style: normalize indentation in Orchestrator.mqh via clang-format

Whitespace-only: consistent 3-space indentation throughout (was a mix of
3/4/5/6.. spaces). Repo conventions preserved: for(/if( spacing, access
labels, attached braces. Adds .clang-format so the style is reproducible
(run: clang-format -i --style=file <file>).

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>

MQL5/Experts/MultiAgentTest/.clang-format

@@ -0,0 +1,15 @@
+BasedOnStyle: LLVM
+IndentWidth: 3
+TabWidth: 3
+UseTab: Never
+ColumnLimit: 0
+BreakBeforeBraces: Attach
+AllowShortFunctionsOnASingleLine: All
+AllowShortBlocksOnASingleLine: Always
+AllowShortIfStatementsOnASingleLine: true
+AllowShortLoopsOnASingleLine: true
+SortIncludes: false
+SpaceBeforeParens: Never
+IndentCaseLabels: true
+AccessModifierOffset: -3
+PointerAlignment: Right

MQL5/Experts/MultiAgentTest/Core/Orchestrator.mqh

@@ -233,32 +233,36 @@ public:
 
    int OpenTradeCount() const {
       int c = 0;
-      for(int i=0; i<maxOpenTrades; i++) if(openTrades[i].active) c++;
+      for(int i = 0; i < maxOpenTrades; i++)
+         if(openTrades[i].active) c++;
       return c;
    }
    int OpenTradeCount(bool isBuy) const {
       int c = 0;
-      for(int i=0; i<maxOpenTrades; i++) if(openTrades[i].active && openTrades[i].isBuy == isBuy) c++;
+      for(int i = 0; i < maxOpenTrades; i++)
+         if(openTrades[i].active && openTrades[i].isBuy == isBuy) c++;
       return c;
    }
    bool HasOpenTrade() const { return OpenTradeCount() > 0; }
    int OpenTicket() const {
-      for(int i=0; i<maxOpenTrades; i++) if(openTrades[i].active) return openTrades[i].ticket;
+      for(int i = 0; i < maxOpenTrades; i++)
+         if(openTrades[i].active) return openTrades[i].ticket;
       return -1;
    }
    double EntryPrice() const {
-      for(int i=0; i<maxOpenTrades; i++) if(openTrades[i].active) return openTrades[i].entryPrice;
+      for(int i = 0; i < maxOpenTrades; i++)
+         if(openTrades[i].active) return openTrades[i].entryPrice;
       return 0;
    }
    void ResetTradeState() {
-      for(int i=0; i<maxOpenTrades; i++) openTrades[i].active = false;
+      for(int i = 0; i < maxOpenTrades; i++) openTrades[i].active = false;
    }
 
-        Orchestrator(double minZ=0.5, double wMin=0.05, double wAlpha=0.05,
+   Orchestrator(double minZ = 0.5, double wMin = 0.05, double wAlpha = 0.05,
-                    bool useNeural=false, string modelFile="",
+                bool useNeural = false, string modelFile = "",
-                    bool trainMode=false, int nnEpochs=100,
+                bool trainMode = false, int nnEpochs = 100,
-                    double nnLR=0.001, int nnHidden=6,
+                double nnLR = 0.001, int nnHidden = 6,
-                    double riskPerTrade=0.01, double riskTotal=0.05) {
+                double riskPerTrade = 0.01, double riskTotal = 0.05) {
       agentCount = 0;
       combinedZ = 0;
       minActionableZ = minZ;
@@ -273,9 +277,9 @@ public:
       // Capacità iniziale: 1000 slot, cresce dinamicamente all'occorrenza
       maxOpenTrades = 1000;
       ArrayResize(openTrades, maxOpenTrades);
-         for(int i=0; i<maxOpenTrades; i++) openTrades[i].active = false;
+      for(int i = 0; i < maxOpenTrades; i++) openTrades[i].active = false;
 
-        for(int i=0; i<MAX_AGENTS; i++) InitCorrelation(i);
+      for(int i = 0; i < MAX_AGENTS; i++) InitCorrelation(i);
 
       // Self-evaluation init
       rollingIdx = 0;
@@ -288,8 +292,8 @@ public:
       convergeCount = 0;
       isConverged = false;
       maxDeltaRho = 0;
-       for(int i=0; i<MAX_AGENTS; i++) prevCorr[i] = 0;
+      for(int i = 0; i < MAX_AGENTS; i++) prevCorr[i] = 0;
-       for(int i=0; i<ROLLING_TRADES; i++) rollingReturns[i] = 0;
+      for(int i = 0; i < ROLLING_TRADES; i++) rollingReturns[i] = 0;
 
       // Neural network init
       m_useNeural = useNeural;
@@ -335,7 +339,7 @@ public:
    }
 
    void ReleaseAgents() {
-       for(int i=0; i<agentCount; i++) {
+      for(int i = 0; i < agentCount; i++) {
          if(agents[i]) {
             agents[i].Release();
             delete agents[i];
@@ -366,20 +370,28 @@ public:
    }
 
    void InitAgents(string symbol, ENUM_TIMEFRAMES tf) {
-      for(int i=0; i<agentCount; i++)
+      for(int i = 0; i < agentCount; i++)
          agents[i].Init(symbol, tf);
    }
 
    int TotalAgents() const { return agentCount; }
 
    double Analyze(const MarketData &data) {
-      if(agentCount == 0) { combinedZ = 0; return 0; }
+      if(agentCount == 0) {
+         combinedZ = 0;
+         return 0;
+      }
 
       double zs[MAX_AGENTS], ws[MAX_AGENTS], corrSign[MAX_AGENTS];
 
