日本語 JA
GitHub で見る

FuturePatternMemoryStrategy

Overview

FuturePatternMemoryStrategy is a StockSharp port of the classic MetaTrader experts FutureMA and FutureMACD. The original robots recorded sequences of indicator differences into CSV files, reused the stored statistics, and decided whether current conditions favored bullish or bearish breakouts. This C# version keeps the same idea, but replaces the file system with an in-memory pattern warehouse and exposes every knob as a strategy parameter. The strategy can operate either on the smoothed moving average spread (the FutureMA logic) or on the MACD histogram (the FutureMACD logic).

The strategy evaluates each finished candle in five stages:

  1. Indicator projection – compute the selected oscillator (MA spread or MACD histogram) and normalise it with a configurable scaling factor. The values are discretised to integers to create compact pattern signatures.
  2. Pattern hashing – maintain a sliding window of the latest AnalysisBars normalised values. Every time a new bar closes, the window is converted into a unique hash string that identifies the current market context.
  3. Historical swing analysis – inspect the previous FractalDepth candles, measure the distance between the oldest open and the highest/lowest extremes, and convert those ranges to points. These distances are the reward expectations that the original robots accumulated in their CSV files.
  4. Weighted memory update – the hash key is used to retrieve or create an entry in the pattern dictionary. The bullish and bearish take-profit expectations are updated with a weighted moving average controlled by ForgettingFactor, which reproduces the “forgetfulness” coefficient (zabyvaemost) from the MQL code.
  5. Signal evaluation and execution – if the bullish expectation dominates, the pattern was seen more than MinimumMatches times, and the projected gain is above MinimumTakeProfit, the strategy enters or adds to a long position. The bearish branch works symmetrically. Protective levels are derived from the stored statistics and optionally trailed as the trade moves in favour.

Conversion notes

  • Both MetaTrader experts are merged into one configurable strategy via the Source parameter, allowing you to switch between the MA-based engine and the MACD-based engine without recompilation.
  • File-based persistence was replaced with a Dictionary<string, PatternStats> that keeps all statistics in memory during the run. This avoids file I/O and stays within the StockSharp sandbox model.
  • Position management replicates the original stop/target placement: the stop uses the full averaged swing, while the take-profit uses StatisticalTakeRatio (the original Stat_Take_Profit). When EnableTrailingStop is true the stop is moved in quarter steps of the profit distance, exactly as the MQL expert modified its orders.
  • Manual mode (ManualMode) disables automated order placement but continues to collect statistics, matching the original Ruchnik flag behaviour.
  • Scaling in (AllowAddOn) mimics the dokupka flag and allows the strategy to add volume whenever the pattern repeats on a new bar.

Trading logic in detail

  • Indicator source
    • MA spread: calculates two smoothed moving averages (SMMA 6 and SMMA 24) on median prices and uses their difference.
    • MACD histogram: calculates the difference between the MACD main line and the signal line (12/26/9 configuration by default).
  • Normalisation: NormalizationFactor reproduces tocnost; it scales the raw difference before converting it to an integer signature. The conversion divides by 100 * MinPriceStep to maintain pip-based units.
  • Pattern memory: the dictionary stores, for each signature, the number of bullish matches, the average bullish distance, the number of bearish matches, and the average bearish distance. Values are updated with the weighted formula (current + input * ForgettingFactor) / (1 + ForgettingFactor).
  • Entry rules:
    • Long: bullish expectation ≥ bearish expectation, bullish matches > MinimumMatches, expected bullish distance > MinimumTakeProfit.
    • Short: bearish expectation ≥ bullish expectation, bearish matches > MinimumMatches, expected bearish distance > MinimumTakeProfit.
  • Risk management: stop-loss is set to one full averaged swing against the position; take-profit uses StatisticalTakeRatio of that swing. Trailing stops move after price travels one quarter of the distance, just like the original trailing routine.

Parameters

Parameter Description Default
CandleType Main timeframe used for calculations. 30-minute candles
Source Choose between MA spread (FutureMA) and MACD histogram (FutureMACD). MaSpread
FastMaLength / SlowMaLength SMMA lengths when Source = MaSpread. 6 / 24
MacdFastLength / MacdSlowLength / MacdSignalLength MACD periods when Source = MacdHistogram. 12 / 26 / 9
AnalysisBars Number of bars forming the pattern signature. 8
FractalDepth Number of past candles used to measure breakout distances. 4
MinimumMatches Required number of stored occurrences before taking a trade. 5
MinimumTakeProfit Minimum expected distance (in points) to accept the signal. 30
NormalizationFactor Scaling factor applied to the indicator difference. 10
ForgettingFactor Weight applied to new measurements in the pattern memory. 1.5
StatisticalTakeRatio Take-profit ratio relative to the measured swing. 0.5
EnableTrailingStop Enables quarter-step trailing stop logic. false
ManualMode Collect statistics but skip order placement. false
AllowAddOn Permit scaling in when a pattern repeats. true
Volume Order size used for entries. 0.1

Practical advice

  • The strategy relies on discretised hashes, so choose NormalizationFactor and AnalysisBars carefully: too large values produce sparse signatures, while too small values blend distinct states together.
  • When running in live trading, consider exporting the pattern dictionary after the session if you need persistence between runs.
  • Because the MQL version stored data per symbol/period, it is recommended to keep a dedicated strategy instance per instrument/timeframe to avoid cross-contamination of statistics.
using System;
using System.Collections.Generic;
using System.Text;

using StockSharp.Algo.Indicators;
using StockSharp.Algo.Strategies;
using StockSharp.BusinessEntities;
using StockSharp.Messages;

namespace StockSharp.Samples.Strategies;

