By: Joe Botana
Proverbs 11:14 New American Standard Bible (NASB)14 Where there is no guidance the people fall,
But in abundance of counselors there is [a]victory.
Proverbs 15:22 New American Standard Bible (NASB) 22 Without consultation, plans are frustrated,
But with many counselors they [a]succeed.
Face it, you never thought you’d see the day where you turned to a baseball blog to read a story about projected divisional standings for the 2016 MLB season, only to encounter a couple of verses from the Bible, and the Old Testament to boot. Trust me, there is a reason for their inclusion, for they illustrate the wisdom which underlies the predictive model upon which this upcoming series of posts is based.
When a group of experts, real or imagined, are called upon to predict the future outcome of an uncertain event, their forecasts will often deviate one from the other. Sometimes the deviation will be small, yet other times they can be significant. Factors which influence the spread include the degree of uncertainty affecting the outcome, the predictive data available, and the assumptions utilized in each forecasting model.
When attempting to forecast the outcome defined as the relative finishing positions in each of the six MLB Divisions, before the start of the season, we are dealing with events with high degrees of uncertainty, a wealth of data with varying degrees of predictive value, and the potential for much variability in the assumptions used in various predictive models. For this reason, different respected models will produce highly divergent forecasts. For example, the Baseball Prospectus PECOTA Model has the Pittsburgh Pirates finishing second in the NLCD with 82 wins, while the USA Today model shows them winning 88 games but finishing third in the Division.
For this series of articles, we’ve borrowed from the ancient wisdom of Proverbs, and developed a consensus forecast model to predict how each of the thirty MLB teams will finish in terms of their won-loss records, and how they will stack up within their respective Divisions. The idea is simple – using four models will smooth out the imperfections in each to produce a forecast which, on the whole, should be better than any individual one. Or, as Proverbs teaches, success comes from drawing from the collective wisdom of multiple counselors.
Consensus Forecasts in Politics and Finance
The practice we are applying here is already in common usage in the arenas of politics and finance. During this highly contentious election season, multiple polls are taken in each state where races will soon be held. Often, these polls show significantly varying results. The web site “RealClearPolitics.com” maintains a rolling average of the polls in each race and displays the “RCP Average” for the race. Often, this is uncannily predictive of the final results.
In finance, we often see consensus estimates of what the quarterly earnings will be for individual publicly traded companies, compiled from the individual estimates of analysts who follow those companies. We also see consensus estimates for such things as how many times the Fed will raise rates during the year, or what the price of WTI petroleum or unleaded gasoline will be, based on the individual estimates of econometric forecasters who follow such things. Again, these consensus estimates often turn out to be uncannily close to the eventual future reality.
The PLA Consensus Model of MLB Baseball Team Record Forecasting
For our model, we’ve selected four established and respected existing models which are commonly available to the general public. They are:
- Baseball Prospectus – PECOTA
- Scout MLB
- USA Today – Federal Baseball
- Bleacher Report
Our methodology is simple and straightforward. Each model is weighed equally. The wins, losses, and divisional finishing position from each model are input into our model for each team. We then calculate the consensus wins, losses, and finishing position in their Division for each team.
Our forecast is unbiased and purely quantitative based on the four models. We re-rank the teams in their divisions based on the consensus wins and also based on consensus position if different. We also compute the standard deviation of the total wins and divisional finishing position of each team. This serves to illustrate the degree of agreement or disagreement in the individual forecasts. Finally, we also compute the consensus expected strength of each division, based on the combined total of expected wins, and the index of standard deviations for each division, meant to illustrate the degree of uncertainty among the experts about the likely outcome of each division.
Highlights from the Model
We will share detailed results by Division over the next several days. However, here are a few summary highlights from the PLA Consensus Model, designed to whet your appetites and keep you coming back.
Strongest and Weakest Divisions – The AL East is expected to be the strongest division in MLB, with 419.3 combined wins, or a 51.3% winning percentage. Meanwhile, the NL East is the weakest division, with only a 47.8% winning percentage and 387.5 total combined wins.
Best and Worst Teams – The Chicago Cubs capture this honor with a consensus 96 wins. The Atlanta Braves, with only 66.5 consensus wins, narrowly edge out their NL East Division mate Philadelphia Phillies at 67.3 wins and the NLCD Cincinnati Reds with 67.8 wins for the “Hall of Shame” trophy as lowliest finisher.
Most Disagreement on a Team Result Prediction – The other Chicago team, the White Sox, comfortably garners this prize, with a standard deviation index of 7.37 on total wins and a Divisional Placement index score of 1.83. It is telling that the consensus model projects them to finish 3rd in the ALCD, but that is the only position in which none of the individual models place them, ranking them instead as first, second, fourth, and last in the Division.
Most Unpredictable Division – The ALCD has by far the highest Division Standing standard deviation index, with a score of 7.3 compared to an average index score of 3.9. It also has the second highest Total Wins standard deviation index, with a score of 24.4 compared to the average score of 19.6. Accordingly, we would expect that when we look back at this in October, the model’s ALCD team position prediction is the one with most uncertainty and disagreement, and thus the one most likely to be wrong.
We will continue to report detailed projected results of the model over the next week, including the PLA Consensus Model predictions for all six divisions, as we count down the days until opening day for the MLB 2016 Season