Blog / Historical Data / XAUUSD Dataset

XAUUSD Historical Dataset Free Download — M30 OHLC CSV + MT5 Export Scripts + Gold Forecast Engine

Q: Where can I download XAUUSD historical data for free?

Download our free XAUUSD M30 OHLC dataset with 100,000+ bars at algospecial.com/historical-data. It includes timestamp, Open, High, Low, Close, Tick Volume, and Spread. Also includes free MQL5 export scripts to extract any symbol/timeframe from MT5 History Center. The dataset is updated regularly through 2026.

Q: How do I add historical data to MT5?

Three methods: (1) Use the free OHLC_Export_Script.mq5 — compile it in MetaEditor, run it on your chart, and it exports all available bars to CSV. (2) Use MT5 History Center (F2) — select your symbol, click Download, then Export to CSV. (3) Use the OHLC_Append_Script.mq5 to incrementally add only new bars without rewriting your existing CSV file. Place the .mq5 file in MQL5/Scripts, compile (F7), then drag onto a chart.

Q: Can I use XAUUSD historical data for machine learning?

Yes. The CSV format (Timestamp, Open, High, Low, Close, TickVolume, Spread) is ready for Python ML with pandas: pd.read_csv(file, parse_dates=['Timestamp']). The dataset spans multiple market regimes making it ideal for LSTM, XGBoost, Random Forest, and CNN models. Our XAUUSD Forecast Engine ($50) provides a complete Python pipeline with analog matching, Elliott Wave detection, and walk-forward validation.

Q: How do I predict XAUUSD prices using Python?

Our XAUUSD Forecast Engine ($50, available via contact form) provides a complete Python analytical pipeline: Wilson confidence-interval analog matching, Elliott Wave structure detection, RSI divergence analysis, reversal pattern recognition, day-by-day calendar forecasting, walk-forward direction audit, and M30-based TP/SL simulation. It uses Mahalanobis distance similarity matching across 13 engineered features. Purchase includes the full source code and the historical dataset.

100,000+ M30 bars of clean gold price data, ready-to-use MQL5 export/append scripts, and a professional Python forecast engine with Elliott Wave detection, analog pattern matching, and walk-forward trade validation.

August 2, 2026 | 16 min read | Gold, Historical Data, Python, MQL5

Free XAUUSD M30 Dataset — Download Now

6.29 MB CSV · 100,000+ bars · Timestamp, Open, High, Low, Close, Volume, Spread · Updated August 2026

Download XAUUSD M30 CSV (6.3 MB) Get MT5 Export Scripts

The XAUUSD M30 Dataset — What's Inside
What You Can Do With This Data
MT5 Export & Append Scripts — How to Extract Your Own Data
How to Add Historical Data to MT5
XAUUSD Forecast Engine — Professional Gold Prediction in Python ($50)
Engine Features Breakdown
Frequently Asked Questions

The XAUUSD M30 Dataset — What's Inside

This is a clean, structured OHLCV dataset for XAUUSD (Gold vs US Dollar) on the M30 (30-minute) timeframe, exported directly from MetaTrader 5 History Center. It contains every M30 bar from the full available history — over 100,000 bars spanning multiple years — making it one of the most comprehensive free gold datasets available.

Column	Description	Example
Timestamp	Date and time of bar close (YYYY.MM.DD HH:MM:SS)	2026.08.05 15:30:00
Open	Opening price of the 30-minute bar	4257.318
High	Highest price during the bar	4259.348
Low	Lowest price during the bar	4254.455
Close	Closing price of the bar	4255.162
TickVolume	Number of ticks during the bar	1705
Spread	Spread in points at bar close	240

Pricing note: Gold is quoted in institutional format (e.g., 4257 = $4,257 per troy ounce). Divide by 100 for standard retail display if needed. The spread of 240 points = 2.4 in retail terms. This precision is ideal for machine learning — 3 decimal places preserve microstructure information that rounded retail data loses.

