MQL5 Algo Trading
Открыть в Telegram
The best publications of the largest community of algotraders. Subscribe to stay up-to-date with modern technologies and trading programs development.
Больше📈 Аналитический обзор Telegram-канала MQL5 Algo Trading
Канал MQL5 Algo Trading (@mql5dev) языкового сегмента Английский является активным участником. Сейчас сообщество объединяет 524 303 подписчиков, занимая 149 место в категории Технологии и приложения и 5 место в регионе Великобритания.
📊 Показатели аудитории и динамика
С момента создания невідомо проект демонстрирует стремительный рост, собрав аудиторию из 524 303 подписчиков.
Согласно последним данным от 20 июля, 2026, канал показывает стабильную активность. За последние 30 дней изменение числа участников составило 11 707, а за последние 24 часа — 437, при этом общий охват остаётся высоким.
- Статус верификации: Не верифицирован
- Уровень вовлечённости (ER): Средний показатель вовлечённости аудитории составляет 3.24%. В первые 24 часа после публикации контент обычно набирает 1.92% реакций от общего числа подписчиков.
- Охват публикаций: В среднем каждый пост получает 16 995 просмотров. В течение первых суток публикация набирает 10 083 просмотров.
- Реакции и взаимодействия: Аудитория активно поддерживает контент: среднее количество реакций на один пост — 36.
- Тематические интересы: Контент сосредоточен на ключевых темах, таких как indicator, chart, mql5, candle, range.
📝 Описание и контентная политика
Автор описывает ресурс как площадку для выражения субъективного мнения:
“The best publications of the largest community of algotraders.
Subscribe to stay up-to-date with modern technologies and trading programs development.”
Благодаря высокой частоте обновлений (последние данные получены 21 июля, 2026) канал поддерживает актуальность и высокий уровень охвата публикаций. Аналитика показывает, что аудитория активно взаимодействует с контентом, что делает его важной точкой влияния в категории Технологии и приложения.
524 303
Подписчики
+43724 часа
+2 7017 дней
+11 70730 день
Архив постов
524 415
This article extends the MCP approach from trading execution to the full development workflow by connecting AI agents to MQL5 Algo Forge, a Git-based host backed by Forgejo. Using its HTTPS REST API, an assistant can programmatically create repos, commit EA files, manage branches, open pull requests, file issues, and publish releases.
The server is a portable Python project (no MetaTrader/Windows dependencies) built in layers: JSON config + token, an httpx-based API client with consistent error normalization, domain handlers, and a FastMCP tool surface exposing 12 tools.
Key implementation details include safe token handling (config or environment variable) and Base64 encoding/decoding for file endpoints. A single “commit file” tool abstracts POST vs PUT by detecting existing files and retrieving sha automatically, enabling reliable updates without t...
👉 Read | Docs | @mql5dev
524 415
Reliable MT5 Expert Advisors require systematic validation before sending trade requests. Common constraints include max positions/orders, volume min/max/step, SL/TP distance, margin, session limits, symbol permissions, and news filters. Market publication tests already enforce many of these cases, so reusable checks reduce duplication and regressions.
A compact MQL5 validator set can cover: lot sizing with step-based normalization (avoids retcode 10014), SL/TP distance vs SYMBOL_TRADE_STOPS_LEVEL, price digit normalization (avoids retcode 10015), freeze-level checks for order modifications, margin/funds checks, and “no changes” guards to prevent TRADE_RETCODE_NO_CHANGES.
Additional utilities include: pending-order limits, lightweight new-bar detection, tradability checks for SYMBOL_TRADE_MODE_DISABLED, calendar-based news windows (live only), and UTC sess...
👉 Read | Signals | @mql5dev
524 415
This part adds a pinned-tools ribbon to complement the deep, tabbed sidebar: the sidebar stays optimized for discovery, while the ribbon provides one-click access to a small set of frequently used drawing tools. The ribbon auto-hides when empty and preserves pin order.
Pinned tools are stored in the engine as an ordered list with a minimal API (count, get, contains, pin/unpin, toggle). Pinning appends; unpinning compacts the array without reordering, keeping behavior predictable.
A shared anti-aliased pushpin glyph is rendered via polygon coverage and alpha blending, reused consistently across flyout rows, the Pinned sidebar tile, and the ribbon.
The ribbon is a floating surface with drag, resize, and horizontal scrolling. It clips overflowing icons using an offscreen canvas and shows a proportional scrollbar thumb, keeping interaction smooth eve...
👉 Read | NeuroBook | @mql5dev
524 415
Walk-Forward Analysis is reframed as a measurable robustness test for MT5 EAs: optimize on in-sample windows, score degradation on forward windows, and summarize it as a reproducible metric instead of eyeballing equity curves.
The core score is Walk-Forward Efficiency (WFE): per-window ratio of out-of-sample to in-sample Sharpe. Windows pass only if they retain at least 50% efficiency, with guards that force failure when in-sample Sharpe is non-positive or too small to be meaningful.
