RayCo Labs — Live Proofs

MVP Mode: Cost-Only, Alpha-Neutral

Right now the system is running in cost-only mode:

Net Edge on this page is calculated as “after spread, fees, and a safety buffer”, assuming zero ML alpha.
Because we don’t add any predictive edge yet, Net Edge will usually be ≤ 0 bps and “Tradeability” will mostly say “No”.
Once live ML models and decay stats are fully wired and validated, this page will start showing positive net edge after costs where the data supports it.

In other words: this dashboard is already net-after-costs and ruthlessly conservative. We’d rather under-promise than over-claim.

Recorder Status

source: /status.json

coinbase — binance.us — last update —

Edge Decay Survival (live)

Tracks whether emitted net edge survives after 10s / 30s / 60s (from edge_decay.jsonl → edge_monitor.json).

Updated: — Rows: — OK: —

10s survival

— —

30s survival

— —

60s survival

— —

Per-exchange survival (smoothed). Low-N cells are dimmed.

Exchange	10s	30s	60s
—

Uses overall survival (combined across exchanges). Breakdown shown below.

Per-exchange breakdown (click to expand)

Same survival definition, shown separately for Coinbase vs BinanceUS.

Exchange	10s	30s	60s
—

Proofs Summary (Baseline)

source: /signals.json

pairs — rows — last update —

Net edge (bps) after fees/spread

Median —

Range — to —

Backtested ML Edge — Summary (offline)

Not live: derived from prior ML training runs and backtests.
sources: /api/signals_ml.json, /api/kpi.json

pair — tradeable — grade — hit-rate (net > 0) —

Net edge (bps) after fees/spread (backtest)

Median —

Range —

Live cards above show current spreads & costs only. This summary reflects historical ML performance and will be wired into live scoring in a later phase.

Net Edge & Costs (session)

source: /signals.json + in-browser history

Last updated: —

Live from signals.json & in-browser history.

Session hit-rate (net > 0): —

View:

BTC-USD

—

Avg Net —

Avg Cost —

Hit % —

ETH-USD

—

Avg Net —

Avg Cost —

Hit % —

SOL-USD

—

Avg Net —

Avg Cost —

Hit % —

Net edge is after spreads, impact, and fees. Costs here use the live cost stack from signals.json. Hit-rate is a simple session-level “net > 0” fraction, not P&L.

*MVP is running in cost-only, alpha-neutral mode. Positive Net Edge will only appear once live ML alpha is wired in and passes our safety checks.

Backtested ML Edge — Sample Rows (offline)

Not live: sample of recent ML-scored trades from backtests.
source: /api/signals_ml.json

ts (UTC)	symbol	net edge (bps)	spread (bps)
—

Edge Trust & Suppression

Live gating layer for testers — requires positive net edge after costs, reasonable hit-rate, and minimum coverage before a symbol is marked safe.
source: /edge_trust.json

Symbols

—

Safe to Trade

—

Suppressed

—

Last Update

—

Edge Trust is currently based on cost-only edges with no ML decay stats yet. “Safe to trade” will stay No until we see robust, positive edge after costs over time.

Symbol	Trust	Status	Top Reason
—

Dual-Tape Alignment

Alignment grades use strict HFT-style thresholds. A “D” does not mean broken — it means latency/coverage is below our ideal machine-speed standard.

source: /alignment.json

Pair	CB	BU	Grade	p50 ms	p95 ms	p99 ms	max gap (s)	coverage	reason
—

Recent Signals (preview)

source: /signals.json (live MVP slice)

live No live signals suppressed — view Show suppressed

Shows recently evaluated signals, including suppressed diagnostics. Expect many “No” rows during early data collection.

ts (UTC)	symbol	net edge (bps)	size (USD)	cap	tradeable	exec	why hidden
—