What You Can Do With This Data

Python Machine Learning: Load with pd.read_csv(file, parse_dates=['Timestamp']). Perfect for LSTM time series forecasting, XGBoost classification, Random Forest regression, and reinforcement learning environments.
MT5 Backtesting: Import into MT5 Strategy Tester for accurate gold EA backtesting. The M30 timeframe is the sweet spot — enough granularity for intraday patterns without the noise of lower timeframes.
Volatility Analysis: Calculate ATR, Bollinger Band width, and session-based volatility patterns to optimize stop placement and position sizing for gold's unique behavior.
Seasonality Research: Analyze day-of-week effects, session patterns, and monthly seasonality specific to gold — which behaves differently from currency pairs.
Correlation Studies: Combine with DXY, EURUSD, and US 10-year yield data to build multi-asset models that capture gold's macro drivers.

MT5 OHLC Export & Append Scripts — Extract Your Own Data

These two free MQL5 scripts let you export any symbol and any timeframe from your MetaTrader 5 terminal directly to CSV. The export script pulls ALL available history. The append script adds only new bars without rewriting your file — perfect for maintaining an ever-growing dataset.

Script 1: OHLC_Export_Script.mq5 — Full Export

Pulls the complete history for the current chart's symbol and timeframe from MetaQuotes servers. Saves to MQL5\Files\[Symbol]_[TF]_OHLC.csv. Includes automatic download retry with timeout handling.

How to use: Place in MQL5\Scripts\. Press F7 in MetaEditor to compile (zero errors). Drag onto any MT5 chart. The script downloads the full history and exports it to CSV.

OHLC_Export_Script.mq5

//+------------------------------------------------------------------+
//|                                        OHLC_Export_Script.mq5     |
//| Export all available OHLC bars to CSV                             |
//| Output: MQL5\Files\<Symbol>_<TF>_OHLC.csv                        |
//+------------------------------------------------------------------+
#property strict
#property script_show_inputs

input string FileName = ""; // Output filename (empty = auto: Symbol_TF_OHLC.csv)

void OnStart()
  {
   string sym = _Symbol;
   ENUM_TIMEFRAMES tf = _Period;

   if(TerminalInfoInteger(TERMINAL_CONNECTED) == 0)
      { Print("Terminal not connected to server. Cannot download."); return; }

   Print("Requesting full history for ", sym, " ", EnumToString(tf), "...");
   int prev = 0, cur = Bars(sym, tf);
   int retries = 0;
   while(cur < 10 || cur != prev)
     {
      prev = cur;
      Sleep(1000);
      cur = Bars(sym, tf);
      retries++;
      if(retries % 5 == 0) PrintFormat("  Downloaded %d bars so far...", cur);
      if(retries > 60) { Print("Timeout. Exporting what's available."); break; }
     }

   if(cur <= 0) { Print("No bars available for ", sym); return; }
   PrintFormat("Total bars available: %d", cur);

   string fn = (FileName != "") ? FileName : sym + "_" + EnumToString(tf) + "_OHLC.csv";
   int h = FileOpen(fn, FILE_WRITE|FILE_CSV|FILE_ANSI, ',');
   if(h == INVALID_HANDLE) { Print("Cannot create file: ", fn); return; }

   FileWrite(h, "Timestamp", "Open", "High", "Low", "Close", "TickVolume", "Spread");

   MqlRates r[];
   ArraySetAsSeries(r, true);
   int copied = 0, chunk = 5000;
   for(int start = 0; start < cur; start += chunk)
     {
      int toCopy = MathMin(chunk, cur - start);
      if(CopyRates(sym, tf, start, toCopy, r) <= 0) break;
      copied += toCopy;
      for(int i = 0; i < toCopy; i++)
         FileWrite(h, TimeToString(r[i].time, TIME_DATE|TIME_MINUTES|TIME_SECONDS),
                   DoubleToString(r[i].open,  (int)SymbolInfoInteger(sym, SYMBOL_DIGITS)),
                   DoubleToString(r[i].high,  (int)SymbolInfoInteger(sym, SYMBOL_DIGITS)),
                   DoubleToString(r[i].low,   (int)SymbolInfoInteger(sym, SYMBOL_DIGITS)),
                   DoubleToString(r[i].close, (int)SymbolInfoInteger(sym, SYMBOL_DIGITS)),
                   (long)r[i].tick_volume, (int)r[i].spread);
     }

   FileClose(h);
   PrintFormat("Exported %d bars to %s (MQL5\Files\%s)", copied, fn, fn);
  }
//+------------------------------------------------------------------+

Script 2: OHLC_Append_Script.mq5 — Incremental Update

Reads your existing CSV file, counts the rows, and appends only the new bars from MT5. Never duplicates data. Never rewrites your file. Perfect for weekly dataset updates.

