Quant Pro Dashboard consolidates trading KPIs into a single on-chart panel, reducing context switching during analysis.

Account intelligence automatically reads account number and broker details. Performance tracking updates balance, equity, and floating PnL in real time. Trade management shows open positions and closed trade counts at a glance. Market awareness includes live spread, daily PnL percentage, and network latency (ping).

Implementation targets minimal overhead with sub-millisecond execution and layout rules designed to avoid text overlap. UI uses a high-contrast black panel with neon outline for consistent readability across chart themes, plus an optional quote section.

Deployment is drag-and-drop onto any chart, auto-aligned top-right with no additional configuration.

👉 Read | Quotes | @mql5dev

BlueMoon EA is an MT5 Expert Advisor built around EMA Envelope boundary checks to trigger reversal-oriented entries, with automated basket recovery for adverse movement. Signal generation and basket control are combined with configurable risk limits and account protection.

Core functions include automatic BUY/SELL entries, adaptive basket management, progressive lot sizing, configurable grid distance, per-trade take-profit, and basket-level profit targets. Operational safeguards cover spread filtering, maximum basket size and lot caps, drawdown protection with optional trading lock, and state recovery after terminal restarts. Support is included for broker symbol suffixes and MT5 hedging.

A real-time dashboard reports EMA/envelope levels, spread, basket metrics, drawdown, status, and recent activity. Configuration covers EMA period, deviation, lot model, gr...

👉 Read | Signals | @mql5dev

The article builds an MQL5 exporter that turns MetaTrader 5 history into a single JSON report of closed trades over a chosen date range, with optional symbol and magic filtering. It targets trade-level records (not raw deals) so the output loads directly in Python, R, or Excel.

Closed trades are reconstructed by pairing entry/exit deals via DEAL_POSITION_ID. Missing stop-loss or take-profit on deals is fixed with a reliable fallback: if DEAL_SL/DEAL_TP are zero, the script reads ORDER_SL/ORDER_TP from the originating order referenced by DEAL_ORDER.

The JSON includes metadata plus a trade array with ISO timestamps, readable direction, separated profit/commission/swap, and precomputed R-multiple and pip profit, using null when risk is undefined. File output uses ANSI text to avoid UTF-16 parsing issues.

👉 Read | AlgoBook | @mql5dev

Mapper converts a point cloud into a compact graph, but the highest sensitivity sits earlier: the lens and the cover. A lens assigns one scalar per point; a cover slices the lens range into overlapping intervals and records point membership.

Three lens options are used: eccentricity (mean distance), density (Gaussian-weighted neighborhood), and a coordinate projection. Intrinsic lenses cost O(N^2) over an existing distance matrix and can collapse on symmetric clouds, producing zero range and a single-bin cover.

The cover is controlled by resolution (interval count) and gain (overlap). With gain>0, points appear in multiple intervals and later become shared membership for edges. Practical starting values: resolution 5–15 and gain 0.2–0.5, with checks for degenerate lenses and expected double counting from overlap.

👉 Read | Signals | @mql5dev

False breakouts often occur during trend transitions, where velocity changes precede visible structure. A proposed approach pairs Divergence Mapping with a Temporal Fusion Transformer style attention layer to produce an alternative forecasting metric and a Trade Robot output.

The divergence engine computes a lookback slope for price and for momentum indicators, normalizes price slope by the window baseline, then takes the slope differential as a structural discrepancy. A sensitivity threshold gates signal validity rather than single-point breakout confirmation.

The TFT proxy consumes three time states and applies fixed attention weights (0.55/0.30/0.15) to form buy and sell scores with strict mutual exclusivity. An MQL5 implementation emphasizes passing arrays by reference to reduce memory churn.

RSI and DeMarker are used to separate closing consens...

👉 Read | Quotes | @mql5dev

The article presents the Bison Algorithm (BIA), a population-based optimizer for single-objective continuous search, built around two behaviors: fast exploration and a defensive clustering phase that stabilizes and refines good candidates.

BIA splits the population into a swarm group (about 80%) that moves toward a target and a runner group (about 20%) that probes new regions with a slightly perturbed direction vector. If a runner outperforms the swarm’s weakest member, it is promoted into the swarm.

