Guide Tutorial
Dashboard Scenarios and Interpretation
This page explains common and edge-case dashboard outcomes with concrete examples. Use this as an interpretation guide, not trading advice.
Core Terms
- Base Price: the `from ...` value shown in Signal Board; this anchors direction and error context.
- Predicted Close: model-estimated target close at the forward horizon.
- Actual Close: observed market value once target time is reached.
- Resolved: target time has passed and row has been reconciled.
- Upcoming: target time has not passed yet.
Signal Lifecycle
- Step 1: a signal is generated at source time with base and target values.
- Step 2: status remains upcoming/live until target time arrives.
- Step 3: status becomes resolved and tracker labels are assigned.
Scenario 1 - BUY with Correct Direction
- Base: `100.00`, Predicted: `100.30`, Actual: `100.18`.
- Both predicted and actual moved upward from base.
- Expected interpretation: directional alignment is positive, even with some distance from exact target.
Scenario 2 - SELL with Correct Direction
- Base: `100.00`, Predicted: `99.70`, Actual: `99.82`.
- Both predicted and actual moved downward from base.
- Expected interpretation: directional call aligned; point error still exists.
Scenario 3 - HOLD in Range-Bound Conditions
- Base: `100.00`, Predicted: `100.01`, Actual: `99.99`.
- Move magnitude is small around base.
- Expected interpretation: HOLD can be the appropriate output in low-move environments.
Scenario 4 - Correct but Non-Zero Error
- Base: `67920.00`, Predicted: `67870.65`, Actual: `67956.00`, Error `0.126%`.
- A row can still be marked directionally acceptable while error is non-zero.
- Expected interpretation: direction and price distance are different quality dimensions.
Scenario 5 - Within Band
- Predicted and actual are close enough under tolerance, though not identical.
- Expected interpretation: acceptable proximity outcome; not exact-match outcome.
Scenario 6 - Outside Band
- Predicted and actual diverge more than tolerance.
- Expected interpretation: target miss beyond accepted proximity.
Scenario 7 - Upcoming vs Resolved
- Upcoming: no final quality label yet.
- Resolved: final tracker labels and error metrics become available.
Scenario 8 - High Confidence but Many HOLD Signals
- Can occur when market is choppy/range-bound and model frequently detects low actionable move.
- Expected interpretation: confidence score and action mix describe different things.
Scenario 9 - Low Hit-Rate but Low Avg Error
- Possible when many direction calls are narrowly wrong but predicted price remains close to realized price.
- Expected interpretation: directional quality and proximity quality must be read together.
Scenario 10 - Session Signal Mix Looks One-Sided
- Example: repeated SELL outputs in one session for one symbol.
- Expected interpretation: reflects current model bias under live conditions; review with historical tracker and not a single row.
Scenario 11 - Why Symbol Filter Matters
- All-symbol view mixes behaviors and can hide symbol-specific dynamics.
- Expected interpretation: evaluate both all-symbol portfolio view and single-symbol view.
Scenario 12 - Why Family Filter Matters
- Different model families can behave differently for the same symbol and session.
- Expected interpretation: compare family-specific tracker behavior before drawing conclusions.
Scenario 13 - Market Closed
- No new actionable updates are expected for closed instruments.
- Expected interpretation: closed status is operational state, not a model error.
Scenario 14 - What to Check Before Trusting a Current Signal
- Check recent resolved history for same symbol + family.
- Check error trend, not only one latest row.
- Check session mix context and market state.
Scenario 15 - Practical Reading Sequence
- Start with Signal Board (current action and expected move).
- Then open Real vs Predicted (short-term behavior).
- Then open Historical Tracker (resolved evidence).
- Finally read Session Signal Mix (context over recent period).
Risk Reminder
- All outputs are guidance signals and probabilistic, not guaranteed outcomes.
- Always use your own risk controls and independent confirmation.