       // Fase 1: Analisi individuale
-      for(int i=0; i<agentCount; i++) {
+      for(int i = 0; i < agentCount; i++) {
-         if(!agents[i].enabled) { zs[i] = 0; ws[i] = 0; corrSign[i] = 1.0; continue; }
+         if(!agents[i].enabled) {
+            zs[i] = 0;
+            ws[i] = 0;
+            corrSign[i] = 1.0;
+            continue;
+         }
          zs[i] = agents[i].Analyze(data);
          double rho = GetCorrelation(i);
          // Competenza = |correlazione| segnale/ritorni (con floor weightMin, così
@@ -390,32 +402,36 @@ public:
       }
 
       // Fase 2: Interazione tra agenti (es. RegimeDetector aggiorna shared context)
-       for(int i=0; i<agentCount; i++) {
+      for(int i = 0; i < agentCount; i++) {
          if(agents[i].enabled) agents[i].Interact(agents, agentCount);
       }
       // Dopo Interact, aggiorna zs[] con lastZScore (alcuni agenti come PatternHunter
       // generano il segnale SOLO in Interact, non in Analyze)
-       for(int i=0; i<agentCount; i++) {
+      for(int i = 0; i < agentCount; i++) {
          if(agents[i].enabled) zs[i] = agents[i].lastZScore;
       }
 
       // Fase 3: Neural o Softmax Gating
       // Se la rete neurale è attiva e addestrata, usa forward pass
-        bool neuralReady = m_useNeural && m_neuralNet != NULL
+      bool neuralReady = m_useNeural && m_neuralNet != NULL && m_neuralNet.IsInitialized() && m_neuralNet.EpochsTrained() > 0;
-                           && m_neuralNet.IsInitialized()
-                           && m_neuralNet.EpochsTrained() > 0;
 
       if(neuralReady) {
          // Estrai feature vector: [z_Hurst, z_ADX, z_MA, z_Mom, z_Consensus, z_Pattern, agreement, trendStrength]
          vector nnInput(NN_FEATURES);
          for(int f = 0; f < NN_FEATURES; f++) nnInput[f] = 0.0;
-           for(int i=0; i<agentCount; i++) {
+         for(int i = 0; i < agentCount; i++) {
-                    if(agents[i].name == "Hurst")     nnInput[0] = agents[i].lastZScore;
+            if(agents[i].name == "Hurst")
-               else if(agents[i].name == "ADX")       nnInput[1] = agents[i].lastZScore;
+               nnInput[0] = agents[i].lastZScore;
-               else if(agents[i].name == "MA")        nnInput[2] = agents[i].lastZScore;
+            else if(agents[i].name == "ADX")
-               else if(agents[i].name == "Momentum")  nnInput[3] = agents[i].lastZScore;
+               nnInput[1] = agents[i].lastZScore;
-               else if(agents[i].name == "Consensus") nnInput[4] = agents[i].lastZScore;
+            else if(agents[i].name == "MA")
-               else if(agents[i].name == "Hunter")    nnInput[5] = agents[i].lastZScore;
+               nnInput[2] = agents[i].lastZScore;
+            else if(agents[i].name == "Momentum")
+               nnInput[3] = agents[i].lastZScore;
+            else if(agents[i].name == "Consensus")
+               nnInput[4] = agents[i].lastZScore;
+            else if(agents[i].name == "Hunter")
+               nnInput[5] = agents[i].lastZScore;
          }
          nnInput[6] = SHARED_regimeAgreement;
          nnInput[7] = SHARED_trendStrength;
@@ -430,14 +446,14 @@ public:
          // (mixing più uniforme, evita di sovra-fidarsi di un singolo agente).
          double wMean = 0;
          int wCount = 0;
-           for(int i=0; i<agentCount; i++) {
+         for(int i = 0; i < agentCount; i++) {
             if(!agents[i].enabled) continue;
             wMean += ws[i];
             wCount++;
          }
          if(wCount > 0) wMean /= wCount;
          double wVar = 0;
-           for(int i=0; i<agentCount; i++) {
+         for(int i = 0; i < agentCount; i++) {
             if(!agents[i].enabled) continue;
             wVar += (ws[i] - wMean) * (ws[i] - wMean);
          }
@@ -446,14 +462,14 @@ public:
          double safeTemp = MathMax(temp, DATA_EPS(temp) * 10.0); // evita overflow di MathExp
 
          double sumExp = 0;
-            for(int i=0; i<agentCount; i++) {
+         for(int i = 0; i < agentCount; i++) {
             if(!agents[i].enabled) continue;
             sumExp += MathExp(ws[i] / safeTemp);
          }
 
          combinedZ = 0;
          double epsExp = DATA_EPS((double)wCount);
-           for(int i=0; i<agentCount; i++) {
+         for(int i = 0; i < agentCount; i++) {
             if(!agents[i].enabled) continue;
             double softmaxW = (sumExp > epsExp) ? MathExp(ws[i] / safeTemp) / sumExp : 1.0 / MathMax(1, wCount);
             // Segnale sign-corretto: agenti anti-correlati contribuiscono invertiti.
@@ -504,8 +520,11 @@ public:
       EnsureTradeCapacity();
       // Trova slot libero
       int idx = -1;
-          for(int i=0; i<maxOpenTrades; i++) {
+      for(int i = 0; i < maxOpenTrades; i++) {
-             if(!openTrades[i].active) { idx = i; break; }
+         if(!openTrades[i].active) {
+            idx = i;
+            break;
+         }
       }
       if(idx < 0) { // non dovrebbe mai succedere dopo EnsureTradeCapacity
          Print("ERROR: slot non disponibile nonostante capacity expansion");
@@ -525,18 +544,24 @@ public:
       openTrades[idx].mfeATR = 0;
       openTrades[idx].active = true;
 
-         for(int i=0; i<agentCount; i++)
+      for(int i = 0; i < agentCount; i++)
          openTrades[idx].entryZScores[i] = agents[i].lastZScore;
 