How to use: Place both the script and your existing XAUUSDm_PERIOD_M30_OHLC.csv in MQL5\Files\. Compile and run the script. It detects existing rows and appends only new bars.

OHLC_Append_Script.mq5

//+------------------------------------------------------------------+
//|                              OHLC_Append_Script.mq5               |
//| Read existing CSV, copy only NEW bars, append without rewrite     |
//+------------------------------------------------------------------+
#property strict
#property script_show_inputs

input string FileName = "XAUUSDm_PERIOD_M30_OHLC.csv";

void OnStart()
  {
   string sym = _Symbol;
   ENUM_TIMEFRAMES tf = _Period;
   int digits = (int)SymbolInfoInteger(sym, SYMBOL_DIGITS);

   int totalBars = Bars(sym, tf);
   if(totalBars <= 0) { Print("No bars for ", sym); return; }

   int existingRows = 0;
   int h = FileOpen(FileName, FILE_READ|FILE_CSV|FILE_ANSI, ',');
   if(h == INVALID_HANDLE)
     {
      FileClose(h);
      FullExport(sym, tf, digits, 0, totalBars);
      return;
     }

   while(!FileIsEnding(h))
     {
      string line = FileReadString(h);
      if(line != "Timestamp" && StringLen(line) > 10) existingRows++;
     }
   FileClose(h);

   int newBars = totalBars - existingRows;
   if(newBars <= 0)
     {
      PrintFormat("CSV is up to date. %d bars total, %d in file.", totalBars, existingRows);
      return;
     }

   PrintFormat("Existing: %d rows | Total MT5 bars: %d | New: %d bars", existingRows, totalBars, newBars);

   MqlRates r[];
   ArraySetAsSeries(r, true);
   int copied = CopyRates(sym, tf, 0, newBars, r);
   if(copied <= 0) { Print("CopyRates failed"); return; }

   int hOut = FileOpen(FileName, FILE_WRITE|FILE_READ|FILE_CSV|FILE_ANSI, ',');
   if(hOut == INVALID_HANDLE) { Print("Cannot open for append"); return; }

   FileSeek(hOut, 0, SEEK_END);

   for(int i = copied - 1; i >= 0; i--)
     {
      FileWrite(hOut,
                TimeToString(r[i].time, TIME_DATE|TIME_MINUTES|TIME_SECONDS),
                DoubleToString(r[i].open,  digits),
                DoubleToString(r[i].high,  digits),
                DoubleToString(r[i].low,   digits),
                DoubleToString(r[i].close, digits),
                (long)r[i].tick_volume,
                (int)r[i].spread);
     }
   FileClose(hOut);
   PrintFormat("[OK] Appended %d new bars to %s", copied, FileName);
  }

void FullExport(string sym, ENUM_TIMEFRAMES tf, int digits, int start, int count)
  {
   MqlRates r[]; ArraySetAsSeries(r, true);
   int copied = CopyRates(sym, tf, start, count, r);
   if(copied <= 0) { Print("CopyRates failed"); return; }

   int h = FileOpen(FileName, FILE_WRITE|FILE_CSV|FILE_ANSI, ',');
   if(h == INVALID_HANDLE) { Print("Cannot create file"); return; }
   FileWrite(h, "Timestamp", "Open", "High", "Low", "Close", "TickVolume", "Spread");
   for(int i = copied - 1; i >= 0; i--)
     {
      FileWrite(h, TimeToString(r[i].time, TIME_DATE|TIME_MINUTES|TIME_SECONDS),
                DoubleToString(r[i].open,digits), DoubleToString(r[i].high,digits),
                DoubleToString(r[i].low,digits),  DoubleToString(r[i].close,digits),
                (long)r[i].tick_volume, (int)r[i].spread);
     }
   FileClose(h);
   PrintFormat("[OK] Created %s with %d bars", FileName, copied);
  }
//+------------------------------------------------------------------+