/// <summary>
/// Pattern memory strategy that records normalized MA spread sequences,
/// tracks fractal outcomes, and trades when a recognized pattern has favorable statistics.
/// </summary>
public class FuturePatternMemoryStrategy : Strategy
{
	private readonly StrategyParam<DataType> _candleType;
	private readonly StrategyParam<int> _fastMaLength;
	private readonly StrategyParam<int> _slowMaLength;
	private readonly StrategyParam<int> _patternLength;
	private readonly StrategyParam<int> _minMatches;

	private readonly Queue<int> _patternWindow = new();
	private readonly Dictionary<string, (int buyCount, int sellCount)> _patterns = new();
	private decimal _entryPrice;
	private int _barIndex;
	private int _lastEntryBar;

	public FuturePatternMemoryStrategy()
	{
		_candleType = Param(nameof(CandleType), TimeSpan.FromMinutes(5).TimeFrame())
			.SetDisplay("Candle Type", "Timeframe for analysis.", "General");

		_fastMaLength = Param(nameof(FastMaLength), 6)
			.SetDisplay("Fast MA", "Fast EMA period.", "Indicators");

		_slowMaLength = Param(nameof(SlowMaLength), 24)
			.SetDisplay("Slow MA", "Slow EMA period.", "Indicators");

		_patternLength = Param(nameof(PatternLength), 5)
			.SetDisplay("Pattern Length", "Number of bars in pattern signature.", "Pattern");

		_minMatches = Param(nameof(MinMatches), 3)
			.SetDisplay("Min Matches", "Minimum pattern occurrences before trading.", "Pattern");
	}

	public DataType CandleType
	{
		get => _candleType.Value;
		set => _candleType.Value = value;
	}

	public int FastMaLength
	{
		get => _fastMaLength.Value;
		set => _fastMaLength.Value = value;
	}

	public int SlowMaLength
	{
		get => _slowMaLength.Value;
		set => _slowMaLength.Value = value;
	}

	public int PatternLength
	{
		get => _patternLength.Value;
		set => _patternLength.Value = value;
	}

	public int MinMatches
	{
		get => _minMatches.Value;
		set => _minMatches.Value = value;
	}

	protected override void OnReseted()
	{
		base.OnReseted();
		_patternWindow.Clear();
		_patterns.Clear();
		_entryPrice = 0;
		_barIndex = 0;
		_lastEntryBar = 0;
	}

	/// <inheritdoc />
	protected override void OnStarted2(DateTime time)
	{
		base.OnStarted2(time);

		_patternWindow.Clear();
		_patterns.Clear();
		_entryPrice = 0;
		_barIndex = 0;
		_lastEntryBar = 0;

		var fastEma = new ExponentialMovingAverage { Length = FastMaLength };
		var slowEma = new ExponentialMovingAverage { Length = SlowMaLength };

		var subscription = SubscribeCandles(CandleType);
		subscription
			.Bind(fastEma, slowEma, ProcessCandle)
			.Start();

		var area = CreateChartArea();
		if (area != null)
		{
			DrawCandles(area, subscription);
			DrawIndicator(area, fastEma);
			DrawIndicator(area, slowEma);
			DrawOwnTrades(area);
		}
	}

	private void ProcessCandle(ICandleMessage candle, decimal fastValue, decimal slowValue)
	{
		if (candle.State != CandleStates.Finished)
			return;

		_barIndex++;
		var close = candle.ClosePrice;

		// Normalize the MA spread into a discrete value
		var spread = fastValue - slowValue;
		var normalized = spread > 0 ? 1 : (spread < 0 ? -1 : 0);

		_patternWindow.Enqueue(normalized);
		while (_patternWindow.Count > PatternLength)
			_patternWindow.Dequeue();

		if (_patternWindow.Count < PatternLength)
			return;

		var key = BuildPatternKey(_patternWindow);

		// Record outcome: if price went up, it's a buy match, otherwise sell
		if (!_patterns.TryGetValue(key, out var stats))
			stats = (0, 0);

		if (close > fastValue)
			stats = (stats.buyCount + 1, stats.sellCount);
		else if (close < fastValue)
			stats = (stats.buyCount, stats.sellCount + 1);

		_patterns[key] = stats;

		// Cooldown: minimum SlowMaLength*4 bars between any trade action
		var cooldownBars = SlowMaLength * 4;
		var cooledDown = _barIndex - _lastEntryBar >= cooldownBars;

		// Position management
		var exitedThisBar = false;
		if (Position > 0)
		{
			if (cooledDown && (spread < 0 || (_entryPrice > 0 && close < _entryPrice * 0.985m)))
			{
				SellMarket();
				_lastEntryBar = _barIndex;
				exitedThisBar = true;
			}
		}
		else if (Position < 0)
		{
			if (cooledDown && (spread > 0 || (_entryPrice > 0 && close > _entryPrice * 1.015m)))
			{
				BuyMarket();
				_lastEntryBar = _barIndex;
				exitedThisBar = true;
			}
		}

		// Entry based on pattern statistics (cooldown between entries)
		if (!exitedThisBar && Position == 0 && cooledDown)
		{
			var total = stats.buyCount + stats.sellCount;
			if (total >= MinMatches)
			{
				if (stats.buyCount > stats.sellCount && spread > 0)
				{
					_entryPrice = close;
					_lastEntryBar = _barIndex;
					BuyMarket();
				}
				else if (stats.sellCount > stats.buyCount && spread < 0)
				{
					_entryPrice = close;
					_lastEntryBar = _barIndex;
					SellMarket();
				}
			}
		}
	}

	private static string BuildPatternKey(IEnumerable<int> values)
	{
		var sb = new StringBuilder();
		var first = true;
		foreach (var v in values)
		{
			if (!first) sb.Append('_');
			sb.Append(v);
			first = false;
		}
		return sb.ToString();
	}
}