       // Feature vector per NN
       for(int f = 0; f < NN_FEATURES; f++) openTrades[idx].entryFeatures[f] = 0.0;
-         for(int i=0; i<agentCount; i++) {
+      for(int i = 0; i < agentCount; i++) {
-                  if(agents[i].name == "Hurst")     openTrades[idx].entryFeatures[0] = agents[i].lastZScore;
+         if(agents[i].name == "Hurst")
-             else if(agents[i].name == "ADX")       openTrades[idx].entryFeatures[1] = agents[i].lastZScore;
+            openTrades[idx].entryFeatures[0] = agents[i].lastZScore;
-             else if(agents[i].name == "MA")        openTrades[idx].entryFeatures[2] = agents[i].lastZScore;
+         else if(agents[i].name == "ADX")
-             else if(agents[i].name == "Momentum")  openTrades[idx].entryFeatures[3] = agents[i].lastZScore;
+            openTrades[idx].entryFeatures[1] = agents[i].lastZScore;
-             else if(agents[i].name == "Consensus") openTrades[idx].entryFeatures[4] = agents[i].lastZScore;
+         else if(agents[i].name == "MA")
-             else if(agents[i].name == "Hunter")    openTrades[idx].entryFeatures[5] = agents[i].lastZScore;
+            openTrades[idx].entryFeatures[2] = agents[i].lastZScore;
+         else if(agents[i].name == "Momentum")
+            openTrades[idx].entryFeatures[3] = agents[i].lastZScore;
+         else if(agents[i].name == "Consensus")
+            openTrades[idx].entryFeatures[4] = agents[i].lastZScore;
+         else if(agents[i].name == "Hunter")
+            openTrades[idx].entryFeatures[5] = agents[i].lastZScore;
       }
       openTrades[idx].entryFeatures[6] = SHARED_regimeAgreement;
       openTrades[idx].entryFeatures[7] = SHARED_trendStrength;
@@ -586,7 +611,7 @@ public:
       d.confidence = MathAbs(combinedZ);
       d.minZ = minZ;
       if(action == "SIGNAL" || action == "ENTRY") {
-            for(int i=0; i<agentCount; i++) {
+         for(int i = 0; i < agentCount; i++) {
             if(agents[i].enabled && MathAbs(agents[i].lastZScore) > 0.1) {
                if((agents[i].lastZScore > 0) == (dir > 0)) d.agreeingCount++;
             }
@@ -596,11 +621,14 @@ public:
       decisionCount++;
    }
 
-void OnTradeClose(int ticket, double closePrice, string exitReason = "MANUAL") {
+   void OnTradeClose(int ticket, double closePrice, string exitReason = "MANUAL") {
       // Trova il trade nell'array
       int idx = -1;
-       for(int i=0; i<maxOpenTrades; i++) {
+      for(int i = 0; i < maxOpenTrades; i++) {
-          if(openTrades[i].active && openTrades[i].ticket == ticket) { idx = i; break; }
+         if(openTrades[i].active && openTrades[i].ticket == ticket) {
+            idx = i;
+            break;
+         }
       }
       if(idx < 0) return;
 
@@ -644,10 +672,10 @@ void OnTradeClose(int ticket, double closePrice, string exitReason = "MANUAL") {
       double sigThr = combinedZStats.Ready()
                           ? combinedZStats.Std() / MathSqrt(MathMax(1, agentCount))
                           : 1.0 / MathSqrt(MathMax(1, agentCount));
-for(int i=0; i<agentCount; i++) {
+      for(int i = 0; i < agentCount; i++) {
          if(MathAbs(openTrades[idx].entryZScores[i]) < sigThr) continue;
 
-UpdateCorrelation(i, openTrades[idx].entryZScores[i], actualReturnZ);
+         UpdateCorrelation(i, openTrades[idx].entryZScores[i], actualReturnZ);
          double rho = GetCorrelation(i);
 
          agents[i].weight = MathMax(weightMin, rho);
@@ -673,15 +701,21 @@ UpdateCorrelation(i, openTrades[idx].entryZScores[i], actualReturnZ);
          double target[NN_TARGETS];
          for(int f = 0; f < NN_FEATURES; f++) features[f] = openTrades[idx].entryFeatures[f];
 
-           target[0] = 0; target[1] = 0; target[2] = 0;
+         target[0] = 0;
+         target[1] = 0;
+         target[2] = 0;
          int tradeDir = openTrades[idx].isBuy ? 1 : -1;
 
          if(actualReturn > 0) {
-              if(tradeDir == 1)     target[0] = 1;
+            if(tradeDir == 1)
-              else                  target[2] = 1;
+               target[0] = 1;
+            else
+               target[2] = 1;
          } else {
-              if(tradeDir == 1)     target[2] = 1;
+            if(tradeDir == 1)
-              else                  target[0] = 1;
+               target[2] = 1;
+            else
+               target[0] = 1;
          }
 
          double sampleWeight = MathAbs(actualReturn) + 1.0; // trade profittevoli/perdenti pesano di più
@@ -700,7 +734,7 @@ UpdateCorrelation(i, openTrades[idx].entryZScores[i], actualReturnZ);
          }
       }
 
-UpdateHealth(actualReturn);
+      UpdateHealth(actualReturn);
       LogHealth();
 
       // Save completed trade record
@@ -732,14 +766,14 @@ UpdateHealth(actualReturn);
 
    // Trova un trade per ticket
    int FindTrade(int ticket) const {
-      for(int i=0; i<maxOpenTrades; i++)
+      for(int i = 0; i < maxOpenTrades; i++)
          if(openTrades[i].active && openTrades[i].ticket == ticket) return i;
       return -1;
    }
 
-// Aggiorna MAE/MFE per tutti i trade aperti (chiamato ogni barra)
+   // Aggiorna MAE/MFE per tutti i trade aperti (chiamato ogni barra)
    void UpdateOpenTrades(double high, double low) {
-        for(int i=0; i<maxOpenTrades; i++) {
+      for(int i = 0; i < maxOpenTrades; i++) {
          if(!openTrades[i].active) continue;
          if(high > openTrades[i].highestPrice) openTrades[i].highestPrice = high;
          if(low < openTrades[i].lowestPrice) openTrades[i].lowestPrice = low;
@@ -749,7 +783,7 @@ UpdateHealth(actualReturn);
 
