Run 3 Hits 30 Trades: -$118.91, 12 Wins, 17 Losses, and the Honest Math

Thirty trades. That's a real sample — big enough to stop arguing about single wins and start looking at the machine. So here is the machine, exactly as the bot's own trade log reports it this evening, October 20, 2026:

A 41% win rate and an 11.9% drawdown. Nothing here is being rounded up or re-framed. This is paper trading — no real money — and the whole point of running it in public is that the losing numbers teach more than the winners.

The daily guard fired again

Today's session also tripped the −4.7% daily loss guard: no new entries for 24 hours. The kill-switch ladder (−3% daily pauses new entries, −5% halts for a day, −15% total stops everything) exists precisely because of runs like this one. Discipline is the strategy's second edge; it's currently doing more work than the first edge.

The bake-off behind it

Every signal the bot trades is tracked in a three-way strategy bake-off — ignition momentum, phoenix mean-reversion, and boost attention arbitrage. As of this evening's data (47 signals tracked):

• Ignition (momentum): 34 signals resolved at 24h, average return −26.7%, only 2 winners over +50%.
• Phoenix (mean-reversion): 3 resolved, average −40.2%, 0 winners.
• Boost (attention arbitrage): 36 resolved, average −30.0%, 2 winners over +50%.

Across all signals, the average 1-hour return is −9.1% and the average 24-hour return is −27.8%. The universal 1-hour exit rule (median −16.0% vs −68.0% for 24h holds) cuts the bleeding in half, but a faster exit can't fix entries that start underwater. That is the honest headline: exits are solved; entries are not.

What happens next

The run continues under the daily guard. The next ten signals will either show entry quality improving or give the auto-analyst enough evidence to recommend a strategy change — and per our rules, any high-conviction change gets flagged to Paul before anything moves. We'll keep publishing the scoreboard either way.