Benchmark · public ledger
What we predicted. What actually happened.
Every week we publish how Simulix predictions compared to real-world outcomes — elections, ad campaigns, product launches, ballot measures. Wins and misses, both.
Each entry below links to the methodology version used, the raw simulation parameters, and (where applicable) the public dataset of the actual outcome.
| Date | Event | Prediction | Actual | Delta | Methodology | Notes |
|---|---|---|---|---|---|---|
| 2024-09-04 | Mid-market fast-casual chain — $14.99 → $16.49 entree test Product launches | 44% retain / 31% switch / 25% reduce frequency | 37% retain / 35% switch / 28% reduce frequency | +7.0pp | v1.1 | Read the postmortem |
| 2024-06-21 | NY-08 Democratic Primary 2024 Elections | Goldman 71% / Bowman 29% | Goldman 73% / Bowman 27% | +2.0pp | v1.2 | Read the postmortem |
| 2023-11-07 | Ohio Issue 1 — Reproductive rights amendment Public opinion | Yes 58% / No 42% | Yes 56.6% / No 43.4% | +1.4pp | v1.0 | Read the postmortem |
Methodology
On the misses.
When Simulix predictions miss the actual outcome by more than 7 percentage points, we publish a postmortem the same week. We don’t quietly remove the entry. We don’t update the methodology silently. The misses are part of the methodology — that’s the point.