The ESPN Fantasy Forecaster is a comprehensive weekly post chocked full of information to help fantasy baseballers in weekly leagues. It is penned (typed?) by FSWA Hall of Famer Tristan Cockcroft who, incidentally, is our favorite ESPN fantasy baseball writer and someone very well-liked and respected among experts.
Each week, the ESPN Fantasy Forecaster estimates the value of every start with a ‘Game Score‘ – a metric devised by Bill James. It is unclear whether Tristan calculates the estimated Game Scores or this comes from someone else at ESPN.
This post is a head-to-head test of our Streamonator vs the ESPN Fantasy Forecaster to see which is better at predicting pitcher starts/stream values.
Note: To see how well Streamonator predicts on a stat by stat basis, please see our Ombotsman.
Overview
| ESPN Fantasy Forecaster | Razzball Streamonator | |
| Update Frequency | Weekly (before the week starts), updated occasionally prior to the start of Monday’s games | Daily |
| Metric For Valuing Start | Bill James’ Game Score | Auction $ (based on 12-team 5×5- $260/team) |
| Other Stats Projected For Starts | n/a | IP, W, L, QS, K, H, BB, HR ERA, WHIP |
| Notes Provided | Yes, a lot of useful notes on pitchers/matchups. | None (just stats) |
| % of Starts Where Correct Starter Is Projected* | 80% | 98% |
| * Based on a test of May 1 through June 29 starts (1,646 total MLB pitcher starts). Note that the Streamonator is updating probable pitchers daily so it should be projecting the correct pitcher more often than ESPN. ESPN also publish Game Score projections in their ‘Daily Notes’. I assume the Daily Notes project the correct SP at a similar rate as Streamonator. | ||
Test Methodology
- Identify all starts where ESPN Fantasy Forecaster and Razzball Streamonator both projected the correct pitcher (1,332 starts out of 1,646 starts)
- Run correlation tests for the ESPN and Razzball values against actual Game Scores and our $ estimates
Test Results
- The Razzball Streamonator is a better predictor of pitcher start values than ESPN Fantasy Forecaster. This is seen using either Game Score or our $ estimates and is true for each month individually and the two months combined.
| Correlation To (r) | ESPN Fantasy Forecaster | Razzball Streamonator |
| Actual Game Score (May/June/Both) | 0.139 / 0.200 / 0.172 | 0.230 / 0.255 / 0.243 |
| Actual Game $ Estimate (May/June/Both) | 0.158 / 0.198 / 0.180 | 0.245 / 0.241 / 0.243 |
Notes
- Click here to see the Google Doc with the 1,332 matched starts.
- While it may appear odd that our $ estimates are better predictors of Game Score than ESPN Fantasy Forecaster’s Game Score projections, it should be noted that Game Score and our 5×5 $ estimates are quite similar – correlating at 0.94. They both credit/measure Hits, BB, ER, K’s and IP. Game Score also factors in unearned runs while our $ estimates credit Wins (the biggest difference between the two metrics)
- The Fantasy Forecaster’ accuracy jumps between May and June where the Streamonator is relatively steady. It could be that Fantasy Forecaster is putting a greater weight on season-to-date performance where Streamonator – with considerable help from Steamer – is weighting in-season performance proportionally with past performance.
- I did a smaller test (148 starts between June 24th to June 29th) comparing ESPN’s Daily Notes Game Score projections to ESPN Fantasy Forecaster and Streamonator. Surprisingly, the Daily Notes (0.074 for $, 0.071 for Game Score) did much worse than the Forecaster (0.205/0.202) for the 6 days. Streamonator came in at 0.272/0.284. I have no idea why the ‘day of’ projections perform so much worse than those estimated before the week starts.