    // Trailing stop: sposta SL dopo che il profitto supera la soglia
    void TrailStops() {
-        for(int i=0; i<maxOpenTrades; i++) {
+      for(int i = 0; i < maxOpenTrades; i++) {
          if(!openTrades[i].active) continue;
 
          double profitATR = openTrades[i].isBuy
@@ -774,7 +808,7 @@ UpdateHealth(actualReturn);
    // Trade da chiudere per inversione di segnale (ritorna array di ticket)
    void GetTradesToClose(int &closeTickets[], double currentZ, double minZ) {
       ArrayResize(closeTickets, 0);
-      for(int i=0; i<maxOpenTrades; i++) {
+      for(int i = 0; i < maxOpenTrades; i++) {
          if(!openTrades[i].active) continue;
          bool shouldClose = false;
          // Chiudi buy se segnale fortemente ribassista
@@ -820,14 +854,14 @@ UpdateHealth(actualReturn);
 
    int MaxTradeSlots() const { return maxOpenTrades; }
    bool IsBuyTrade(int ticket) const {
-      for(int i=0; i<maxOpenTrades; i++)
+      for(int i = 0; i < maxOpenTrades; i++)
          if(openTrades[i].active && openTrades[i].ticket == ticket)
             return openTrades[i].isBuy;
       return true; // default: buy
    }
 
    double GetTradeSL(int ticket) const {
-      for(int i=0; i<maxOpenTrades; i++)
+      for(int i = 0; i < maxOpenTrades; i++)
          if(openTrades[i].active && openTrades[i].ticket == ticket)
             return openTrades[i].slPrice;
       return -1;
@@ -836,7 +870,7 @@ UpdateHealth(actualReturn);
    // Ritorna il ticket dell'n-esimo trade attivo (per iterazione esterna)
    int GetTrackedTicket(int nth) const {
       int count = 0;
-      for(int i=0; i<maxOpenTrades; i++) {
+      for(int i = 0; i < maxOpenTrades; i++) {
          if(openTrades[i].active) {
             if(count == nth) return openTrades[i].ticket;
             count++;
@@ -845,7 +879,7 @@ UpdateHealth(actualReturn);
       return -1;
    }
 
-   FinalSignal GetFinalSignal(double overrideMinZ=0) {
+   FinalSignal GetFinalSignal(double overrideMinZ = 0) {
       double thr = (overrideMinZ > 0) ? overrideMinZ : AdaptiveMinZ();
       int dir = 0;
       if(combinedZ > thr) dir = 1;
@@ -859,7 +893,7 @@ UpdateHealth(actualReturn);
       double sigThr = combinedZStats.Ready()
                           ? combinedZStats.Std() / MathSqrt(MathMax(1, agentCount))
                           : 1.0 / MathSqrt(MathMax(1, agentCount));
-      for(int i=0; i<agentCount; i++) {
+      for(int i = 0; i < agentCount; i++) {
          if(!agents[i].enabled) continue;
          if(MathAbs(agents[i].lastZScore) > sigThr) {
             int agentDir = (agents[i].lastZScore > 0) ? 1 : -1;
@@ -876,7 +910,7 @@ UpdateHealth(actualReturn);
 
    void PrintAgentLearningSummary() const {
       Print("=== Agent Learning Summary ===");
-        for(int i=0; i<agentCount; i++) {
+      for(int i = 0; i < agentCount; i++) {
          string name = agents[i].name;
          double bias = agents[i].predictionError.Mean();
          double biasN = agents[i].predictionError.Count();
@@ -892,7 +926,7 @@ UpdateHealth(actualReturn);
    void PrintAgentStatus() const {
       Print("=== Orchestrator Status ===");
       Print("combinedZ: ", StringFormat("%+.4f", combinedZ));
-      for(int i=0; i<agentCount; i++) {
+      for(int i = 0; i < agentCount; i++) {
          Print("  ", agents[i].SignalInfo(),
                " | ρ=", StringFormat("%+.3f", GetCorrelation(i)),
                " | w=", StringFormat("%.3f", agents[i].weight));
@@ -903,7 +937,7 @@ UpdateHealth(actualReturn);
    }
 
    void Reset() {
-       for(int i=0; i<agentCount; i++) {
+      for(int i = 0; i < agentCount; i++) {
          agents[i].Reset();
          InitCorrelation(i);
       }
@@ -911,7 +945,7 @@ UpdateHealth(actualReturn);
       combinedZ = 0;
 
       if(m_trainBuffer) m_trainBuffer.Clear();
-        for(int i=0; i<maxOpenTrades; i++) openTrades[i].active = false;
+      for(int i = 0; i < maxOpenTrades; i++) openTrades[i].active = false;
       maeStats.Reset();
       mfeStats.Reset();
       maeWinStats.Reset();
@@ -926,8 +960,8 @@ UpdateHealth(actualReturn);
       convergeCount = 0;
       isConverged = false;
       maxDeltaRho = 0;
-       for(int i=0; i<MAX_AGENTS; i++) prevCorr[i] = 0;
+      for(int i = 0; i < MAX_AGENTS; i++) prevCorr[i] = 0;
-       for(int i=0; i<ROLLING_TRADES; i++) rollingReturns[i] = 0;
+      for(int i = 0; i < ROLLING_TRADES; i++) rollingReturns[i] = 0;
 
