V6.9.1 is closed: no edge confirmed. The price-band idea didn’t replicate.
V6.10 now uses pre-trade microstructure to avoid toxic fills.
Control: -$10.36
Filter: -$9.96
Delta: +$0.40
Both red. Only 11/200 candidate fills. We keep collecting 🧐
V6.9.1 update:
All 3 arms are red.
A — Control: -$14.52
B — Toxic-band filter: -$6.56
C — Execution-aware model: -$6.55
B and C are roughly +$8 vs control on the same markets.
Promising damage reduction. Absolutely zero proof of edge yet 😖
We’re no longer trying to predict the market harder.
V6.9 is running 3 paper strategies on the same Polymarket BTC 5m feed:
A — V6.8 control
B — rejects the toxic $0.61–$0.70 zone
C — execution-aware EV model
The question: why does the market choose to fill us? 😵💫
I’m shocked by the difference between GPT-5.6 SOL on Very High and ULTRA
I expected better reasoning. I didn’t expect it to feel like a completely different model.
ULTRA is expensive. But holy shit, it cooks 😮
Today the market punched back.
11 settled fills: 6W / 5L
PnL: -$8.94
ROI on filled notional: -49.4%
V6.8 is still +$11.71 across 94 settled markets.
No cope. No mid-run tweaks. An edge is a distribution, not a straight green line.
We keep collecting 😃
A lot of people think they have an edge on Polymarket.
What they actually have is a small sample, overfitted rules, and imaginary fills.
The market hasn’t paid them yet. Their spreadsheet has 😴