ForecastMap Repaired 6.80
A local historical-analogue map, built from closed bars.
Load the map
- Remove older ForecastMap copies, then attach MIMIFX_ForecastMap_Repaired_v6.80.
- Allow history to load. The panel reports loading, insufficient matches or results.
- Enable MT5 Chart Shift or leave space to the right to see the projected lines.
What the lines mean
The indicator compares the last closed chart candle’s features with historical examples. Selected cumulative paths are normalized by their historical ATR and rescaled to current volatility. This makes the calculation suitable for comparing different DOGEUSD price scales, without assuming equal price changes across regimes.
P50 is the median path, shown in Gold. P10 and P90 are outer historical percentile paths in DeepSkyBlue. Optional P25/P75 paths add the inner band. They describe the selected historical sample; they are not calibrated future confidence intervals.
Similarity is 100 / (1 + distance). It is not a win rate. UP/DOWN/FLAT describes median direction, not an order instruction.
Main inputs
| Input | Default / meaning |
|---|---|
| HistoryBars | 5,000; allowed 500–20,000. |
| ForecastBars | 24; allowed 1–100. |
| MinimumMatches / MaximumMatches | 8 / 30; too few valid matches clears the old forecast. |
| MaximumDistance | 2.5; limits acceptable analogue distance. |
| UseHigherTimeframes | true; H1/H4 confirmation only when above the chart timeframe, using closed historical bars. |
| MatchSession | false; optional grouping uses server time plus SessionOffsetHours. Weekends are not filtered out. |
| ShowInnerBands / ShowTarget | false / true. Target is the final median reference. |
Panel controls
PanelTransparency=30 uses 0 opaque to 100 clear. The default panel is top-right at X=10, Y=50, width 420. Inputs set the starting location; select and drag for session positioning. Hide Forecast Info hides the panel while leaving forecast lines visible. ToggleX and ToggleY position the button.
Updates and limitations
A two-second timer checks for a new bar or history reset; incomplete history retries every five seconds. Selected historical feature/outcome windows do not overlap, reducing duplicated examples without proving independence. Future timestamps use ordinary timeframe spacing and do not skip future market closures.
This build uses local MT5 history. It does not load legacy CSV/JSON, TimesFM or ONNX overlays and does not run an external AI model. It places no trades.
Validation
The repaired build compiled with zero errors and warnings, and 13 numerical regression checks passed. Live chart operation and walk-forward predictive accuracy have not been verified. A clean compile does not validate a trading edge.