PW Trading — Diagnostics

How to Read This Data

▼

What the Pipeline Does

Every time a market's price updates, the V6.3 pipeline evaluates whether the price overshot — moved further than fundamentals justify — and whether it's likely to revert (snap back). It runs 16 sequential checks ("gates"). A signal must pass ALL 16 to become a trade candidate. Currently in Stage 0: observation only, no live trading.

Key Metrics Explained

Pipeline Runs — Total evaluations since startup. Higher = more price activity across monitored markets.
Candidates >1.2 — Runs where the elasticity score (ε) exceeded 1.2, meaning the price moved disproportionately to the driver (volume/attention). These are potential overshoot signals.
Epsilon (ε) — Measures how much price moved relative to what the underlying driver activity justified. ε < 0.8 = proportional (normal). ε 1.2–2.0 = elastic (possible overshoot). ε > 2.5 = hyperelastic (extreme overshoot). Tracked at two timescales: 15s (fast snap) and 60s (confirmed move).
Driver Z — Combined z-score of all drivers (volume spike, attention surge, order flow imbalance). Higher = stronger fundamental reason for the price move. Gate 7 requires Z ≥ 1.0.
Max Price Move — Largest 60-second price change observed. Shown in cents. Gate 2 requires ≥ 1¢ to even start evaluation.
Memory (RSS) — Process memory usage. Should stay under 300MB. Rising steadily over 24h = memory leak.

Gate Reference

Gates are ordered cheapest-first. Most candidates fail at Gates 1–3 in under 2ms. The pass rate shows what percentage survived each gate. 0% at a gate means nothing gets through — that gate is the current bottleneck.

Gate	Name	What It Checks	Threshold
G1	Regime	Is the market in a fadeable regime?	ATTENTION_SPIKE or EXHAUSTION
G2	Price Floor	Did price actually move?	≥ 1¢ in 60 seconds
G3	Liquidity	Can we actually trade this market?	Volume ≥ $5K, depth ≥ $300, spread ≤ 1.5¢
G4	Resolution	Is the market about to resolve?	≥ 24h to resolution
G5	Propagation Guard	Are correlated markets also moving?	< 2 correlated markets moving
G6	Propagation Window	Is there a pending propagation?	No unexpired windows
G7	Driver Validity	Is there a real driver behind the move?	Combined Z ≥ 1.0
G8	Elasticity	Is the price move disproportionate?	ε15 > 1.3 AND ε60 > 1.3
G9	Reflexivity	Is the market self-reinforcing?	R ≤ 4.0
G10	Classifier	Does the overshoot pattern match?	≥ 2/5 features strong, conf ≥ 0.65
G11	Fee Edge	Is there profit after fees?	Net edge per share > 0
G12	Stop Geometry	Is stop loss outside the spread?	Stop distance > spread
G13	Impact	Would our order move the market?	Impact share ≤ 30%
G14	Risk Limits	Within daily loss and position limits?	< 2% daily loss, < 3 concurrent
G15	Portfolio	Group/portfolio exposure within caps?	≤ 6% equity per group
G16	Size Viability	Is the computed order size valid?	Notional ≥ $10

Live = computed from real WebSocket data. Gates 4–6 and 13–16 are stubbed in Stage 0 (default to PASS since we're not trading).

Distribution Charts

Epsilon 15s / 60s — Histogram of ε values. Healthy system: most values in <0.3 (price moves are proportional), with a small tail above 1.2 (real overshoot signals).
Driver Z — How strong the underlying drivers are. Values >2.0 indicate significant volume/attention spikes.
Price Move — Distribution of 60s price changes. Most should be under 0.5¢. Values >5¢ are the ones that matter for fading.

What to Watch For

Good: Steady pipeline runs, candidates appearing in the >1.2 bucket, memory stable, some signals reaching Gates 4+
Watch: Gate 3 blocking 100% — means only illiquid markets are overshooting. Need liquid markets to produce signals too.
Problem: Memory growing steadily (leak), zero candidates over 12h+ (pipeline dead), or pipeline runs dropping to zero (WS disconnected)