Been running the weather strategy for about 10 days on default settings, ~$5 per trade across ~15 cities. I'm a bit confused: my win rate is only around 18%. That feels really low — I'm used to thinking a good system should win most of its trades. Yet my balance is slowly going up, not down. Am I doing something wrong, or is a sub-20% win rate actually normal for this bot? Should I tighten the filters to win more often?
An ~18% win rate is exactly how this strategy is designed to work — it's the point, not a problem. Let me explain why.
The bot runs in UNDERPRICE mode: it buys YES options that the multi-model ensemble (GFS + ECMWF + ICON + NOAA) believes are worth more than the market is charging. By default it only enters when all of these hold:
- Edge (ensemble fair value − market price) is at least 6.5%
- Market price is between 6¢ and 18¢
- Confidence is HIGH (strong model agreement)
Because we deliberately enter options priced at 6–18¢, we're taking positions on outcomes the market prices as unlikely (6–18% implied). Even with our edge the true probability is only ~15–25%, so most individual positions don't resolve our way — by design.
Why it's profitable anyway — the payoff asymmetry:
- A YES option entered at 18¢ that resolves true pays 100¢ → 5.5× the capital at risk.
- At 10¢ → 10×. At 7¢ → 14×. At 6¢ → 16×.
- When a position goes against you, the Stop Loss (default 70%) caps the loss before the option bleeds to zero.
The break-even math: at the most expensive entry we allow (18¢), with the 70% stop loss, you only need about a 13% win rate to break even. The strategy targets options whose true probability is ~20–25%. So an 18–20% realized win rate sits comfortably above break-even — that's a winning system, not a broken one.
This is why chasing a "high win rate" is the wrong instinct here. To win 70–80% of trades you'd have to enter expensive, near-certain options (70–90¢) where a winner returns only 1.1–1.4× and a single adverse resolution wipes out many winners. That inverts the edge. Low win rate + large asymmetric payoff + capped downside is the entire thesis.
The settings that actually matter (Settings panel): Min Edge (raise for fewer, higher-quality entries), Min/Max entry price (the 6–18¢ band — lowering the max raises your return per winner), Stop Loss, and Confidence. The bot is edge-driven, not favorite-driven — there's no "win more often" dial.
This clicked for me after a losing-looking week that was actually green. Quick example from my own log: out of 10 trades, 8 hit the stop loss and 2 resolved in my favor. Sounds terrible. But the 2 winners were entered around 9–11¢ and resolved at 100¢ — roughly 10× each. The 8 stop-losses were small (capped at 70%). Net for the cycle was clearly positive. The "win rate" number is the most misleading stat in this bot — watch P&L and ROI, not win %.
That completely reframes it — thanks. So if anything I should be looking at the entry price cap, not the win rate. Does lowering the max entry price (the 18¢ cap) improve things, or just reduce the number of trades?
Exactly the right question. The max entry price is a direct return-vs-frequency dial:
- Lower it (e.g. 12–14¢): every winner pays more (7–8×+) and your break-even win rate drops — more cushion. The cost is fewer entries per day, since fewer options fall in the lower band with enough edge.
- Raise it (toward/over 20¢): more entries, but each winner pays less and your break-even win rate creeps up toward your actual win rate — the margin of safety shrinks. We don't recommend going above 18¢.
The 6–18¢ default band is the balance we ship for a reason. If you want to lean on quality, drop the max to ~14¢ and/or raise Min Edge to ~0.08, and accept fewer trades. Change one dial at a time and judge it on ROI over a week, not win rate.
One more practical lever: the City Filter. It's not about "high win-rate cities" — it's about which cities the ensemble maps cleanly to a reliable observation station. Cities with a single well-defined airport station and stable model agreement give the cleanest edge. Leave it empty to trade everything the bot supports, or list specific cities (comma-separated) once you've seen which ones produce the most HIGH-confidence entries in your Pool.