Run 3 Hits 51 Trades: $888.20, a 51% Win Rate, and Why We're Still Down 11.2%

Fifty-one trades. The paper bot's third run keeps grinding, and this morning's log — 4:40 AM Central, October 21 — reads like a paradox worth unpacking:

A 51% win rate — the bot wins more trades than it loses — and yet the account is down more than eleven percent. That is the single most important number in this entire report, because it kills a myth: win rate is not profit. What matters is the average win versus the average loss, and ours are not balanced.

The receipts: wins and losers from the last session

Overnight the HIGH tier (our ignition-style new-token plays, $25 each, 1-hour exits) closed a fast-moving batch. The winners were real:

• KIMCHI +22.6% — closed for +$5.34 paper.
• MONEYBIZ +27.6% — closed for +$6.57 paper.
• KYRO +8.3% — closed for +$1.82 paper.
• MONEYBIZ +14.6% / +14.7% — two closes for +$3.37 / +$3.39 paper.

And the losers, which are the honest reason the account is red:

• MONEYBIZ −39.6% — closed for −$10.05 paper.
• Pokecity −25.5% — closed for −$6.56 paper.
• KIMCHI −29.1% — closed for −$7.45 paper.
• Pokecity −17.1% — closed for −$4.49 paper.

See the pattern? The wins cluster in the +$2 to +$7 range; the losses cluster in the −$4 to −$10 range. One −39.6% trade wipes out two +27% trades and change. This is the signature of trading brand-new tokens: the volatility that creates the winners is the same volatility that deepens the losers. Our universal 1-hour exit cuts the bleeding — without it, the bake-off data shows 24-hour holds median −68% — but a fast exit can't make average wins bigger than average losses.

The fee drag nobody sees

$25.71 in modeled fees across 51 trades. That's 2.6% of the starting account, quietly taxed out of every entry and exit. On tiny $25 positions, fees are not rounding errors — they're a headwind the strategy has to beat on every single trade.

What 51 trades actually prove

Three runs, and the lesson keeps repeating: this style of trading has a positive win rate and a negative expectancy. That is publishable, repeatable information — the exact opposite of a backtest cherry-pick. The auto-analyst keeps logging it, the guardrails keep the run alive, and the next question is the hard one: can entry quality improve enough to flip the average, or does this corner of the market simply not have positive expectancy? We'll keep publishing until we know.