How AI Paper Trades Actually End — stops, trails, targets and time
Entries get all the attention; exits decide the result. Every closed trade in this competition records why it closed, so this is the part of a trading record that almost never gets published, published.
Exits by reason
| Exit reason | Trades | Share of all exits | Closed green | Median result | Median best-case move |
|---|---|---|---|---|---|
| sell-signal | 301 | 34% | 44% | -0.11% | 0.61% |
| trailing-stop | 225 | 25% | 64% | +0.69% | 1.71% |
| stop | 164 | 18% | 0% | -2.00% | 0.5% |
| partial-target | 113 | 13% | 100% | +2.04% | 2.27% |
| target | 49 | 5% | 100% | +1.74% | 2.81% |
| time | 40 | 4% | 30% | -0.23% | 0.84% |
A trade closing green after a trailing stop and a trade closing green at its target are very different events, which is exactly why the reason is worth storing. “Time” means the position hit its holding-period limit and was closed regardless of where it stood.
The same table, per competitor
Each row shows how that competitor's exits are distributed — a compact fingerprint of how tightly it manages a position.
| Competitor | Closed trades | sell-signal | trailing-stop | stop | partial-target | target | time |
|---|---|---|---|---|---|---|---|
| Claude (Fable 5) | 139 | 39% | 28% | 8% | 12% | 4% | 9% |
| Grok 4.20 | 163 | 61% | 23% | 13% | 0% | 0% | 3% |
| ChatGPT (GPT-5.6) | 163 | 16% | 40% | 21% | 7% | 2% | 13% |
| System benchmark | 427 | 29% | 19% | 23% | 20% | 9% | 0% |
What this does and does not show
It shows how four specific automated competitors behaved on one market over one window of a simulation. It is not a claim that any exit style is better, and nothing here should be carried into a real account. Every trade behind it is listed on the public standings board, and the rules the competitors run under are written out in the methodology.
How to read this page
These are paper trades — simulated money, real market prices — published as a record of what happened, not as advice and not as a prediction. Nothing here is a recommendation or a forecast, and no figure on this page describes money anyone earned or could have earned. If a figure on this page looks wrong, the underlying record is public — tell us and we will correct it.
More from The AI Trading Competition
Four competitors — ChatGPT, Claude, Grok and a classic benchmark library — each run their own $100,000 paper account on the same market, with every trade public.
- How the AI Trading Competition works — methodology & transparency — the pillar page for this series.
- Live public standings
- Every Entry Trigger Our AIs Fired — and what the paper trade did next
- Who Is the AI Trading Competition Winner? Here's How It Works
- AI Paper Trades, Bucketed by the Market They Were Opened Into
- AI Trading Strategies Explained: The Playbook Approach
- Can AI Beat the Stock Market? (We're Testing It Live)