MQL5 Refactor Job – NIKA_FTS (REV + PSAR + HA only, closed-candle, max speed)

Refactor my current NIKA_FTS.mq5 into a minimal, fast, backtest-deterministic indicator that keeps:

• Tilson T3 pair (fast/slow) with separate alphas → drives REV state

• PSAR (plot + up/down cross signals)

• Heikin-Ashi overlay (candles/bars/hollow), colored by REV state

Remove everything else (RSI, TSI, TRM, MA1/2/3, MA crosses, Curl, bar-color logic from RSI/TRM, etc.).
All logic must be closed-candle only (no live/intrabar behavior).

Scope of Work

KEEP & Implement

1. Tilson T3 (double-alpha)

   • Inputs:

     input int aLength = 5; // fast T3 length input int Length = 8; // slow T3 length input double alphaFast = 0.70; // fast T3 alpha input double alphaSlow = 0.70; // slow T3 alpha

   • Buffers: anT3Average[] (fast), nT3Average[] (slow)

   • Plots: T3 Fast color line, T3 Slow color line, T3 Fill between them

   • Calls:

             calc_t3_series(src_arr, total, aLength, alphaFast, anT3Average);
     calc_t3_series(src_arr, total, Length , alphaSlow, nT3Average);
     

2. REV state (Tilson-based)

   • For bar i (closed bars only):

     • uc : anT3Average[i] >= nT3Average[i] && Close[i] > nT3Average[i]

     • dc : anT3Average[i] <= nT3Average[i] && Close[i] < nT3Average[i]

     • ur : anT3Average[i] <= nT3Average[i] && Close[i] > nT3Average[i] (cross up)

     • dr : anT3Average[i] >= nT3Average[i] && Close[i] < nT3Average[i] (cross down)

   • Output buffer: Rev[] = +1 (bull: uc/ur), -1 (bear: dc/dr), 0 otherwise.

3. PSAR

   • Inputs: start , inc_v , max_v

   • Use iSAR() handle; copy to sar[]

   • Signal buffer: SARsig[] = +1 if Close[i] >= sar[i] , else -1

   • Plot: dots/arrows (bull/bear color by state)

4. Heikin-Ashi overlay (KEEP)

   • Inputs:

     enum CandleType {CT_Candles }; input CandleType type =

             CT_Candles 

     input bool haover = true; input bool rPrice = false; // optional real-close line

   • Buffers: hac_* and/or hab_* + color index buffer

   • Color rule: map REV → HA color

     • Rev>0 → bull; Rev<0 → bear; Rev==0 → neutral

5. Alerts (closed-bar only)

   • Mode input:

     enum AlertMode { AM_PER_BAR=0, AM_PER_BAR_CLOSE }; input AlertMode i_alert_mode = AM_PER_BAR_CLOSE; input bool REVUP=false, REVDWN=false, SARUP=false, SARDWN=false;

   • Trigger only on last fully closed bar:

     • REVUP : Close[i-1] <= nT3[i-1] && Close[i] > nT3[i]

     • REVDWN : Close[i-1] >= nT3[i-1] && Close[i] < nT3[i]

     • SARUP : cross above PSAR

     • SARDWN : cross below PSAR

6. 2-row table (Rev, SAR)

   • Inputs:

             input bool showTable=false;
     input int  textSize=12, textWidth=100, textHeight=25, tableYpos=10, tableXpos=10;
     

   • Rows: “Rev”, “SAR”; icons ✅ / ❌ / 🔸 .

   • Update only when a new bar closes (no per-tick redraw).

REMOVE / PRUNE

• All RSI/TSI/TRM/MA/Curl code: enums, inputs, buffers, plots, alerts, palette colors, paint logic, table cells.

• Any leftover live/intrabar logic (e.g., using TimeCurrent , per-tick alerts, on-the-fly objects).

• Any plotting/objects unrelated to T3/PSAR/HA/table.

Closed-Candle-Only Rules (hard requirement)

• In OnCalculate , set:

  const int last_closed = rates_total - 2; int start = (prev_calculated>0 ? prev_calculated-1 : warmup); for(int i=start; i<=last_closed; ++i) { ... } // NEVER use rates_total-1 in logic

• Use data from arrays time[]/open[]/high[]/low[]/close[] only.