A native MQL5 pipeline implements the full loop: an EA logs per-bar equity to CSV, a fast reader ingests it, WFE_Engine.mqh computes Sharpe/WFE with numerical stability and validity flags, and a CCanvas histogram renders pass/fail windows and reference lines directly on-chart for immediate diagnosis.
👉 Read | Calendar | @mql5dev
524 415
KCI Embedded Sniper is an algorithmic reversal-entry EA built around a fully embedded Kinetic Compression Index engine. KCI math (Velocity Quotients, Kinetic Displacement, Energy Dispersion, Phase Velocity) runs inside the EA, avoiding iCustom() latency and thread desynchronization. Signals are computed on closed bars only, targeting non-repainting execution with Singularity exhaustion validation and a Williams %R momentum gate.
The design removes external indicator files, reducing memory overhead and eliminating indicator path, loading, and buffer read errors. Computation is event-driven via rates_total and internal matrices, optimized for multi-asset operation in a single process loop.
Risk and filtering are parameterized: fixed lot sizing, ED-based dynamic SL/TP scaling, WPR period and extreme levels, plus KCI sensitivity controls (ZScorePeriod, Compress...
👉 Read | Docs | @mql5dev
524 415
The Kinetic Compression Index (KCI) is a custom oscillator for detecting market exhaustion and localized compression events. It computes kinematic-style metrics directly inside the indicator loop, reducing reliance on multiple external indicator handles and simplifying EA buffer management.
The design exposes reproducible components such as Velocity, Deviation, and Dispersion, with Z-Score normalization applied over a rolling window before forming a composite KCI value. Signal buffers are built to validate on closed bars to reduce intrabar repainting and async tick timing issues.
A unified buffer map is central to integration: KCI main line and color index, buy/sell signal buffers, plus internal calculation arrays including Velocity Quotient, Kinetic Displacement, Energy Dispersion (usable as a volatility proxy for SL/TP logic), Phase Velocity, and R...
👉 Read | Forum | @mql5dev
524 415
A Smart Money Concepts (SMC/ICT) market-structure indicator for MetaTrader 5 that derives structure from swing-high/swing-low sequencing and renders the current state directly on the chart.
Swing points are detected via N-bar fractals and plotted as arrows. Breaks are classified as BOS when price closes beyond the prior swing in the trend direction, and as CHoCH when the close exceeds the prior swing against the trend. On each structure break, the last opposite candle is flagged as an Order Block. Fair Value Gaps are identified as 3-candle imbalance zones. After a CHoCH, QML draws a dotted retrace level at the left-shoulder swing.
Key inputs include swing sensitivity, close vs wick confirmation, lookback bars, and per-feature toggles with configurable colors. Runs on any symbol and timeframe, recalculating on every new bar with close-confirmed markings.
👉 Read | Docs | @mql5dev
524 415
A trading-journal export script pulls raw account history from the terminal and writes it to a single CSV in one run.
The script loads deals for the last N days, then aggregates them into closed positions by position ID. It works on both netting and hedging accounts and supports partial fills. Entry and exit prices are volume-weighted across all entry and exit deals per position.
Each CSV row includes: position ID, symbol, direction, volume, open time and VWAP open price, close time and VWAP close price, points result (direction-aware), commission, swap, net profit, duration (minutes), and the first deal comment for strategy-level filtering.
Inputs cover: period (days), file name (or auto TradeJournal_YYYY-MM-DD.csv), CSV separator (default “;” for Excel), optional common-folder output, and an optional current-symbol filter. Output goes to MQL5\Files o...
👉 Read | Signals | @mql5dev
524 415
A position sizing script calculates lot size directly on the active chart symbol using a user-defined risk budget (percent of equity or fixed account currency) and stop-loss distance (points or explicit price level).
The calculation pulls live contract specs: tick size and tick value for money-per-point conversion, then applies volume min/max/step rules. Output is normalized by rounding down to the volume step so risk does not exceed the limit, and the exact risk at the normalized size is reported.
Margin is estimated for the intended direction using OrderCalcMargin. Clear warnings are generated when free margin is insufficient, when the computed size is below the symbol minimum (minimum volume exceeds the risk budget), or when the result is capped at the symbol maximum.
A full report is printed to the chart Comment, Experts journal, and a single-line Alert....
👉 Read | NeuroBook | @mql5dev
524 415
Spread Monitor Panel is a lightweight on-chart panel for tracking the live spread of the current chart symbol, refreshed every second. Spread is calculated from Ask-Bid in points, avoiding issues with integer symbol properties on float-spread accounts.
The panel reports current spread plus minimum, average, and maximum since attachment, including sample count and start time. Current value is color-coded against user thresholds: green below warning, orange between warning and danger, red at or above danger.
Optional alerting triggers when spread remains at or above the danger threshold for a configurable number of consecutive seconds, with a cooldown to prevent repeated notifications during spikes. Configuration includes thresholds, alert settings, panel corner, offsets, and font size.