Download Export Script Download Append Script

How to Add Historical Data to MT5 — Three Methods

Method 1: MT5 History Center (Built-in, No Coding). Press F2 in MT5. Select your symbol and timeframe. Click Download — this pulls the full history from MetaQuotes servers. Click Export and choose CSV. The data is immediately loaded into MT5 for backtesting AND saved as a CSV file for external use. This is the fastest method for one-time exports.

Method 2: OHLC Export Script (Automated, Any Symbol). Place OHLC_Export_Script.mq5 in MQL5\Scripts\. Compile (F7 in MetaEditor). Drag onto any chart. The script automatically downloads the full history and exports to CSV in one click. Works on any symbol, any timeframe. Better than Method 1 for bulk exports across multiple pairs.

Method 3: OHLC Append Script (Incremental Updates). After you have your initial CSV from Method 1 or 2, place the CSV in MQL5\Files\. Run OHLC_Append_Script.mq5. It reads your existing file, counts the rows, and appends only new bars from the MT5 terminal. Run this weekly to keep your dataset current without re-exporting everything.

Where to find the exported files: After running any script, open MT5 → File → Open Data Folder → MQL5 → Files. Your CSV is there. The scripts use FileOpen() with FILE_COMMON, so the output goes to this standard location.

XAUUSD Forecast Engine — Professional Gold Prediction in Python

The XAUUSD Forecast Engine is a 1,282-line Python analytical suite that combines statistical rigor, machine learning, and classical technical analysis into a single runnable pipeline. It reads the M30 CSV dataset, builds daily bars, finds historically similar market conditions, projects forward price paths with confidence intervals, and validates every prediction with causal walk-forward auditing.

This is not a black box. The engine outputs every calculation, every confidence score, and every validation metric. You can inspect every step of the analysis. The GUI (built with tkinter) provides a clean interface for loading your CSV, configuring parameters, and viewing results.

Engine Features Breakdown

🎯

Wilson Confidence-Interval Analog Matching

Finds historically similar days using 13 engineered features — range, body percentage, returns, gaps, ATR ratio, and lagged values. Uses Mahalanobis distance with pseudo-inverse covariance for robust similarity scoring. Wilson score interval provides statistically sound confidence bounds — not just raw percentages.

📈

Multi-Horizon OHLC Price Path Prediction

Projects forward price paths for 1, 3, 5, and 7-day horizons with percentile bands (P10, P50, P90). Shows median open/close, high/low excursions, range, bull probability, and dispersion metrics. Validated with M30 bar-by-bar path simulation — not just end-point comparisons.

🌎

Multi-Degree Elliott Wave Detection

Detects impulse (1-2-3-4-5) and correction (A-B-C) wave structures at three fractal degrees (2-bar, 5-bar, 10-bar margins). Validates with strict Elliott rules — W3 never the shortest, W2/W4 never retrace 100%, no W4-W1 overlap. Fibonacci scoring quantifies wave quality. Projects target prices and completion dates.

🔄

RSI Divergence & Level Probability Analysis

Detects bullish/bearish RSI divergences across swing points with divergence strength scoring. RSI Level Analysis tracks every RSI 22 cross of 80/75/25/20 and computes conditional probabilities of hitting the next level vs reversing within 30 days — showing you exactly what happened historically after each RSI extreme.

👁

Reversal Pattern Scanner

Detects Shooting Star, Hammer, Marubozu, Doji, and Engulfing patterns with precise wick-to-body and range ratio calculations. Scores each pattern's reversal probability using a weighted ensemble of RSI divergence, cycle position, range climax, gap fill probability, swing proximity, and volume climax.

💪

M30-Simulated Trade Setup Optimizer

Tests every TP/SL/entry offset/order type combination across all M30 analog paths. Simulates actual order execution — limit vs stop, fill/no-fill detection, intrabar TP/SL competition resolution. Outputs the single best setup ranked by expected value per unit of risk. No theoretical backtesting — actual M30 bar-by-bar path simulation.