• No SymbolInfoDouble(BID/ASK) or TimeCurrent() in logic/alerts.

• Alerts and table update only when time[last_closed] changes (track static datetime last_bar_time ).

Performance & Code Quality (hard requirement)

• Max speed / zero “dirty” code:

  • Single pass loop, no nested loops per bar.

  • Respect prev_calculated . Start at max(prev_calculated-1, warmup) .

  • Pre-allocate & reuse buffers; ArraySetAsSeries(..., true) once.

  • No dynamic ObjectCreate/Delete inside the bar loop. Table draw only once per closed bar; otherwise disabled.

  • Move branches out of hot loops; precompute booleans reused multiple times.

  • Keep #property indicator_plots and indicator_buffers to exact counts used.

  • Avoid repeated switch / Math* in hot path when possible; cache inputs.

  • Use built-in iSAR handle; a single CopyBuffer per call; fill EMPTY_VALUE for overflow.

  • No sleeps, prints, file I/O, or terminal calls in calc loop.

• Backtest determinism:

  • No tick-dependent state. Results must be identical in visual/non-visual and multi-core tester.

  • No dependence on chart scaling/zoom or user objects.

• Deinit hygiene: ObjectsDeleteByPrefix(0,"NIKA-") ; do not nuke user objects.

Final Inputs (only)

// T3
input int    aLength   = 5;
input int    Length    = 8;
input double alphaFast = 0.70;
input double alphaSlow = 0.70;

// PSAR
input double start = 0.043;
input double inc_v = 0.043;
input double max_v = 0.34;

// Visuals
enum CandleType { CT_Hollow=0, CT_Bars, CT_Candles };
input CandleType type = CT_Hollow;
input bool t3vis   = true;
input bool psarvis = true;
input bool haover  = true;
input bool rPrice  = false;

// Alerts (closed-bar only)
enum AlertMode { AM_PER_BAR=0, AM_PER_BAR_CLOSE };
input AlertMode i_alert_mode = AM_PER_BAR_CLOSE;
input bool REVUP=false, REVDWN=false, SARUP=false, SARDWN=false;

// Optional table
input bool showTable=false;
input int  textSize=12, textWidth=100, textHeight=25, tableYpos=10, tableXpos=10;

Required Buffers/Plots

• Calc: anT3Average[] , nT3Average[] , sar[] , src_arr[] , HA arrays ( hac_* / hab_* )

• Signals: Rev[] , SARsig[]

• Plots:

  • Color-line T3 Fast , color-line T3 Slow , T3 Fill (2 buffers)

  • PSAR dots

  • HA candles/bars/hollow (color index)

  • Real Close line (optional)

• Palette: only colors used by T3/PSAR/HA.

Acceptance Tests

1. Compilation: MT5 current build, 0 errors / 0 warnings.

2. Search-clean: No occurrences of RSI , TSI , TRM , MA1 , MA2 , MA3 , HCROSS , Curl .

3. Closed-bar only: Changing tester from “Every tick” → “1 minute OHLC” yields identical buffers.

4. Performance: On M1 with 50k bars, first load < 50ms and incremental updates scale O(1) per bar (no visible lag).

5. Alerts: Fire once per qualifying closed bar (when AM_PER_BAR_CLOSE ), never on forming bar.

6. HA coloring: Follows Rev state correctly; table (if enabled) shows only “Rev”, “SAR”.

7. Determinism: Visual vs non-visual backtests produce the same Rev[]/SARsig[]/sar[]/T3 values.

Deliverables

1. Refactored NIKA_FTS.mq5 .

2. Short changelog of removed modules + final public API (inputs/plots/buffers).

3. QA note with 3 screenshots:

   • Bull REV + SARUP

   • Bear REV + SARDWN

   • Neutral REV

• Export a consolidated Signal[] buffer: +1 bull / 0 neutral / −1 bear (mirrors Rev ).

• Inputs for line widths and PSAR dot size.
  
  
  input int    WarmupBars   = 500;     // bars to prefill T3 before signaling
  input int    MaxHistory   = 10000;  // cap processing for speed; 0 = unlimited
  input bool   UseSeries    = true;    // ArraySetAsSeries on calc buffers