Eight ghosts. Two controls. Zero live money (yet).
Lab
Before any ghost places a real trade, each (profile × strategy) pair must accumulate enough paper-trade outcomes for its win-probability calibrator to mean something. Until then every signal is journaled, every gate is logged, and the sizer stays at its minimal bootstrap size — 1 share per stock signal, a small fixed notional per crypto signal. Stocks trade during market hours; crypto runs 24/7. This page is the live look at that bootstrap — updated every 60 seconds.
Articles classified
66,745
Catalysts logged
115,890
Active right now
121
Signals generated
482,643
Outcomes accumulated
1,722
Gated today
1,694
Stocks signals
10,641
📈
Crypto signals
472,002
₿
Active strategy instances
Phase 1 of the lab seeds one default-param instance per strategy. Each row below tracks one of those instances. Once an instance has 100 closed trades with PF > 1.1 and Brier < 0.22, it gets promoted into a ghost's live mix.
📈 Stocks
| Instance | Trades | Win% | PF | P&L | Status |
|---|---|---|---|---|---|
Hollow Prophet Catalyst Scalper #1f1a | 30 | 30% | 1.02 | +$5 | live |
Gilded Gambler Crypto/Equity Coupling #cbe1 | 0 | — | — | — | live |
Neon Mongoose Gap & Go #0e76 | 65 | 37% | 0.64 | −$227 | live |
Howling Wraith Mean Reversion (RSI/VWAP) #6c79 | 208 | 37% | 0.97 | −$64 | live |
Pale Reaper Momentum Breakout #35b5 | 266 | 38% | 1.00 | −$2 | live |
Pale Ghoul Political News #d431 | 98 | 24% | 0.98 | −$15 | live |
₿ Crypto
| Instance | Trades | Win% | PF | P&L | Status |
|---|---|---|---|---|---|
Midnight Phantasm Crypto Breakout #2155 | 15 | 87% | 12.09 | +$93 | live |
Ectoplasmic Gambler Crypto Momentum #e795 | 485 | 26% | 1.51 | +$1132 | live |
Lurking Duchess Crypto News #22aa | 4 | 0% | 0.00 | −$10 | live |
Feral Reaper Stablecoin Arbitrage #17a8 | 16 | 25% | 1.28 | +$13 | live |
Grim Vagrant Stablecoin Arbitrage #08d1 | 0 | — | — | — | live |
📈 Stock strategies
All six stock strategies are anchored here. Each needs 100 observed outcomes before the calibrator can produce a meaningful win-probability curve.
| Strategy | Pack | Signals | Outcomes | Win rate | Progress to calibration | Status | ||
|---|---|---|---|---|---|---|---|---|
| 🏛Political News | Political | 7050 | 98 | 24.5% | −$15 | 0.98 | 98/100 | calibrating |
| 🪞Mean Reversion (RSI/VWAP) | Core Technical | 1428 | 208 | 36.5% | −$519 | 0.75 | calibrated | ready for live |
| 📈Momentum Breakout | Core Technical | 917 | 266 | 37.6% | −$5 | 1.00 | calibrated | ready for live |
| 🌅Gap & Go | Core Technical | 841 | 65 | 36.9% | −$227 | 0.64 | 65/100 | calibrating |
| 🎯Catalyst Scalper | News-Driven | 116 | 30 | 30.0% | +$5 | 1.02 | 30/100 | calibrating |
| ₿Crypto/Equity Coupling | Crypto Macro | — | 0 | — | — | — | no signals yet | calibrating |
₿ Crypto strategies
Four crypto-native strategies running 24/7. Same calibration bar: 100 outcomes per (profile × strategy) pair before real sizing kicks in.
| Strategy | Pack | Signals | Outcomes | Win rate | Progress to calibration | Status | ||
|---|---|---|---|---|---|---|---|---|
| 📈Crypto Momentum | Crypto | 456399 | 578 | 29.8% | +$12 | 1.00 | calibrated | ready for live |
| ⚓Stablecoin Arbitrage | Crypto | 7406 | 17 | 29.4% | +$15 | 1.32 | 17/100 | calibrating |
| 🔔Crypto Breakout | Crypto | 5946 | 15 | 86.7% | +$93 | 12.09 | 15/100 | calibrating |
| 📰Crypto News | Crypto | 1550 | 4 | 0.0% | −$10 | 0.00 | 4/100 | calibrating |
⚖️ The fair trials
Two months in, the fleet's edge was statistically indistinguishable from zero-minus-friction. So the experiment grew a control group and a clean re-trial: every number below is measured against "no skill at all", on windows where the machinery actually worked.
