The wave-31 narrowing of cheap_conviction to [0.20, 0.30] plus the wave-34 wallet reputation gate predicts a fire rate of ~0.5 per day at current market conditions. Yesterday's 24-hour audit found something more uncomfortable: the entire 0.20–0.30 band had zero whale-sized buys for 48 consecutive hours. Not "zero that passed our gate" — zero candidates whatsoever, before we even applied a filter.

Here's the daily candidate count across the last 14 days, for $25K+ BUYs on YES/NO markets, by entry price band:

16 candidates in 14 days site-wide before our gate. After the wallet-rep gate filters ~50% of those, we'd see ~8 fires in 14 days, or ~0.5/day on average. With 0 days clustered around quiet markets.

Meanwhile the 0.80+ ("favorites") band has 805 candidates over the same 14 days. Whales overwhelmingly bet favorites on Polymarket. Not 2x. Not 5x. 50x.

Why this matters

The about page used to say "Backtest expectation: 2–5 fires/day." That was wrong by an order of magnitude. The backtest gave us the directional shape of the strategy — cheap is better than expensive at whale sizes — but it didn't tell us how rare the cheap entry would be in live markets.

That has two consequences worth being explicit about:

1. Calibration cycles are slow. With ~0.5 fires/day, accumulating 100 resolved trades takes ~200 days. To meaningfully test whether the wave-34 reputation gate improves the hit rate from the baseline 53% (live band, no gate) to the predicted 62%, we need 30–50 resolved trades at minimum. That's 2–3 months of waiting, not days.

2. The "live signal feed" framing was overpromised. A feed that fires 0–1 times per day is a feed in name only. We're rethinking the public surface around this.

What it tells us about the market

Polymarket's whale flow is structurally consensus-seeking. Big money chases markets that have already moved. The cheap band is rare for a reason — it's where bettors disagree most, where information asymmetry might actually exist, but where everyone's incentive structure says to wait until the price moves up. Whales rarely lead. They follow.

That's also the reason the cheap-band edge should exist in theory: it's where informed early-movers buy before consensus arrives. The pattern we're trying to capture is the rarest one. Empirically, when whales do buy cheap, the live hit rate is ~53%. That's the part of the original thesis that still holds.

What we're doing about it

This audit fed directly into a strategic decision documented in docs/superpowers/specs/2026-05-13-audit-transparency-pivot-design.md: ohh.bet's public-facing identity reorients around audit/transparency rather than live signals.

• The narrow cheap_conviction detector still runs and the gate still filters. That's the in-house high-conviction surface.
• The new public-facing entry point is the whale activity feed, showing all $25K+ BUYs across all bands, with our methodology badges per row. You see the broad market and the audit framework simultaneously.
• This /audits section is the depth layer — what we've learned, what we changed, what we got wrong.

The strategy may still produce alpha. It will just take longer to prove than we initially budgeted, because the market structure doesn't give us the candidate flow our backtest implied.

We owe you the receipt for that. This is it.

See the receipts: the backtest, after fees · stop-loss audit · live transparency