       histIdx = 0;
       histCount = 0;
@@ -952,7 +986,7 @@ UpdateHealth(actualReturn);
       // Rolling Sharpe (corrMinSamples = numero minimo per una correlazione stabile)
       if(rollingCount >= corrMinSamples) {
          double sum = 0, sumSq = 0;
-         for(int i=0; i<rollingCount; i++) {
+         for(int i = 0; i < rollingCount; i++) {
             sum += rollingReturns[i];
             sumSq += rollingReturns[i] * rollingReturns[i];
          }
@@ -969,8 +1003,11 @@ UpdateHealth(actualReturn);
 
       bool allHaveSamples = true;
       maxDeltaRho = 0;
-      for(int i=0; i<agentCount; i++) {
+      for(int i = 0; i < agentCount; i++) {
-         if(corrCount[i] < corrMinSamples) { allHaveSamples = false; break; }
+         if(corrCount[i] < corrMinSamples) {
+            allHaveSamples = false;
+            break;
+         }
          double rho = GetCorrelation(i);
          double delta = MathAbs(rho - prevCorr[i]);
          if(delta > maxDeltaRho) maxDeltaRho = delta;
@@ -1007,8 +1044,7 @@ UpdateHealth(actualReturn);
             // z_0.95 ≈ invNormalCDF(0.95) tramite approssimazione di Abramowitz & Stegun 26.2.23
             double p05 = 0.95;
             double t = MathSqrt(-2.0 * MathLog(1.0 - p05));
-              double z095 = t - (2.515517 + 0.802853*t + 0.010328*t*t)
+            double z095 = t - (2.515517 + 0.802853 * t + 0.010328 * t * t) / (1.0 + 1.432788 * t + 0.189269 * t * t + 0.001308 * t * t * t);
-                              / (1.0 + 1.432788*t + 0.189269*t*t + 0.001308*t*t*t);
             if(sharpeZ > z095)
                Print("  ✓ Sharpe significativamente positivo (z=", StringFormat("%.2f", sharpeZ), ")");
          }
@@ -1018,7 +1054,7 @@ UpdateHealth(actualReturn);
    void LogBarHistory(datetime time) {
       history[histIdx].time = time;
       history[histIdx].combinedZ = combinedZ;
-      for(int i=0; i<agentCount; i++) {
+      for(int i = 0; i < agentCount; i++) {
          history[histIdx].z[i] = agents[i].lastZScore;
          agentHistory[histIdx][i].lastZScore = agents[i].lastZScore;
          agentHistory[histIdx][i].weight = agents[i].weight;
@@ -1056,7 +1092,7 @@ UpdateHealth(actualReturn);
    // Somma del rischio (in valuta conto) di tutte le posizioni aperte
    double TotalRiskUsed() const {
       double total = 0;
-         for(int i=0; i<maxOpenTrades; i++)
+      for(int i = 0; i < maxOpenTrades; i++)
          if(openTrades[i].active) total += openTrades[i].riskAmount;
       return total;
    }
@@ -1078,7 +1114,10 @@ UpdateHealth(actualReturn);
       double maxTotalRisk = bal * m_riskTotal;
       double usedRisk = TotalRiskUsed();
       double remainingRisk = maxTotalRisk - usedRisk;
-         if(remainingRisk <= 0) { Print("  Risk budget esaurito (", StringFormat("%.2f", usedRisk), "/", StringFormat("%.2f", maxTotalRisk), ")"); return 0; }
+      if(remainingRisk <= 0) {
+         Print("  Risk budget esaurito (", StringFormat("%.2f", usedRisk), "/", StringFormat("%.2f", maxTotalRisk), ")");
+         return 0;
+      }
       riskAmount = MathMin(riskAmount, remainingRisk);
 
       // Costo in valuta conto per 1 lotto a questa distanza SL
@@ -1107,25 +1146,27 @@ UpdateHealth(actualReturn);
       int n = histCount;
       if(n < corrMinSamples) return 0;
       int lag = 1;
-       double sum=0, sumSq=0, sumShift=0, sumSqShift=0, sumCov=0;
+      double sum = 0, sumSq = 0, sumShift = 0, sumSqShift = 0, sumCov = 0;
       // Usa fino a n/2 per bilanciare stabilità e reattività, minimo corrMinSamples
       int cnt = MathMin(n, MathMax(corrMinSamples, n / 2));
-      for(int i=0; i<cnt-lag; i++) {
+      for(int i = 0; i < cnt - lag; i++) {
          int ii = (histIdx - 1 - i + MAX_HISTORY) % MAX_HISTORY;
          int jj = (ii - lag + MAX_HISTORY) % MAX_HISTORY;
          double x = history[ii].combinedZ;
          double y = history[jj].combinedZ;
-         sum += x; sumSq += x*x;
+         sum += x;
-         sumShift += y; sumSqShift += y*y;
+         sumSq += x * x;
-         sumCov += x*y;
+         sumShift += y;
+         sumSqShift += y * y;
+         sumCov += x * y;
       }
-      double mean = sum/cnt, meanS = sumShift/cnt;
+      double mean = sum / cnt, meanS = sumShift / cnt;
-      double var = sumSq/cnt - mean*mean;
+      double var = sumSq / cnt - mean * mean;
-      double varS = sumSqShift/cnt - meanS*meanS;
+      double varS = sumSqShift / cnt - meanS * meanS;
-      double cov = sumCov/cnt - mean*meanS;
+      double cov = sumCov / cnt - mean * meanS;
-      double denom = MathSqrt(var*varS);
+      double denom = MathSqrt(var * varS);
       double epsDenom = DATA_EPS(MathSqrt(MathMax(0, var) + MathMax(0, varS)));
-      return (denom > epsDenom) ? cov/denom : 0;
+      return (denom > epsDenom) ? cov / denom : 0;
    }
 
    double TradeSharpe() const {
@@ -1188,8 +1229,10 @@ UpdateHealth(actualReturn);
       string fn = (m_modelFilename != "") ? m_modelFilename
                                           : "TR_Agent_NN_" + Symbol() + "_" + EnumToString(Period()) + "_" + m_runTimestamp + ".dat";
       bool ok = m_neuralNet.Save(fn);
-       if(ok) Print("NN model saved: ", fn);
+      if(ok)
-       else   Print("NN model save FAILED: ", fn);
+         Print("NN model saved: ", fn);
+      else
+         Print("NN model save FAILED: ", fn);
       return ok;
    }
 
