CreditFloor publishes one trade at a time: the single highest-conviction put credit spread on a highly liquid stock (≥$250M/day — deep, fillable options markets), machine-selected and published only when it clears a bar validated at 96.9% accuracy and ~24.7% return-on-risk per trade on untouched 2019–2026 data. No firehose — it fires only when a genuine setup appears, about once every 2–3 weeks. Every pick is a real, verified contract (actual strikes, expiry, live bid/ask, collectible natural credit), held to expiry, logged forever wins and losses alike. Weekly cadence at this accuracy is impossible — forcing more trades drops below 95% — so the pick waits for conviction. Not financial advice.
The one trade, right now. Sell the put spread shown, hold to expiry; it wins if the stock closes at or above the short strike. Published only when the highest-confidence, high-liquidity setup clears the 96.9%-validated bar — otherwise nothing, which is most days. When a pick also clears the stricter Elite (≈99%) band — top-tier model confidence, which went 73 for 73 in backtest (~10/year) — it’s marked with a gold badge.
Every Conviction Pick the frozen rule would have made on the same high-liquidity names it trades in production, over the untouched 2019–2026 validation window (one per week when the bar was met, conservative fills, held to expiry). This is the backtest; the live record starts accruing now and is shown further below.
| Year | Picks | Losses | Win rate | P&L (1 contract/trade) |
|---|
Every backtested Conviction Pick (newest first) — actual strikes, expiries, model confidence, credits and outcomes:
| Published | Ticker | Short strike | Long strike | Expiry | Credit | ROR | Conf. | Close@exp | Result |
|---|
Every Conviction Pick the frozen rule produced, back to 2019, resolved at its actual expiry close — so the record isn’t empty while the forward run is young. Picks marked BACKTEST are the frozen rule (model fit only on 2008–2018) applied to real market history; picks marked LIVE were published here in real time before their outcome was known and append going forward. Same rule, same strikes — only the clock differs.
| Published | Kind | Ticker | Short strike | Spot at publish | Expiry | Close at expiry | Result |
|---|
One put credit spread at a time: sell a put near the money-safe distance below the stock, buy a further put 2.5% of spot below it (that long leg caps your loss), same expiry, and hold to expiry. It wins if the stock closes at or above the short strike on the expiry date. Maximum loss is the width between strikes minus the credit — known and capped the moment you open it.
Candidates are restricted to stocks trading at least $250M/day in dollar volume — roughly the top ~450 U.S. names, with deep, continuous options markets. Every published pick is then snapped onto the real live chain: an actual listed expiration, real listed strikes, both legs with ≥25 contracts of open interest and a tight short-leg bid/ask, priced at the natural credit (sell the bid, buy the ask) — the fill available without negotiating. If it isn’t a real, fillable contract, it isn’t shown.
A systematic scan of 1.5M+ historical candidate trades against 14 causal technical features found one edge that replicates out of sample: selling premium when a stock’s volatility is elevated versus its own long-run level and already calming (the post-spike “vol-crush” regime) wins far more often than the market prices in. A gradient-boosted model over those features — fit only on 2008–2018, then frozen — scores every candidate. Each week, the single highest-confidence high-liquidity put spread is published only if it clears a bar set at the design 97th-percentile of confidence.
On untouched 2019–2026 data, picks clearing that bar validated at 96.9% accuracy and ~24.7% return-on-risk per trade, about 21 per year (one every 2–3 weeks). A stricter Elite band (top-tier confidence) validated even higher — 73 for 73, 0 losses, ~10 per year — and those picks are marked with a gold badge. Weekly cadence above 95% is impossible: forcing a pick every week drops accuracy to ~92%, so the strategy waits for conviction rather than lowering the bar.
Five kinds of exit management were tested with the same design/ validation discipline: fixed stops, fixed take-profits, GTC resting entries, GTC exits, and a fully learned exit policy. Every one reduced both win rate and P&L; the learned policy, given eleven years of examples, converged to never exiting. At two-week tenors an exit pays slippage, commissions, and the market’s already-repriced threat. Hold to expiry is the optimum.
The validated 96.9% (and the Elite band’s 73/73) are estimates from one carefully de-biased history on today’s surviving liquid universe, with conservative modeled fills — not a floor and not a promise of 100%. The trades that lose do so on exactly the events no price-history rule can rule out in advance: systemic crashes, M&A pops, earnings collapses. Defined-risk spreads cap each loss at width minus credit. Size to survive the worst trade (backtest worst: about −$160/contract), and assume the future can be worse than the past. Not financial advice.
This page is research output, not investment advice, and not a solicitation. Options involve substantial risk and are not suitable for all investors. Estimated credits, losses, and returns are model outputs from conservative assumptions, not quotes; real fills will differ. Backtest and replay results are hypothetical, computed on today’s surviving universe, and do not guarantee future results. The live log is append-only and includes every signal ever published on this page, including all losses. Past performance does not predict future results. Do your own research.