ohh.bet
AnalyticsTrack asymmetric-payoff whale trades on Polymarket.
We watch $25K+ whale BUYs at price 0.20–0.30 and hold them to actual market resolution. A 135K-trade historical backtest motivated this strategy; live calibration is what we publish on. The track record page shows everything — wins, losses, drawdowns.
The thesis
The cheap band has structural edge. Polymarket BUYs at price 0.20–0.30 carry an asymmetric (1−P)/P payoff: each win pays 3–5×, each loss costs 1×. Even at a 34% hit rate the math clears. Buys above 0.60 fail the same arithmetic even at 70%+ hit rates. This is the part of the thesis we have receipts for — a 135K-trade resolution backtest motivated it; live data is grading it.
Wallet reputation is descriptive, not predictive. We tested an "informed capital" filter (the wave-34 gate: proven wallets only) and the rep-alpha audit found it filtered AGAINST alpha. A U-shape refinement (wave 62) located the real sweet spot — wallets with modest prior PnL/$ (0–5%, ≥5 resolved). Paper portfolio D forward-tests that cohort, and the U-shape is currently replicating in live PnL (mild_winner 80% wins, reputation cohort 0%).
Full math, per-detector logic, and the live narrowing history live on /methodology. Running track record on /transparency.
How it works
We ingest every $1K+ trade in real time
Polymarket Data API polled every 10 seconds for fresh whale fills. ~5min upstream lag is a Polymarket characteristic, not ours.
Detectors fire on calibrated patterns
- Cheap Conviction — flagship. $25K+ buys at price 0.20–0.30, held to resolution. Currently the only detector firing daily.
- Arbitrage — cross-venue spread vs Kalshi after fees + slippage. Arithmetic edge, no modeling.
Four research detectors (fresh_whale, whale_cluster, volume_anomaly, copy_trade) run as data collection but have rare or zero recent fires — see /methodology for each one's trigger and current status. Three legacy detectors (smart_money, reversal, momentum) were calibrated as anti-predictive and removed from the worker in wave 64.
Three strategies trade in parallel as paper portfolios
Each portfolio is virtual $100K with proper risk management (max open notional, max market exposure, max drawdown auto-pause, stop-loss). Auto-traded by cron. Marked-to-market every 5min. Public leaderboard at /strategies.
You can take any signal manually
Every signal card has a "Take this trade" button that opens a position in your designated portfolio (default: M: Manual / Discretionary). Risk caps still apply — the system won't let you over-extend.
The honesty rule
Backtest motivated the thesis. Live data is the test. The 135K-trade backtest is historical context, not a forward promise — we narrow the band when live data disagrees, as waves 29 and 31 did. Current calibration lives on /transparency.
Arithmetic edges are real. Arbitrage's net edge is just spread minus fees minus slippage. No modeling.
When something doesn't work, we say so. Smart money, reversal, and momentum had over 5,000 calibration signals between them. The data showed they don't have edge on Polymarket. Wave 64 removed them from the worker; their historical fires stay visible on /transparency.
Heuristic multipliers are banned. No "confidence score" derived from a vibe. If we can't measure it, we don't display a number.