@@ -1209,9 +1252,7 @@ UpdateHealth(actualReturn);
    }
 
    bool IsNeuralReady() const {
-       return m_useNeural && m_neuralNet != NULL 
+      return m_useNeural && m_neuralNet != NULL && m_neuralNet.IsInitialized() && m_neuralNet.EpochsTrained() > 0;
-              && m_neuralNet.IsInitialized() 
-              && m_neuralNet.EpochsTrained() > 0;
    }
 
    string NeuralInfo() const {
@@ -1222,18 +1263,15 @@ UpdateHealth(actualReturn);
    // Salva summary CSV con learning stats di ogni agente
    void SaveAgentLearningCsv() const {
       string fn = "AgentLearning_" + Symbol() + "_" + EnumToString(Period()) + "_" + m_runTimestamp + ".csv";
-       int fh = FileOpen(fn, FILE_TXT|FILE_WRITE|FILE_COMMON);
+      int fh = FileOpen(fn, FILE_TXT | FILE_WRITE | FILE_COMMON);
       if(fh == INVALID_HANDLE) return;
       FileWriteString(fh, "agent,bias,bias_n,rho_learn,rho_learn_n,weight\n");
-       for(int i=0; i<agentCount; i++) {
+      for(int i = 0; i < agentCount; i++) {
          double bias = agents[i].predictionError.Mean();
          int biasN = agents[i].predictionError.Count();
          double rho = agents[i].predCorr.Ready() ? agents[i].predCorr.Correlation() : 0;
          int rhoN = agents[i].predCorr.Count();
-       FileWriteString(fh, agents[i].name + ","
+         FileWriteString(fh, agents[i].name + "," + StringFormat("%+.6f", bias) + "," + (string)biasN + "," + StringFormat("%+.6f", rho) + "," + (string)rhoN + "," + StringFormat("%.6f", agents[i].weight) + "\r\n");
-            + StringFormat("%+.6f", bias) + "," + (string)biasN + ","
-            + StringFormat("%+.6f", rho) + "," + (string)rhoN + ","
-            + StringFormat("%.6f", agents[i].weight) + "\r\n");
       }
       FileClose(fh);
       Print("Agent learning CSV saved: ", fn);
@@ -1246,7 +1284,7 @@ UpdateHealth(actualReturn);
 
    void SaveState(string symbol, ENUM_TIMEFRAMES tf) const {
       string fn = ModelFilename(symbol, tf);
-       int fh = FileOpen(fn, FILE_WRITE|FILE_BIN|FILE_COMMON);
+      int fh = FileOpen(fn, FILE_WRITE | FILE_BIN | FILE_COMMON);
       if(fh == INVALID_HANDLE) {
          Print("Save: impossibile creare ", fn, " errore ", GetLastError());
          return;
@@ -1258,11 +1296,11 @@ UpdateHealth(actualReturn);
       FileWriteInteger(fh, agentCount);
 
       // Agenti (RunningStats + extra)
-      for(int i=0; i<agentCount; i++)
+      for(int i = 0; i < agentCount; i++)
          agents[i].Save(fh);
 
       // Correlazioni per agente
-      for(int i=0; i<agentCount; i++) {
+      for(int i = 0; i < agentCount; i++) {
          FileWriteDouble(fh, corrMeanX[i]);
          FileWriteDouble(fh, corrMeanY[i]);
          FileWriteDouble(fh, corrCov[i]);
@@ -1279,11 +1317,11 @@ UpdateHealth(actualReturn);
       FileWriteInteger(fh, histIdx);
       int nHist = MathMin(histCount, MAX_HISTORY);
       int start = (histCount >= MAX_HISTORY) ? histIdx : 0; // dal più vecchio
-      for(int i=0; i<nHist; i++) {
+      for(int i = 0; i < nHist; i++) {
          int ii = (start + i) % MAX_HISTORY;
          FileWriteInteger(fh, (int)history[ii].time);
          FileWriteDouble(fh, history[ii].combinedZ);
-         for(int j=0; j<agentCount; j++)
+         for(int j = 0; j < agentCount; j++)
             FileWriteDouble(fh, history[ii].z[j]);
       }
 
@@ -1313,7 +1351,7 @@ UpdateHealth(actualReturn);
          return;
       }
 
-      int fh = FileOpen(fn, FILE_READ|FILE_BIN|FILE_COMMON);
+      int fh = FileOpen(fn, FILE_READ | FILE_BIN | FILE_COMMON);
       if(fh == INVALID_HANDLE) {
          Print("Load: impossibile aprire ", fn);
          return;
@@ -1323,10 +1361,10 @@ UpdateHealth(actualReturn);
       int savedCount = FileReadInteger(fh);
       int n = MathMin(savedCount, agentCount);
 
-      for(int i=0; i<n; i++)
+      for(int i = 0; i < n; i++)
          agents[i].Load(fh);
 
