a comprehensive and universally accepted formula for stratifying team performance. Most rankings are determined by the eyeball or litmus test in which random value is applied to one of many different categories to determine performance. I set out to standardized value sets so that from one set of stratification to the next, teams are being evaluated on a consistent scale. To select these variables I looked to baseball statisticians such as Bill James to determine what factors play the most significant role in team performance. By weighing those categories (OPS, BA, WHIP, ERA, and Fielding Percentage), I generated a formula which produced a composite score that closely correlates with a team’s record. For this analysis, I am going to repeat my previous approach to see how the teams stacked up at the end of the year and verify whether or not my approach has validity. Additionally, I will make assertions about ball clubs by breaking down the data set and identify weaknesses in the methodology.
In my June rankings, I found that the formula had a close correlation with the league standings. At that point in the season, I was able to determine that the Tampa Bay Rays were underperforming by comparing their composite score to their record, at the time of the initial analysis, the Tampa Bay Rays wear a slightly better than .500 team (28-27), But they were ranked as my sixth best team. As the season unfolded, it turned out that my metric was accurate, as the Rays finished the season With 90 wins in a tough AL East Division that produced two 100-win teams and the World Series champions.
Similarly, I found that The Los Angeles Dodgers were underperforming according to my metric. At the time of the initial analysis, they were 4 games under .500 (26-30) but ranked in the middle of the pack which means they should have had at least two to three more wins at that point in the season. The Dodgers finished the regular season with 92 wins and the highest run differential in the National League (+194).
New Data Set:
Click here to view the original data set.
End of the Season Stratification:
- Houston – 4. (103-59)
- Boston – 4.2 (108-54)
- Cleveland – 4.9 (91-71)
- LA Dodgers – 5.95 (92-71)
- Tampa Bay – 7.25 (90-72)
- New York Yankees – 7.8 (100-62)
- Oakland – 7.95 (97-65)
- Washington – 8.89 (82-80)
- Atlanta – 9.4 (90-72)
- Milwaukee – 9.55 (96-67)
- Chicago Cubs – 9.55 (95-68)
- Colorado – 10.3 (91-72)
- Seattle – 13.65 (89-73)
- Arizona – 14.3 (82-80)
- Pittsburgh – 15.35 (82-79)
- St. Louis – 16.7 (88-74)
- LA Angels – 17.5 (80-82)
- Cincinnati – 17.75 (67-95)
- Minnesota – 18.9 (78-84)
- NY Mets – 19.5 (77-85)
- Toronto – 20.2 (73-89)
- San Francisco – 20.85 (73-89)
- Philadelphia – 21 (80-82)
- Detroit – 22.7 (64-98)
- Texas Rangers – 23.5 (67-95)
- Kansas City – 23.55 (58-104)
- San Diego – 24.05 (66-96)
- Chicago Sox – 24.7 (62-100)
- Miami – 24.8 (63-98)
- Baltimore – 27.1 (47-115)
- Philadelphia – Hard to imagine saying a sub .500 team overperformed, but according to our metrics they should’ve lost a few more games.
- St. Louis – 88 wins may have been the upper end of their feasible spectrum, especially in one of the most competitive divisions in baseball.
- Washington – This shouldn’t come as a surprise, Washington could barely hold a lead all year. The only team with a bullpen that gave away more leads was the Mets.
- Pittsburgh – If the Cardinals overperformed then the Pirates underperformed. They stratified as a better ballclub than their division rivals, yet managed to lose five more games. Managing might be the unmeasurable factor in their failure to top the Cards.
Findings, Data Set Problems, and Trends:
- Teams with a composite score <10 tended to have more than 90 wins, Washington being the only exception.
- Fielding percentage is being evaluated too highly in the data set. The deviation from the best defensive team to the worst was only .008% (Astros .989 and White Sox .981), yet it is being graded at 10% of the weighted composite score. In future data sets, I will need to weigh it less severely and allocate the free percentage toward other metrics.
- Findings may at times be spurious due to some metrics being part of others (OPS and BA).
- Composite scores are built off rankings from other statistical categories, the deviation between teams may be exaggerated based off actual performance in each category, but the ranking of teams remains accurate because they are all being evaluated on the same scale.
A Better Way?:
This is by no means the best way to stratify teams, as I highlighted in the previous article, regression analysis can tell us more about which of these metrics is both statistically significant and impactful (coefficient value). Previous baseball analyses have largely done this work for us, identifying the value of these metrics which has become a staple in player analysis but rarely used to evaluate overall team performance. To improve upon the CGMs Stratification System, a regression identifying the impact of each variable on team performance would help in better weighing the metrics to hone in on a more accurate composite score. Until then, this methodology is among the few to evaluate team performance outside of team record.