EPIC-MRA
Graded against the actual result across 38 races (from 98 polls, through 2024).
Head-to-head vs the VotePredictor model
The fair, apples-to-apples test: on the 26 races EPIC-MRA actually polled, how its final poll's margin compared to what the VotePredictor model predicted for those same races.
| Model | Avg miss (pts) | Called right |
|---|---|---|
| EPIC-MRA | 4.25 | 89% |
| VotePredictor | 4.12 | 92% |
VotePredictor aggregates all the pollsters, so it's expected to beat any single one on margin — that's the value of averaging. The honest comparison among forecasters is on the combined board.
Every race (38)
Each race EPIC-MRA polled, scored on its final poll — the call right before the vote — against the actual Dem−Rep result. Click a race for its full detail.
| Race | Their call | Result | Miss | Call |
|---|---|---|---|---|
| 2024 MI President | D+3.0 | R+1.4 | 4.4 | ✗ |
| 2024 MI Senate | D+5.0 | D+0.3 | 4.7 | ✓ |
| 2022 MI Governor | D+11.0 | D+10.5 | 0.5 | ✓ |
| 2020 MI President | D+7.0 | D+2.8 | 4.2 | ✓ |
| 2020 MI Senate | D+5.0 | D+1.7 | 3.3 | ✓ |
| 2018 MI Governor | D+5.0 | D+9.6 | 4.6 | ✓ |
| 2018 MI Senate | D+7.0 | D+6.5 | 0.5 | ✓ |
| 2016 MI President | D+4.0 | R+0.2 | 4.2 | ✗ |
| 2014 MI Governor | R+2.0 | R+4.1 | 2.1 | ✓ |
| 2014 MI Senate | D+15.0 | D+13.3 | 1.7 | ✓ |
| 2012 MI President | D+6.0 | D+9.5 | 3.5 | ✓ |
| 2012 MI Senate | D+21.0 | D+20.8 | 0.2 | ✓ |
| 2010 MI Governor | R+18.0 | R+18.2 | 0.2 | ✓ |
| 2010 IN-2 House | D+5.0 | D+1.3 | 3.7 | ✓ |
| 2010 IN-7 House | D+13.0 | D+21.1 | 8.1 | ✓ |
| 2010 MI-1 House | R+2.0 | R+11.1 | 9.1 | ✓ |
| 2010 MI-15 House | D+17.0 | D+16.7 | 0.3 | ✓ |
| 2010 MI-3 House | R+9.0 | R+22.2 | 13.2 | ✓ |
| 2010 MI-7 House | D+6.0 | R+4.8 | 10.8 | ✗ |
| 2010 MI-9 House | D+5.0 | D+2.5 | 2.5 | ✓ |
| 2010 IN Senate | R+18.0 | R+14.6 | 3.4 | ✓ |
| 2008 MI President | D+12.0 | D+16.5 | 4.5 | ✓ |
| 2008 MI Senate | D+18.0 | D+28.8 | 10.8 | ✓ |
| 2006 MI Governor | D+9.0 | D+14.1 | 5.1 | ✓ |
| 2006 MI Senate | D+13.0 | D+15.7 | 2.7 | ✓ |
| 2004 MI President | D+6.0 | D+3.4 | 2.6 | ✓ |
| 2002 MI Governor | D+9.0 | D+4.0 | 5.0 | ✓ |
| 2002 MI-11 House | R+15.0 | R+17.5 | 2.5 | ✓ |
| 2002 MI Senate | D+30.0 | D+22.7 | 7.3 | ✓ |
| 2000 MI-8 House | D+1.0 | R+0.1 | 1.1 | ✗ |
| 2000 IL President | D+15.0 | D+12.0 | 3.0 | ✓ |
| 2000 MI President | D+1.0 | D+5.1 | 4.1 | ✓ |
| 2000 OH President | R+2.0 | R+3.5 | 1.5 | ✓ |
| 2000 PA President | D+18.0 | D+4.2 | 13.8 | ✓ |
| 2000 MI Senate | R+5.0 | D+1.6 | 6.6 | ✗ |
| 2000 OH Senate | R+21.0 | R+24.1 | 3.1 | ✓ |
| 2000 PA Senate | R+4.0 | R+6.9 | 2.9 | ✓ |
| 1998 MI Governor | R+32.0 | R+24.4 | 7.6 | ✓ |
Accuracy by time to election
Lower is better. Time to election runs right (election week) to left (~2 months out).
By the numbers
| Time to election | Polls | Avg miss | vs field | Called right |
|---|---|---|---|---|
| ≤1 wk | 5 | 5.45 | +1.17 | 60% |
| 1–3 wk | 45 | 4.56 | -0.51 | 82% |
| 3–6 wk | 26 | 6.17 | +0.43 | 81% |
| 6–9 wk | 22 | 5.49 | -0.66 | 86% |
vs field is this pollster's average miss minus all pollsters' at the same lead time — green beats the field, redtrails it. Our historical polls reach ~2 months out; earlier polling isn't in the record.
Track record by cycle — getting better?
| Year | Polls | Avg miss | Lean (house effect) |
|---|---|---|---|
| 2000 | 18 | 7.0 | R+4.2 |
| 2002 | 10 | 6.7 | D+6.7 |
| 2006 | 11 | 4.7 | R+4.1 |
| 2008 | 7 | 5.9 | R+5.9 |
| 2010 | 14 | 5.3 | D+1.5 |
| 2012 | 7 | 3.5 | R+3.4 |
| 2014 | 6 | 2.7 | R+1.4 |
| 2016 | 4 | 6.5 | D+6.5 |
| 2018 | 4 | 5.8 | D+2.7 |
| 2020 | 8 | 4.5 | D+4.5 |
| 2022 | 4 | 1.7 | D+1.7 |
Do we credit a pollster for fixing its bias? Each cycle, the model re-estimates every pollster's lean from all its earlier polls (walk-forward) and subtracts it before using the poll. We tested weighting recent cycles more — it doesn't help: a pollster's lean in one cycle barely predicts the next (correlation 0.28), so the swings above are mostly noise, and averaging over more history beats chasing the latest cycle. The all-time estimate we use came out within ~0.5% of the best option.