🪙🐒 Fleet vs. the controls — since Aug 5, 2026
Coinflip and Dart Monkey trade at random with the fleet's exact sizing, costs and exits. If a track's ghosts can't out-earn their control per trade, their signals carry no information — friction is the whole story.
| Track | Cohort | Trades | Win rate | P&L | Avg / trade |
|---|---|---|---|---|---|
| ₿ Crypto | control | 384 | 40% | -$16.45 | -$0.04 |
| ₿ Crypto | fleet | 270 | 33% | +$650.35 | +$2.41 |
| 📈 Stocks | control | 57 | 46% | +$16.44 | +$0.29 |
| 📈 Stocks | fleet | 269 | 32% | -$290.46 | -$1.08 |
📰 The news-edge trial — clean window from Jul 28, 2026
The one place this system might have a real informational edge is LLM-classified news speed — and the pipeline feeding it was broken until Jul 28 (classifier fields dropped, catalyst RVOL gate dead). Lifetime stats measured the bug. This table scores only post-fix trades; each strategy gets its verdict at 50 closed trades.
| Strategy | Trades | Win rate | P&L | Verdict |
|---|---|---|---|---|
| 📰Catalyst scalper | 12 | 25% | +$29.36 | 12/50 — collecting |
| 🏛️Political news | 98 | 24% | -$14.83 | ❌ no edge |
| 🔗Crypto coupling | 0 | — | — | 0/50 — collecting |
| ₿Crypto news | 4 | 0% | -$10.18 | 4/50 — collecting |
Today's signal funnel
Every signal goes through six gates before a single share is sized. Here's how many fell out at each step today.
Gate 1 — Score
Raw confidence below minimum threshold
signals whose confidence score didn't clear the bar
Gate 2 — Regime
Strategy not allowed in current market regime
bootstrap bypass lets some through anyway — the ghost needs data from every regime
Gate 3 — Daily cap
Per-symbol daily trade limit reached
prevents piling into the same ticker multiple times in one session
Gate 4 — Sizer
Calibrator unfit → 0 shares sized
the win-probability model isn't confident enough to size a position
Gate 5 — Fan-out
Per-profile cap or price ceiling hit
Per-trade notional caps: Conservative $300, Clockpuncher $500, Newshound $2,000, Gunslinger $3,000 (the last two run the $10k large-capital track).
Gate 6 — Execution
Broker reject or market closed
paper broker accepts all, so this gate rarely fires during market hours
Gate reasons logged today
The ghosts during bootstrap
Five stock ghosts (market hours) + five crypto ghosts (24/7) — eight real personas and the two random-entry controls. Most start with $1,000 notional; Newshound and Gunslinger run the $10k large-capital track. During bootstrap stock ghosts trade 1 share per signal; crypto ghosts trade a small fixed notional per signal. Coinflip 🪙 and Dart Monkey 🐒 never graduate — controls stay controls.
📈 Stock ghosts
Conservative
the careful one
Bankroll: $1,000 notional
Per-trade cap: $300
PAPER · 8 outcomes
Clockpuncher
the daytrader
Bankroll: $1,000 notional
Per-trade cap: $500
ELIGIBLE FOR LIVE
Newshound
the headline chaser
Bankroll: $10,000 notional
Per-trade cap: $2,000
ELIGIBLE FOR LIVE
Gunslinger
the reckless one
Bankroll: $10,000 notional
Per-trade cap: $3,000
ELIGIBLE FOR LIVE
Coinflip
the control group
Bankroll: $1,000 notional
Per-trade cap: $150
PAPER · 57 outcomes
₿ Crypto ghosts
Stablecoin Stoic
the careful one
Bankroll: $1,000 notional
Max position: 20%
PAPER · 14 outcomes
HODLer
the patient one
Bankroll: $1,000 notional
Max position: 20%
PAPER · 27 outcomes
Pump Chaser
the momentum one
Bankroll: $1,000 notional
Max position: 12%
ELIGIBLE FOR LIVE
Degen
the chaos one
Bankroll: $1,000 notional
Max position: 20%
ELIGIBLE FOR LIVE
Dart Monkey
the blindfolded one
Bankroll: $1,000 notional
Max position: 12%
ELIGIBLE FOR LIVE
How the ghost is learning during bootstrap
Every signal goes through 6 gates (score · regime · per-symbol cap · sizer · fan-out · per-profile cap). While each strategy has fewer than 100 observed outcomes:
- The regime gate is informational only — signals get through even when the market is in the "wrong" regime for the strategy.
- The position sizer short-circuits to a bootstrap size — 1 share per stock signal, a small fixed notional per crypto signal — so it fits under each ghost's hard cap.
- Stock per-trade caps: Conservative ≤ $300 · Clockpuncher ≤ $500 · Newshound ≤ $2,000 · Gunslinger ≤ $3,000. Conservative & Clockpuncher run on $1k of working capital, Newshound & Gunslinger on $10k. Crypto ghosts size by position percentage (20% · 20% · 12% · 20%), backstopped by a hard notional cap.
When the first ghost graduates
The fan-out sizer switches from 1-share bootstrap to fractional-Kelly real sizing. Per-profile equity curves on /ghosts will start to diverge dramatically. The morgue opens.
Live promotion eligibility
A (profile × strategy) pair becomes live-eligible when ALL of these clear:
- outcomes ≥ profile threshold (50–100 depending on profile)
- profit_factor ≥ profile threshold (0.85–1.15 depending on profile)
- max_drawdown ≤ 8% (all profiles)
- last 30-day profit factor ≥ 0.90 (all profiles)
Applies equally to stock and crypto (profile × strategy) pairs.
Table populates once the first pair clears all gates.