Телеграм канал «MQL5 Algo Trading»

A trend-following trading system built around OpenAI models typically separates signal generation from execution and risk controls. Market data is normalized into features, prompts are constrained to structured outputs, and model responses are treated as probabilistic inputs rather than deterministic rules.

Key design points include strict data boundaries to prevent leakage, reproducible backtests with walk-forward validation, and hard limits on position sizing, leverage, and drawdown. Execution should rely on deterministic order logic, while the model is limited to regime classification, filtering, or parameter selection.

Operational requirements include latency and rate-limit handling, audit logs of inputs and outputs, and continuous monitoring for drift. Any deployment should include fail-closed behavior and clear rules for disabling automation un...

👉 Read | NeuroBook | @mql5dev

DeepSeek Quantum Apex Color is a chart-based indicator intended to visualize the dominant trend while reducing noise through a rules-based filter stack.

Regime filtering is driven by ADX. Trend-following logic is active only when directional strength exceeds a threshold (default >20), limiting signals during sideways conditions.

Momentum and exhaustion checks combine RSI and Z-Score. RSI provides directional bias (RSI >50 long, <50 short), while Z-Score measures price deviation from a moving average to reject statistical outliers and reduce early entries.

Signal history is computed on the last closed bar (Bar 1) to avoid repainting and keep backtest results consistent.

👉 Read | Forum | @mql5dev

Mistral Trend Candles color-code trend state using Hull MA slope and an ADX regime filter.

Green marks a confirmed uptrend when the slope is above the configured threshold and ADX indicates a trend regime. Red marks a confirmed downtrend under the same conditions.

In weak or lateral markets, candles retain the last confirmed trend color to keep continuous green/red zones for faster visual reading.

The logic is non-repainting: closed candles keep their final color. Only the currently forming candle can change until close.

Rendering uses chart objects placed above native candles, fully covering them.

For M1–M5, lowering ADX_TrendMin to ~18–19 can improve responsiveness. Higher timeframes typically need a higher ADX_TrendMin. HMA_Period and Slope_MinATR control reactivity and noise tolerance.

👉 Read | AlgoBook | @mql5dev

Perplexity AI is being used to accelerate development of rule-based trading systems by generating code drafts, test cases, and documentation for common patterns such as trend filters, mean reversion triggers, and risk controls. Output still requires strict validation: confirm indicator math, verify order sizing, and reconcile bar indexing to avoid lookahead.

A production-ready setup typically includes deterministic backtests, walk-forward checks, and Monte Carlo analysis on execution assumptions. Logging, parameter constraints, and fail-safe handling for connectivity and rejected orders are mandatory.

AI assistance can reduce iteration time, but strategy quality remains dependent on data hygiene, robust statistics, and disciplined change control.

👉 Read | Signals | @mql5dev

A new indicator package is available featuring a “Good Tendency Estimator” combined with coloured candlesticks. The estimator is designed to quantify directional bias and highlight potential transitions in trend conditions, while the candle colouring provides a fast visual filter for momentum and regime changes.

Typical usage focuses on aligning entries with the estimator’s direction and using colour shifts as confirmation or early warning. Key points to validate during testing include behaviour in low-volatility ranges, sensitivity to parameter changes, repainting risks, and performance across symbols and timeframes.

👉 Read | Freelance | @mql5dev

Multi-period indicators with shifted lines require converting the shift into the current chart’s bar scale. A 1-bar shift on H4 becomes 4 bars on H1, and buffer copying must start from a shift-adjusted reference index to align values correctly.

Library updates: buffer shift setters now allow shifts on calculated buffers, and calculated buffer copy uses the line shift with an inverted sign to select the correct source start. Indicator factory methods (MA/AMA/DEMA/TEMA/VIDYA/Bands/Alligator, etc.) compute shiftBars = shift * periodBarRatio, pass that to the native handle, and apply the same shift to drawn and calculated buffers.

Common logic was consolidated: standard line types were normalized across indicators, universal preparation/clearing paths were reduced to 1/2/3-line cases, and a buffer-clear routine prevents artefacts from inactive indicators...

👉 Read | AppStore | @mql5dev

Market structure can be quantified by tracking sequences of higher highs/higher lows or lower highs/lower lows, and flagging when that pattern changes. This indicator applies fixed swing rules and prints non-repainting labels.

Swing highs/lows are confirmed only after N candles close to the right (default N=5), so markings appear with a delay and never shift. Each confirmed swing is classified as HH/LH versus the prior swing high, and HL/LL versus the prior swing low.

Break logic supports BOS and CHoCH. BOS is triggered when a candle closes beyond the last swing in the current trend direction and is drawn as a dashed line. CHoCH is triggered when a candle closes beyond the last swing against the active trend and is drawn as a solid line. Break detection can use close (default) or wick.

Inputs include swing strength, bars to scan (default 3000), disp...

👉 Read | CodeBase | @mql5dev