Track Record
We track every contract to the end.
Every morning we surface about 50 contracts. We follow every one from that morning until it expires — nothing gets edited out. Here is what actually happens to them.
The win is real.
Between the morning we surfaced them and the day they expired, most of these contracts had a real moment:
were up +40% or better at some point
were up +63% or more at their best moment
doubled at some point
measured on all 1,554 contracts that have expired so far · updated daily
The win doesn't wait forever.
When did each contract hit its best price, counting from the morning we surfaced it?
Most best moments come early — but 1 out of 6 contracts doesn't hit its best price until more than two weeks in.
If you never sell, you lose.
Held all the way to the end, about 6 out of every 10 of these contracts finished nearly worthless — down 90% or more. The ride is violent too: the typical contract was at some point down more than 90% from where it started, even when it also had a big up moment.
We even tested buying every single contract with one fixed selling rule. It loses money (-5.2% per contract on average). We publish that on purpose — it's why we sell data, not picks.
The win is real, and it's temporary. Finding the contracts is our job. Selling at the right time is yours — or your agent's.
See the full data — charts and tables, for people and agents who want the detail
median best moment (peak premium return)
the top 10% of contracts peaked here or higher
median worst moment along the way
median value if held to expiration
Best moment before expiration
Each contract's best price after surfacing, vs the 10:00 ET surfacing price. Share of all 1,554 tracked contracts per bucket.
Value if held to expiration
The same pool, never sold — settlement value vs the surfacing price. Blue = loss side, gold = gain side. N = 1,499.
| bucket | contracts | share |
|---|---|---|
| <+5% | 277 | 17.8% |
| +5–20% | 182 | 11.7% |
| +20–40% | 170 | 10.9% |
| +40–70% | 183 | 11.8% |
| +70–100% | 130 | 8.4% |
| +100–200% | 233 | 15% |
| +200%+ | 379 | 24.4% |
| bucket | contracts | share |
|---|---|---|
| <−90% | 939 | 62.6% |
| −90–−50% | 64 | 4.3% |
| −50–0% | 85 | 5.7% |
| 0–+50% | 74 | 4.9% |
| +50–100% | 59 | 3.9% |
| +100–200% | 94 | 6.3% |
| +200%+ | 184 | 12.3% |
Peak figures are realized per-contract extremes — profit potential, not a return anyone earned; nobody sells at the exact top. Distributions cover 2026-04-10 to 2026-06-26 and exclude 878 contracts too illiquid to price cleanly at entry, 1,208 not yet expired or awaiting the nightly labeler, and 4 whose labeling failed — all counted, none hidden. Separately, a small paper-traded cohort exercises the engine's selection daily under fixed mechanical rules as a measurement instrument; we make no marketing claims from it.
Paper-trading data · Educational only · Not investment advice
Want to work this data?
Everything on this page — and the contract-by-contract detail behind it — is what your AI agent gets over MCP. Our Lab publishes the experiments we run on it, including the ideas that failed.
Paper-trading data, educational content only. Not investment advice. Excursion figures are realized per-contract extremes, conditional on the stated tracking windows — not returns achieved by any account. Past performance is not a guarantee of future results.