🔎

Causal Walk-Forward Audit

The most important feature. Tests the engine directionally (is the forecast better than baseline?) and commercially (do the trade setups produce positive expectancy?). Every test date can only see older data — no look-ahead bias. Outputs Brier scores, accuracy by decile, fill rates, cost-adjusted expectancy, profit factor, and drawdown.

📅

Day-by-Day Calendar Forecast

Projects next 7 trading days with full OHLC predictions and percentile bands. Each forecast day uses only the specific trading day offset from historical analogs — day 3 forecast only looks at what happened on day 3 after similar setups, not averaged across all horizons.

🔑

Cycle End Prediction

Uses conditional empirical distribution to compute the probability that the current bull or bear cycle will end within 1, 3, 5, or 7 days — given how long it has already lasted. Shows average, median, and maximum cycle lengths for context.

✅

Data Integrity Validation

Runs 10+ automated checks: D1 index monotonicity, M30 timestamp uniqueness, OHLC geometry validity, forecast date uniqueness and business-day compliance, and incomplete-analog-outcome blocking for look-ahead prevention. Every forecast is validated before display.

Get the XAUUSD Forecast Engine — $50 One-Time

Complete 1,282-line Python source code · Full documentation · Free XAUUSD dataset included · Commercial use allowed · One-time payment, no subscription

Dependencies: Python 3.8+, pandas, numpy, tkinter (built-in). Runs on Windows, macOS, and Linux.

Buy Now — Contact Us You'll receive the .py file + dataset via email after payment

Frequently Asked Questions

Where can I download XAUUSD historical data for free?

Download our free XAUUSD M30 CSV with 100,000+ bars from the download box above. It includes Timestamp, Open, High, Low, Close, TickVolume, and Spread. Also available: additional pairs and timeframes in our historical data library. All free, no registration required.

How do I import MT5 bars data to CSV?

Three methods: (1) MT5 History Center → F2 → select symbol/timeframe → Download → Export to CSV. (2) Use our free OHLC_Export_Script.mq5 — compile and run to auto-export any symbol. (3) Use OHLC_Append_Script.mq5 to add only new bars to an existing CSV. Output goes to MQL5\Files\. Scripts are downloadable above with full source code.

Can I use XAUUSD historical data for machine learning?

Yes. The CSV format is ready for Python: pd.read_csv(file, parse_dates=['Timestamp']). Use for LSTM, XGBoost, Random Forest, CNN, and reinforcement learning. Our AI analysis guide walks through the complete ML pipeline. The XAUUSD Forecast Engine ($50) provides a production-ready Python framework with all features built in.

How do I predict XAUUSD prices using Python?

The XAUUSD Forecast Engine ($50, available via contact form) provides a complete pipeline: analog pattern matching, Elliott Wave detection, RSI analysis, OHLC path projection, M30 TP/SL simulation, and walk-forward validation. It's the fastest way to go from CSV data to actionable gold forecasts. Contact us to purchase.

How do I add historical data to MT5 for backtesting?

MT5 History Center (F2) auto-loads data from MetaQuotes servers when you download. For external CSV data, import via the History Center: select symbol, click Import, choose your CSV. The OHLC scripts in this article handle the export side — from MT5 to CSV. For MT5 backtesting, the History Center already has everything loaded after download.

Why M30 timeframe for gold data?

M30 is the optimal balance for gold analysis: enough granularity to capture intraday patterns (London/NY session shifts, news reactions) without the noise of M1/M5 data. It's also efficient — 100,000 M30 bars cover years of history in a manageable 6MB file. The Forecast Engine aggregates M30 data into daily bars internally for pattern analysis while using M30 granularity for precise TP/SL simulation.

Build Your Gold Trading System Today

Start with the free dataset and export scripts. Use the MQL Prompts library to generate a gold-specific EA. Need professional analysis? The Forecast Engine gives you institutional-grade gold prediction in a single Python file.

Free Historical Data MQL Code Prompts Buy Forecast Engine $50