The Tyson Group
Graded against the actual result across 14 races (from 27 polls, through 2020).
Head-to-head vs VotePredictor Elections
The fair, apples-to-apples test: on the 14 races The Tyson Group actually polled, how its final poll's margin compared to what VotePredictor Elections predicted for those same races.
| Model | Avg miss (pts) | Called right |
|---|---|---|
| The Tyson Group | 3.52 | 93% |
| VotePredictor Elections | 1.98 | 86% |
VotePredictor Elections 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 (14)
Each race The Tyson Group 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 |
|---|---|---|---|---|
| 2020 LA President | R+6.0 | R+18.6 | 12.6 | ✓ |
| 2016 FL President | D+3.0 | R+1.2 | 4.2 | ✗ |
| 2016 FL Senate | R+5.0 | R+7.7 | 2.7 | ✓ |
| 2014 FL Governor | R+3.0 | R+1.1 | 1.9 | ✓ |
| 2012 NC Governor | R+14.0 | R+11.4 | 2.6 | ✓ |
| 2012 NC-8 House | R+5.0 | R+7.8 | 2.8 | ✓ |
| 2012 NC President | R+4.0 | R+2.0 | 2.0 | ✓ |
| 2010 NC Senate | R+10.0 | R+11.8 | 1.8 | ✓ |
| 2008 NC Governor | D+2.0 | D+3.4 | 1.4 | ✓ |
| 2008 NC President | D+1.0 | D+0.3 | 0.7 | ✓ |
| 2008 NC Senate | D+2.0 | D+8.5 | 6.5 | ✓ |
| 2004 NC Governor | D+15.0 | D+12.7 | 2.3 | ✓ |
| 2004 NC President | R+8.0 | R+12.4 | 4.4 | ✓ |
| 2004 NC Senate | R+1.0 | R+4.6 | 3.6 | ✓ |
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 |
|---|---|---|---|---|
| 1–3 wk | 12 | 3.25 | -1.82 | 83% |
| 3–6 wk | 8 | 3.43 | -2.31 | 75% |
| 6–9 wk | 7 | 5.21 | -0.94 | 43% |
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) |
|---|---|---|---|
| 2004 | 3 | 3.4 | D+3.4 |
| 2008 | 14 | 4.1 | R+3.0 |
| 2012 | 3 | 2.5 | R+0.6 |
| 2016 | 4 | 2.9 | D+2.7 |
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.