Strategy Library
Playbooks bound to the live record. Every number on this page is rendered from /proof when you load it, and every card stands down automatically the moment its cell drops out of the proven set. Nothing here is a backtest.
How this library works
Every playbook is bound to the live forward record. Nothing on this page is a backtest, a static configuration, or an opinion — a card exists only because the measured record supports it, and it withdraws itself the moment the record stops doing so:
- Live-gated. A playbook is tradeable only while its signal cell passes the public gate (positive excess vs BTC over the same window, 95% CI lower bound above zero, sufficient sample). Each card checks the gate when the page loads and flips to stand down on its own.
- Measured exits only. Published stats are measured holding to the stated horizon — the only exit the record currently supports. TP/SL ladders publish when the live exit corpus is large enough to defend them, and not before. Frame risk with /levels zones and /check's stop-geometry flags.
- Dated behavior notes. Prose findings cite the forward ledger through 2026-07-17 (entries since 2026-06-03, the current clean-measurement floor). Live numbers in the stat blocks supersede them.
Security-event fade short the named asset
Thesis. When a hack / exploit signal fires on an asset, the market reprices down over the first day. The measured edge lives in the FIRST 24 HOURS — the forward ledger shows the move more than retraces by day 7, so this is strictly an intraday-to-24h play, never a swing short.
Entry. A hack signal on a named asset with conviction “strong” (free tier sees it 15 min delayed; Pro real-time). Corroborated events (corroborated: true) are the cleaner subset.
Exit. 24h — close at the horizon. The published stats are measured holding exactly that long.
What the ledger measured. Ledger through 2026-07-17 (entries since 2026-06-03): +0.48% avg excess vs BTC at 24h (n=255, t=2.4). Held to 7 days the same entries measure −2.2% excess (n=139), and −4.2% at 14 days — the edge does not just decay, it reverses. Exit discipline is the strategy.
agent prompt for this playbook
Watch n0brains for hack signals. API: https://api.n0brains.com, header X-API-Key: YOUR_API_KEY
Every 5 minutes: GET /signals?signal_type=hack&urgency=high&limit=5
Rules:
1. Act only on conviction "strong", and only while /proof lists hack as
a proven type (GET /proof, check proven_horizon on the hack row).
2. Direction: bearish on the named asset. The measured edge is the FIRST
24 HOURS - plan the exit at the 24h mark. NEVER hold past 48h: the
forward ledger shows the move retraces and reverses by day 7.
3. Frame the stop with GET /levels/<COIN> (nearest resistance zone above).
4. Read me the signal, the plan, and the invalidation. Analysis only,
not financial advice - I place my own trades.Deleverage rebound long the flushed asset
Thesis. After a leverage flush (cascade of forced liquidations), price tends to snap back once forced sellers are done. This is the fastest cell in the library — the gate proves it at the 4-hour horizon.
Entry. A deleverage signal with bullish direction and conviction “strong”. These fire after the cascade is measured, not during the knife.
Exit. 4h — this cell is proven at four hours, not a day. Take the bounce and be done.
What the ledger measured. Ledger through 2026-07-17: +0.26% avg excess vs BTC at 4h and 64% win rate (n=59 since 2026-06-03, t≈2.0 at both 4h and 24h). Small but consistent sample — the gate sizes the claim accordingly.
agent prompt for this playbook
Watch n0brains for post-cascade rebounds. API: https://api.n0brains.com, header X-API-Key: YOUR_API_KEY
Every 5 minutes: GET /signals?signal_type=deleverage&limit=5
Rules:
1. Act only on bullish direction + conviction "strong", and only while
/proof lists deleverage as proven (GET /proof, proven_horizon set).
2. The proven horizon is 4 HOURS. Plan entry immediately on signal and
the exit at the 4h mark - this is a bounce, not a position.
3. Frame the stop below the nearest support zone: GET /levels/<COIN>.
4. Report signal, plan, invalidation. Analysis only, not financial
advice - I place my own trades.Whale-position follow follow bullish prints only
Thesis. Large on-chain / exchange position builds by tracked whales lean the market their way over the next day. The edge is asymmetric: the ledger backs FOLLOWING BULLISH builds. Bearish whale prints measure flat-to-negative as shorts — treat them as context, not entries.
Entry. A whale_position signal, bullish direction, conviction “strong”. Skip bearish prints entirely as trade triggers.
Exit. 24h — proven at the one-day horizon, held to the horizon.
What the ledger measured. Ledger through 2026-07-17: bullish side +0.11% excess at 24h with 57% win rate (n=687, t=2.2); the bearish side measures −0.16% at 24h (t=−2.9) and degrades brutally at swing horizons (−6.6% excess at 7d). Following only the bullish half is what the record supports.
agent prompt for this playbook
Track whale positioning with n0brains. API: https://api.n0brains.com, header X-API-Key: YOUR_API_KEY
Every 10 minutes: GET /signals?signal_type=whale_position&limit=10
Rules:
1. Trade triggers = bullish direction + conviction "strong" ONLY, and
only while /proof lists whale_position as proven. Bearish prints are
context - mention them, never short them (measured anti-edge).
2. Horizon: 24 HOURS, exit at the mark.
3. Frame stops with GET /levels/<COIN>; sanity-check the whole idea with
POST /check before flagging it to me.
4. Analysis only, not financial advice - I place my own trades.What the record says not to do
The same ledger that proves the playbooks above measures these as negative-edge behaviors. Knowing them is worth as much as the playbooks:
- Chasing listings. Buying listing announcements measures ~
−0.3%to−0.5%excess at every horizon we track (and worse net of the pop you pay up for). - Buying funding squeezes. Bullish funding signals measure
−0.67%excess at 24h (n=476, t=−5.1). Crowded-long funding is a warning, not an invitation. - Holding event shorts. The hack fade's own entries flip to
−2.2%excess by day 7. Event edges are measured in hours, not weeks. - Shorting whale distribution at swing horizons. Bearish whale_position held a week
measures
−6.6%excess. The bullish half of that signal carries the edge.
Agents can pull the current inverted cells programmatically:
get_anti_predictive_cells over MCP — cells where the measured
edge points against the signal's own direction.
All figures: forward ledger through 2026-07-17, entries since 2026-06-03, vs same-window BTC baseline.
Methodology & what's coming
- Where the numbers come from. Every directional signal is measured against real post-signal prices at fixed horizons, vs a same-window BTC buy-hold baseline, cost-adjusted — the identical pipeline behind /proof. No curation, no survivor bias; weak types are demoted publicly.
- Why only three playbooks. The gate currently proves three cells. When another cell earns its way in — and holds — it gets a card. When one drops out, its card stands down automatically rather than lying to you.
- Swing horizons (7d / 14d / 30d) are being measured now, but they are not yet gate-eligible: the corpus spans a single market regime and overlapping windows inflate significance. They'll graduate here when a fair test exists — not before.
- Exit ladders. A live walk-forward exit tuner (MAE/MFE on real fills) is accumulating data. If it ever produces an exit rule that beats hold-to-horizon out of sample, it will be published here with its record — until then, hold-to-horizon is the honest exit.
Everything on this page is analytical market data on public markets — not financial advice, not a recommendation, and not tailored to you. Signals can and will be wrong. Size for survival; keep a human on the trigger. See terms §8.