Signal overview
A clear view of the data behind your strategy.
This console does not place orders or change your bot’s settings. All times shown in UTC.
Signal lifecycle
Latest lifecycle eventResearch observations do not reactivate expired signals.
Data environments
Keep outcomes distinctSimulation and backtest records are not actual trades.
Reporting sources
| Source | Environment | Events | Last received · UTC |
|---|
A source appears after its first upload. Recent uploads do not prove that a bot is currently connected.
Signal explorer
Inspect what your strategy saw, and what happened next.
| Signal / source | Setup | Environment | State | Age at last event | Trigger price | Last event · UTC |
|---|
VQA ML research
A first setup-quality model for MT5 Gold.
Scores estimate a positive final net outcome if the bot fills the setup. They do not authorize orders.
Fresh demo scores from the backtest model
Gold demo · Observation onlyMT5 reports each completed prediction using the setup's frozen arming inputs. The model was trained on historical tester data. Outcomes appear after the demo trade is reconciled.
| Signal / source | Scored · UTC | Model | Score | Audited outcome | Evidence |
|---|
Demo-trained model predictions
Gold demo · Shadow onlyScores reported by the bot at arming from a model trained on demo outcomes. The dashboard checks prediction timing against order events. An upload can arrive later than the order.
| Signal / source | Scored · UTC | Model | Score | Audited outcome | Evidence |
|---|
VQA ONNX backtest
Historical · Shadow onlyA frozen model scores held-out MT5 Strategy Tester setups. Outcomes reflect the tester's simulated execution and cost settings. Demo predictions appear in the separate ledger above.
Native tester segments
| Signal / source | Armed · UTC | Setup | ONNX score | Tester net outcome | Closed · UTC |
|---|
Automatic training
—Checking scheduler status…
Checks run hourly on the server. The first eligible model is installed for shadow scores only. An existing model is preserved; repeated training is disabled.
Data readiness
MT5 · XAU · demoDefault training gate: 200 training outcomes, 50 validation outcomes, and at least 10 wins and non-wins in each period. Counts alone do not establish predictive value.
Data quality audit
Heldout validation
Retrospective comparison against a constant score learned from training. Predictive value is unproven until reviewed.
No trained model. Collect and audit outcomes first.
What the first model learns
Frozen arming features: setup family, side, ATR, RSI, spread, price distances, candle body, existing score and UTC hour.
Both score views start at arming. Dashboard research scores are calculated after upload when this page is opened.
Dashboard research scores
Calculated after upload for research. Signals armed before the model was created remain unscored.
| Signal / source | Armed · UTC | Setup | Score | Audited outcome | Status |
|---|
Backups & health
Know when your data was saved—and verified.
Latest backup
—Backup policy
- Schedule
- Daily, 00:30–00:35 UTC
03:30–03:35 Kuwait - Retention
- 14 days
- Format
- PostgreSQL custom archive + SHA-256
- Verification
- Restored to a temporary database on every run
- Storage
- Same VPS · root-only archive access
- Off-server copy
- Not configured
Same-server backups do not protect against loss of the VPS. Restore operations are performed over SSH, not from this console.