The MT5-style implementation exposes clear controls (population size, swarm ratio, elite size, max step). Iterations combine weighted elite-centroid tracking, bounded step updates with rollback on worse fitness, and sorting to preserve top solutions—useful for robust parameter optimization in trading systems.

👉 Read | Forum | @mql5dev

A trend identification indicator is available for classifying buy-side and sell-side conditions using signals derived from the stochastic oscillator.

The logic relies on stochastic readings to determine direction and potential turning points, helping separate upward momentum phases from downward momentum phases.

Typical usage includes filtering entries to align with the detected direction, combining with a separate confirmation method, and validating behavior across multiple market regimes. Parameters such as %K, %D, smoothing, and threshold levels should be reviewed for the target instrument and timeframe.

👉 Read | AlgoBook | @mql5dev

Deterministic backtests produce a single realized path and can overstate risk-adjusted metrics. A validation battery addresses this by evaluating distributions, not one sequence.

Permutation testing uses sign-randomization on per-trade P&L to build a no-edge null for order-independent statistics like Sortino. Bootstrap resampling estimates metric stability via percentile and BCa confidence intervals. Monte Carlo trade-sequence shuffling keeps trade outcomes fixed but randomizes order to quantify drawdown sensitivity to sequencing.

All modules consume a trade-level CSV exported from MT5 deal history. An EA-side include writes one row per closed trade and supports multi-core optimization via a contention-safe file-open. Only Trade_Profit_USD and Indicator_Name are required; no strategy logic changes needed.

👉 Read | Freelance | @mql5dev

Chart Replay Pro (MT5 Strategy Tester, visual mode) received a major update focused on manual backtesting parity with live-trading controls.

Changes include runtime lot adjustment without restarting tests, per-position trade management, and an order workflow supporting buy/sell stop and buy/sell limit orders. UI is built with bitmap-based CbmpButton plus #resource inclusion, and a read-only edit field for lots.

Key implementation notes: OnChartEvent is not processed in Strategy Tester, so entry and TP/SL are adjusted via plus/minus buttons instead of draggable lines. Global state variables are used to prevent tick-driven UI bugs and to keep selected ticket, TP, and SL stable while actions execute.

👉 Read | Quotes | @mql5dev

Symbolic Aggregate approXimation (SAX) converts a price window into a short word via z-normalization, PAA segmentation, and Gaussian breakpoint discretization. The result is countable patterns with a lower-bounding distance (MINDIST) for pruning similarity search.

A practical MQL5 implementation fills missing platform primitives: probit-based Gaussian breakpoints, robust PAA for non-divisible lengths, flat-window rejection, and integer-coded symbols to keep distance math stable.

Applied to markets, normalization removes level and volatility, which can invalidate naive trading use. This is handled by evaluating forward outcomes in ATR units and enforcing a strict no-lookahead constraint when collecting precedents.

The analog search uses MINDIST as a fast filter, then computes full distances only for survivors, producing a distribution-based forecas...

👉 Read | Calendar | @mql5dev

Forex returns rarely have constant variance, so MSE-based regressors silently optimize the wrong objective. This article replaces that assumption with a probabilistic MLP that outputs both the conditional mean and a feature-dependent variance, trained via Gaussian negative log-likelihood.

The network uses a linear head for the mean and Softplus for strictly positive variance. Backprop is extended with explicit gradients for both outputs, enabling custom-loss training instead of relying on built-in MQL5 loss helpers.

Training is integrated with ALGLIB’s L-BFGS through parameter packing/unpacking plus a separate callback to log true per-iteration loss. A sample MT5 indicator trains on normalized price increments and plots forecasts with 95% confidence intervals, giving traders risk-aware signals, not just point estimates.

👉 Read | VPS | @mql5dev

Stop-loss and take-profit placement often defaults to round numbers or fixed ratios. Closed trade history already contains usable measurements for where price moved against and in favor of each entry.

The MAE/MFE Excursion Analyzer EA rebuilds round-trips from MT5 deal records (optionally filtered by magic), then scans M1 candles between entry and exit to compute per-trade MAE, MFE, and efficiency (captured move divided by MFE). Output includes on-chart stats and a CSV per trade.

Key readouts come from distributions split by winners/losers: winners’ p90 MAE as a stop distance, winners’ median MFE as a target zone, and efficiency to flag premature exits. The tool reads history only and does not place orders.

👉 Read | Freelance | @mql5dev