-      for(int i=0; i<n; i++) {
+      for(int i = 0; i < n; i++) {
          corrMeanX[i] = FileReadDouble(fh);
          corrMeanY[i] = FileReadDouble(fh);
          corrCov[i] = FileReadDouble(fh);
@@ -1342,12 +1380,12 @@ UpdateHealth(actualReturn);
          int savedHist = FileReadInteger(fh);
          int savedIdx = FileReadInteger(fh);
          int nHist = MathMin(savedHist, MAX_HISTORY);
-          for(int i=0; i<nHist; i++) {
+         for(int i = 0; i < nHist; i++) {
             int ii = (savedHist >= MAX_HISTORY) ? (savedIdx + i) % MAX_HISTORY : i;
             if(ii < 0 || ii >= MAX_HISTORY) { ii = 0; }
             history[ii].time = (datetime)FileReadInteger(fh);
             history[ii].combinedZ = FileReadDouble(fh);
-             for(int j=0; j<agentCount && j<MAX_AGENTS; j++)
+            for(int j = 0; j < agentCount && j < MAX_AGENTS; j++)
                history[ii].z[j] = FileReadDouble(fh);
          }
          histCount = nHist;
@@ -1389,8 +1427,11 @@ UpdateHealth(actualReturn);
    // Salva CSV con TUTTI i parametri derivati per analisi periodica
    void SaveAnalysisCSV(string symbol, ENUM_TIMEFRAMES tf) const {
       string fn = "TR_Agent_Analysis_" + symbol + "_" + EnumToString(tf) + "_" + m_runTimestamp + ".csv";
-        int fh = FileOpen(fn, FILE_TXT|FILE_WRITE|FILE_COMMON);
+      int fh = FileOpen(fn, FILE_TXT | FILE_WRITE | FILE_COMMON);
-        if(fh == INVALID_HANDLE) { Print("SaveAnalysis: errore apertura ", fn); return; }
+      if(fh == INVALID_HANDLE) {
+         Print("SaveAnalysis: errore apertura ", fn);
+         return;
+      }
 
       // ===================== INTESTAZIONE =====================
       FileWriteString(fh, "=== TR_Agent Analysis Report ===\r\n");
@@ -1466,7 +1507,7 @@ UpdateHealth(actualReturn);
       // ===================== AGENTI =====================
       FileWriteString(fh, "=== Agent Details ===\r\n");
       FileWriteString(fh, "Name,Weight,Rho,LastZ,Bias,BiasN,RhoLearn,RhoLearnN\r\n");
-        for(int i=0; i<agentCount; i++) {
+      for(int i = 0; i < agentCount; i++) {
          double rho = GetCorrelation(i);
          double bias = agents[i].predictionError.Mean();
          double biasN = agents[i].predictionError.Count();
@@ -1515,12 +1556,15 @@ UpdateHealth(actualReturn);
    // ===================== SAVE BAR HISTORY CSV =====================
    void SaveBarHistoryCSV(string symbol, ENUM_TIMEFRAMES tf) const {
       string fn = "BarHistory_" + symbol + "_" + EnumToString(tf) + "_" + m_runTimestamp + ".csv";
-        int fh = FileOpen(fn, FILE_TXT|FILE_WRITE|FILE_COMMON);
+      int fh = FileOpen(fn, FILE_TXT | FILE_WRITE | FILE_COMMON);
-        if(fh == INVALID_HANDLE) { Print("BarHistoryCSV: errore apertura ", fn); return; }
+      if(fh == INVALID_HANDLE) {
+         Print("BarHistoryCSV: errore apertura ", fn);
+         return;
+      }
 
       string header = "bar,time,combinedZ";
       string agentNames[MAX_AGENTS];
-        for(int i=0; i<agentCount; i++) {
+      for(int i = 0; i < agentCount; i++) {
          agentNames[i] = agents[i].name;
          header += "," + agentNames[i] + "_z";
       }
@@ -1529,10 +1573,10 @@ UpdateHealth(actualReturn);
 
       int n = MathMin(histCount, MAX_HISTORY);
       int start = (histCount >= MAX_HISTORY) ? histIdx : 0;
-        for(int i=0; i<n; i++) {
+      for(int i = 0; i < n; i++) {
          int ii = (start + i) % MAX_HISTORY;
          string line = (string)i + "," + TimeToString(history[ii].time) + "," + StringFormat("%+.6f", history[ii].combinedZ);
-           for(int j=0; j<agentCount; j++)
+         for(int j = 0; j < agentCount; j++)
             line += "," + StringFormat("%+.6f", history[ii].z[j]);
          line += "," + StringFormat("%.4f", SHARED_regimeH);
          line += "," + StringFormat("%.1f", SHARED_adxRaw);
@@ -1549,21 +1593,24 @@ UpdateHealth(actualReturn);
    // ===================== SAVE TRADE HISTORY CSV =====================
    void SaveTradeHistoryCSV(string symbol, ENUM_TIMEFRAMES tf) const {
       string fn = "TradeHistory_" + symbol + "_" + EnumToString(tf) + "_" + m_runTimestamp + ".csv";
-        int fh = FileOpen(fn, FILE_TXT|FILE_WRITE|FILE_COMMON);
+      int fh = FileOpen(fn, FILE_TXT | FILE_WRITE | FILE_COMMON);
-        if(fh == INVALID_HANDLE) { Print("TradeHistoryCSV: errore apertura ", fn); return; }
+      if(fh == INVALID_HANDLE) {
+         Print("TradeHistoryCSV: errore apertura ", fn);
+         return;
+      }
 
       string header = "ticket,isBuy,entryTime,closeTime,barsHeld,entryPrice,closePrice,entryATR,entryZ,slPrice,highestPrice,lowestPrice,maeATR,mfeATR,actualReturn,exitReason";
       string agentNames[MAX_AGENTS];
-        for(int i=0; i<agentCount; i++) {
+      for(int i = 0; i < agentCount; i++) {
          agentNames[i] = agents[i].name;
          header += "," + agentNames[i] + "_entryZ";
       }
-        string featLabels[NN_FEATURES] = {"Hurst","ADX","MA","Momentum","Consensus","Hunter","Agreement","TrendStr"};
+      string featLabels[NN_FEATURES] = {"Hurst", "ADX", "MA", "Momentum", "Consensus", "Hunter", "Agreement", "TrendStr"};
-        for(int f=0; f<NN_FEATURES; f++) header += ",feat_" + featLabels[f];
+      for(int f = 0; f < NN_FEATURES; f++) header += ",feat_" + featLabels[f];
       FileWriteString(fh, header + "\r\n");
 
