Benchmarking across time-series
Intraday liquidity heatmap, composite liquidity score, and cross-metal quality comparison across the London session.
How to read this page
Use this page to compare intraday liquidity conditions across metals and time buckets rather than looking at a single snapshot in isolation. The heatmap shows where liquidity tends to be strongest or weakest during the London session, while the cross-metal table helps benchmark current conditions against the rest of the session. This is a compact benchmarking layer built on the same shared prompt engine and top-of-book stats pipeline used elsewhere in the project.
Metric glossary
Liquidity scoreComposite 0-100 score using weighted normalized tightness (40%), depth (40%), and balance (20%). Formula:
LiquidityScore = 100 * (0.4 * Snorm + 0.4 * Dnorm + 0.2 * Inorm). Higher is better. This is a benchmarking summary, not a replacement for the underlying spread and size detail.Tightness (Snorm)Spread is normalized against a spread-percentage cap so tighter markets score higher. Inputs:
mid = (bid + ask) / 2, spread = ask - bid, spread_pct = spread / mid. Formula: Snorm = clamp(1 - (spread_pct / spread_pct_cap), 0, 1). This helps compare metals with different outright price levels.Depth (Dnorm)Visible bid plus ask size normalized against a per-metal depth cap. Inputs:
depth_total = bid_qty + ask_qty. Formula: Dnorm = clamp(depth_total / depth_cap, 0, 1). More displayed size scores higher, but the normalization caps out once visible size is considered strong enough.Balance (Inorm)Derived from order-book imbalance so more balanced visible books score higher. Inputs:
imbalance = (bid_qty - ask_qty) / (bid_qty + ask_qty). Formula: Inorm = 1 - abs(imbalance). Strongly one-sided books score lower.Intraday heatmapEach cell shows the average liquidity score for a metal in a given London time bucket. Brighter/greener cells indicate stronger liquidity. Bucket size only changes aggregation, not the underlying source observations.
Cross-metal comparisonSummarizes the current, average, best, and worst liquidity score for each metal across today’s London session buckets. This gives a compact quality benchmark for comparing metals side by side.
Spread percentage capCurrent backend benchmark uses
spread_pct_cap = 0.005 (0.5%). This is the cap used in Snorm = clamp(1 - (spread_pct / spread_pct_cap), 0, 1), where spread_pct = (ask - bid) / mid and mid = (bid + ask) / 2. Once spread percentage reaches that threshold, the tightness component bottoms out near zero. The purpose is to keep very wide markets from still scoring well on tightness, while making spread comparison more portable across metals with different outright price levels. At the moment this is a practical benchmark cap rather than a fully calibrated statistical threshold. The intended longer-term approach is to calibrate it from real spread-percentage distributions, potentially by metal or by prompt family, so the cap reflects a realistic “clearly poor tightness” boundary instead of a fixed placeholder.Per-metal depth capsCurrent backend caps are metal-specific: aluminium 20, copper 20, zinc 20, lead 10, nickel 8, tin 5. These are current benchmarking caps used in
Dnorm = clamp(depth_total / depth_cap, 0, 1). They stop larger-lot metals from automatically looking better just because they naturally display more size at top of book. At the moment they are practical benchmark values rather than a fully calibrated statistical model. The intended longer-term approach is to derive them from real depth distributions by metal (for example using a recent 75th or 80th percentile of depth_total), so each cap represents a realistic “strong enough” visible depth level for that metal.London day
-
Bucket size
-
Metals covered
-
Best current score
-
Benchmarking controls
Loading...
Liquidity score heatmap
Cross-metal comparison
| Metal | Current score | Average score | Best score | Worst score | Bucket count | Signal |
|---|