Signal Transparency
Every signal RefineX generates is logged here in real time. We show what fired, what was suppressed, and why — including signals we chose not to deliver.
System Status
API
Operational
Signal Engine
Healthy
Active Signals
—
Signals (24h)
—
Suppression Rate
55%
Signal Field Reference
regionAWS availability zone where the spot market condition was detected.
instance_typeEC2 instance class evaluated. RefineX scores each instance independently per region.
confidenceRegime-adjusted probability this represents a genuine arbitrage opportunity. Calculated after volatility gating and market state classification.
savings_pctEstimated cost reduction vs on-demand pricing at time of signal generation. Not a guarantee — spot prices shift continuously.
actionRecommended response: buy_spot (conditions favourable), hold (marginal), or avoid (volatility too high).
suppressedIf true, the signal was generated internally but not delivered. Suppression reasons are logged below.
regimeCurrent market state classification: STABLE_DISCOUNT, VOLATILE, TRANSITIONING, or COMPRESSED. Regime affects confidence scoring.
Confidence Bands
RefineX uses three confidence bands instead of raw scores. Bands are recalculated every cycle based on current regime state — a signal that scores HIGH in a stable market may be reclassified to WATCH if volatility increases before delivery.
>= 0.75
Delivered to autoscaler. Regime is stable, price spread is significant, false positive risk is low. Act on this.
0.50 - 0.74
Delivered with advisory flag. Conditions are favourable but regime shows early instability markers. Monitor closely.
< 0.50
Suppressed. Signal detected but confidence insufficient for delivery. Logged for audit purposes only.
Recent Signal Log
Last 20 signals — 9 delivered, 11 suppressed (55% suppression rate)
| Region | Instance | Confidence | Action | Savings | Suppressed | Reason |
|---|---|---|---|---|---|---|
| us-east-2 | t3.large | MEDIUM | buy_spot | 72.96% | Yes | ttl expired |
| us-west-2 | c6i.xlarge | HIGH | buy_spot | 64.12% | No | — |
| us-west-2 | c6i.xlarge | HIGH | buy_spot | 63.18% | No | — |
| us-west-2 | c7g.xlarge | HIGH | buy_spot | 70.8% | No | — |
| us-east-2 | c7g.xlarge | HIGH | buy_spot | 67.89% | No | — |
| us-east-2 | c7g.xlarge | HIGH | buy_spot | 65.54% | No | — |
| us-east-2 | c7g.xlarge | MEDIUM | buy_spot | 65.54% | Yes | ttl expired |
| us-west-2 | c6i.xlarge | HIGH | buy_spot | 63.18% | No | — |
| us-west-2 | c6i.xlarge | HIGH | buy_spot | 64.12% | No | — |
| us-east-2 | c7g.xlarge | MEDIUM | buy_spot | 67.89% | Yes | ttl expired |
Why We Suppress Signals
Overfiring destroys trust faster than underfiring. RefineX suppresses signals that pass initial detection but fail one of four quality gates before delivery.
Confidence below threshold
Signal confidence score fell below 0.75 after regime adjustment. Raw price spread was present but market state classification reduced conviction.
Six-hour clustering window
A higher-confidence signal for the same region and instance type was already active. Duplicate suppression prevents autoscaler noise.
High-volatility regime
Market state classified as VOLATILE or TRANSITIONING. Price spreads in volatile regimes revert faster than autoscaler reaction time.
Stale underlying data
Source pricing data was older than the acceptable freshness threshold. RefineX does not deliver signals derived from stale inputs.
Conflict-zone suppression
Region is under active military conflict or sustained physical infrastructure damage. Confidence band suppressed until the physical cause resolves and suppression rates return to pre-conflict baselines.
Data refreshes every 60 seconds. Last updated: Tue, 14 Apr 2026 00:51:09 GMT