       // Completed trades
-        for(int t=0; t<completedTradeCount; t++) {
+      for(int t = 0; t < completedTradeCount; t++) {
          string line = (string)completedTrades[t].ticket;
          line += "," + (string)(completedTrades[t].isBuy ? 1 : 0);
          line += "," + TimeToString(completedTrades[t].entryTime);
@@ -1580,15 +1627,15 @@ UpdateHealth(actualReturn);
          line += "," + StringFormat("%.4f", completedTrades[t].mfeATR);
          line += "," + StringFormat("%+.4f", completedTrades[t].actualReturn);
          line += "," + completedTrades[t].exitReason;
-           for(int j=0; j<agentCount; j++)
+         for(int j = 0; j < agentCount; j++)
             line += "," + StringFormat("%+.4f", completedTrades[t].entryZScores[j]);
-           for(int f=0; f<NN_FEATURES; f++)
+         for(int f = 0; f < NN_FEATURES; f++)
             line += "," + StringFormat("%+.4f", completedTrades[t].entryFeatures[f]);
          FileWriteString(fh, line + "\r\n");
       }
 
       // Open trades (still active)
-        for(int i=0; i<maxOpenTrades; i++) {
+      for(int i = 0; i < maxOpenTrades; i++) {
          if(!openTrades[i].active) continue;
          string line = (string)openTrades[i].ticket;
          line += "," + (string)(openTrades[i].isBuy ? 1 : 0);
@@ -1604,9 +1651,9 @@ UpdateHealth(actualReturn);
          line += "," + StringFormat("%.4f", openTrades[i].maeATR);
          line += "," + StringFormat("%.4f", openTrades[i].mfeATR);
          line += ",,OPEN"; // no return, exit=OPEN
-           for(int j=0; j<agentCount; j++)
+         for(int j = 0; j < agentCount; j++)
             line += "," + StringFormat("%+.4f", openTrades[i].entryZScores[j]);
-           for(int f=0; f<NN_FEATURES; f++)
+         for(int f = 0; f < NN_FEATURES; f++)
             line += "," + StringFormat("%+.4f", openTrades[i].entryFeatures[f]);
          FileWriteString(fh, line + "\r\n");
       }
@@ -1617,11 +1664,14 @@ UpdateHealth(actualReturn);
    // ===================== SAVE AGENT INTERACTION CSV =====================
    void SaveAgentInteractionCSV(string symbol, ENUM_TIMEFRAMES tf) const {
       string fn = "AgentInteraction_" + symbol + "_" + EnumToString(tf) + "_" + m_runTimestamp + ".csv";
-        int fh = FileOpen(fn, FILE_TXT|FILE_WRITE|FILE_COMMON);
+      int fh = FileOpen(fn, FILE_TXT | FILE_WRITE | FILE_COMMON);
-        if(fh == INVALID_HANDLE) { Print("AgentInteractionCSV: errore apertura ", fn); return; }
+      if(fh == INVALID_HANDLE) {
+         Print("AgentInteractionCSV: errore apertura ", fn);
+         return;
+      }
 
       string header = "bar,time,combinedZ";
-        for(int i=0; i<agentCount; i++) {
+      for(int i = 0; i < agentCount; i++) {
          string n = agents[i].name;
          header += "," + n + "_z," + n + "_weight," + n + "_rho," + n + "_bias," + n + "_biasStd," + n + "_rhoLearn," + n + "_rawSignal";
       }
@@ -1629,10 +1679,10 @@ UpdateHealth(actualReturn);
 
       int n = MathMin(histCount, MAX_HISTORY);
       int start = (histCount >= MAX_HISTORY) ? histIdx : 0;
-        for(int i=0; i<n; i++) {
+      for(int i = 0; i < n; i++) {
          int ii = (start + i) % MAX_HISTORY;
          string line = (string)i + "," + TimeToString(history[ii].time) + "," + StringFormat("%+.6f", history[ii].combinedZ);
-           for(int j=0; j<agentCount; j++) {
+         for(int j = 0; j < agentCount; j++) {
             line += "," + StringFormat("%+.4f", agentHistory[ii][j].lastZScore);
             line += "," + StringFormat("%.4f", agentHistory[ii][j].weight);
             line += "," + StringFormat("%+.4f", agentHistory[ii][j].rho);
@@ -1650,11 +1700,14 @@ UpdateHealth(actualReturn);
    // ===================== SAVE DECISION LOG CSV =====================
    void SaveDecisionLogCSV(string symbol, ENUM_TIMEFRAMES tf) const {
       string fn = "DecisionLog_" + symbol + "_" + EnumToString(tf) + "_" + m_runTimestamp + ".csv";
-        int fh = FileOpen(fn, FILE_TXT|FILE_WRITE|FILE_COMMON);
+      int fh = FileOpen(fn, FILE_TXT | FILE_WRITE | FILE_COMMON);
-        if(fh == INVALID_HANDLE) { Print("DecisionLogCSV: errore apertura ", fn); return; }
+      if(fh == INVALID_HANDLE) {
+         Print("DecisionLogCSV: errore apertura ", fn);
+         return;
+      }
 
       FileWriteString(fh, "time,action,direction,zScore,combinedZ,price,ticket,agreeingCount,totalAgents,confidence,minZ\r\n");
-        for(int d=0; d<decisionCount; d++) {
+      for(int d = 0; d < decisionCount; d++) {
          string line = TimeToString(decisionLog[d].time);
          line += "," + decisionLog[d].action;
          line += "," + (string)decisionLog[d].direction;
@@ -1671,5 +1724,5 @@ UpdateHealth(actualReturn);
       FileClose(fh);
       Print("Decision log salvato: ", fn, " (", decisionCount, " decisioni)");
    }
- };
+};
- #endif
+#endif