Market-making the Limitless prediction-market book with real money
We put real money into quoting both sides of a prediction market, aiming to earn the gap between the buy and sell price. Our orders did fill at good prices. The problem is what happens next: there is almost nobody to sell the position back to, so every fill has to be held until the market settles, where the outcome is close to a coin flip worth fifty cents a share against a gap worth under two. And the fills land on the wrong side, because on a falling market everyone sells into your buy order and nobody touches your sell order. We lost about a hundred and forty-five dollars in thirty-five hours and stopped. The deeper mistake was arming it with no maximum loss, no position cap and no exit plan, because the criteria we had written down were about statistics, not safety.
An earlier measurement was right that our quotes filled at good prices. The inference drawn from it was wrong. A good fill price only becomes profit if you can exit the inventory near fair value, and on these books you cannot. The whole top of the book is roughly one clip, the passive exit almost never completes, so every fill rides to settlement, where a binary contract swings about 50 cents a share against an edge worth about 1.6 cents. Worse, the fills arrive on the losing side by construction: on a falling market the crowd sells into your bid all the way down and nobody lifts your offer.
Armed 2026-08-09, clean two-sided from 20:35 UTC, halted 2026-08-11 at 07:13 UTC. Live loss about $145 on roughly $200 deployed, $138 marked at the halt decision, about $167 across the venue lifetime. Day-one fills ran 552 losing-side shares against 52 winning-side, better than ten to one, on a trending day. Depth was measured properly rather than from a single snapshot: 968 polls per symbol at 30-second cadence across five markets, median top-of-book capacity $43.10 on the deepest and $0.00 on the thinnest, with the tenth percentile near zero everywhere, against a clip of $30 to $60. Filled inventory then locks the capital: 8,529 quotes skipped for want of funds. Two order-path defects on day zero cost about $7.21 and were diagnosed and fixed the same day. The structural failure is the one that matters: the position was armed with no cumulative-loss line, no inventory cap and no exit path, because the kill criteria written in advance were purely statistical. Spec-lock is not a safety layer, and that is now a standing rule here. Honesty about the statistics: one day is about 1.3 standard deviations and is not proof of a negative edge. What is settled is the absence of a viable exit and the capital lock on fills. What is not settled is whether the fill edge survives to settlement across market regimes, and a zero-capital paper arm now measures exactly that against a metric and a read date both fixed in advance, on 2026-08-25.
- Kill date
- 2026-08-11
- Sample
- live, 35 hours
- Method
- Pre-registered live test
- Verdict
- fill edge is not exit edge
Pre-registered before the data, judged on a criterion locked in advance, and published whatever the result.
See all kills