Our Track Record

Why publish this

Any site covering prediction markets could cherry-pick the wins. We publish every market we wrote about once it resolves, because long-term credibility only comes from readers seeing the misses as well as the hits.

Format

Each entry shows:

• Race or market name
• Final Polymarket implied probability (last trading day)
• Final Kalshi implied probability where available
• Actual outcome
• Whether the favorite won (✓) or an upset occurred (✗)

Log

2024 cycle (retroactive entries)

Coming soon. We'll retroactively record every market we would have covered for 2024 presidential, senate, house, and governor races, using Polymarket's historical-prices endpoint to snapshot the final implied probabilities.

2025 cycle

Pending.

Current cycle

Each major market we cover will be added here on resolution. See our weekly analysis for ongoing commentary.

What counts as "right"

We measure calibration, not just majority-favorite accuracy. If Polymarket prices a race at 70% and the favorite wins, the market is not "right" — a 70% forecast is only well-calibrated if it's wrong 30% of the time across many races priced that way. A single outcome doesn't validate or invalidate a probability.

The honest metric is: across all markets priced near X%, does the outcome happen X% of the time on average? Published calibration curves for prediction markets suggest yes, within a few points, for liquid markets. We'll publish our own calibration curve once we have enough resolved markets to make it statistically meaningful.

Corrections policy

If we mis-state a final price or an outcome, email us and we'll correct with a dated note inline. We do not silently edit historical analysis.