Designed for monitoring only: no trades, buffers, or plots. Suitabl...
👉 Read | Quotes | @mql5dev
524 415
o you have Expert Advisors written in MQL4 and think migrating them to MetaTrader 5 will take too much time? Let's check this out right now.
We'll take an existing Expert Advisor from the MQL5.com CodeBase, feed it into ChatGPT, generate an MQL5 version, compile it in MetaEditor, and run it in the Strategy Tester. The entire process will take just a few minutes.
If you don't want to work with the code or handle the migration manually, there's an even easier option. MQL5.com offers a Freelance service where you can hire professional developers to convert Expert Advisors, indicators, and other applications.
Discuss the video:
👉 MQL5.community for traders
👉 MetaQuotes official YouTube channel
524 415
MetaTrader 5 EA portability often fails due to broker-specific symbol identifiers, not strategy logic. Variants like EURUSD, EURUSDm, EURUSD.fx, GOLD, XAUUSD.a, or US30.cash can break SymbolSelect, bid/ask reads, and OrderSend, sometimes without explicit errors.
A broker-agnostic symbol layer addresses this with deterministic translation, terminal validation, caching, and CSV persistence. Resolution is defined as testable behavior: Resolve("EURUSD") returns a selectable symbol with non-zero quotes, or an empty result with logging.
Architecture splits responsibilities into modules: mapping storage, resolver with hash-based lookup plus controlled discovery, a small resolution cache, a host verification EA for live checks, and a benchmark tool for latency measurement.
👉 Read | AlgoBook | @mql5dev
524 415
This article implements the core step of persistent homology in MQL5: standard column reduction of a Vietoris–Rips boundary matrix over Z/2, turning sparse column relationships into a readable persistence diagram of (birth, death, dimension) pairs.
The reduction tracks each column’s pivot, uses a pivot-to-owner table for O(1) lookups, and performs fast XOR “additions” via symmetric difference of two sorted face lists. Empty reduced columns mark feature creation; unique pivots mark feature death. Unpaired creators become essential features with infinite death.
A compact data model (SPersistencePair, CTDADiagram) enables persistence, Betti numbers at any epsilon, and barcode/diagram views. A CTDA facade runs the full pipeline (Takens embedding → distances → filtration → boundary → reduction) from a price window in one call.
Correctness is cross-chec...
👉 Read | NeuroBook | @mql5dev
524 415
Brain-Computer Interfaces are moving from lab demos to clinical trials, with systems such as Neuralink N1 and the Blackrock Utah Array already decoding motor-cortex activity into discrete commands.
A practical gap remains: no standard software bridge from neural command streams into trading terminals such as MetaTrader 5.
A reference prototype can be built today using simulated neural commands. A Python Flask service emits BUY/SELL/CLOSE/HOLD over JSON via HTTP, while an MQL5 Expert Advisor polls with WebRequest, parses commands, deduplicates execution, and routes actions through CTrade.
This simulation validates integration points, safety behavior (reset-to-HOLD), and transport latency without requiring BCI hardware or clinical access.
👉 Read | Docs | @mql5dev
524 415
Mamba4Cast targets real-time market forecasting where RNNs miss impulses and Transformers become too heavy. It combines Mamba state-space blocks (linear cost over sequence length) with Prior-data Fitted Networks, pretrained on many synthetic market-like tasks to enable zero-shot generalization across symbols and timeframes.
The pipeline normalizes inputs, adds timestamp-aware positional features (minute/hour/day/month/year) via multi-harmonic sine/cosine components, then uses causal and multi-kernel convolutions plus inception-style mixing before stacked Mamba blocks. It outputs multi-step forecasts in one pass, reducing autoregressive drift, with optional variance heads for uncertainty and regime classification losses.
The MQL5 implementation focuses on computing temporal encodings in an OpenCL kernel directly from timestamps, avoiding lookup t...
👉 Read | CodeBase | @mql5dev
524 415
Prop Firm Risk Dashboard is a lightweight, read-only account monitoring panel designed for prop-firm risk limits. Attach it to a single chart to monitor the entire trading account across any symbol and timeframe.
The panel displays balance and equity, floating P/L, today’s P/L based on equity change since the trading day start, daily loss usage versus a configurable daily-loss limit, max drawdown usage versus a configurable max-drawdown limit measured from a configurable starting balance, and current margin level. Status is color-coded (green/orange/red) as thresholds are approached.
Configuration inputs include daily-loss and max-drawdown percent limits, starting balance (0 uses current balance), warning and danger thresholds, and UI settings such as corner position, offsets, font, colors, and background. Day-start equity is stored in a terminal global var...
👉 Read | Calendar | @mql5dev
524 415
MQTTFive is an MQTT 5.0 client library for MQL5, delivered as an #include package for MetaTrader 5 expert advisors and scripts. It connects directly to MQTT brokers such as Mosquitto, EMQX, and HiveMQ, enabling outbound publishing of prices/signals and inbound command handling for EA control and status monitoring.
The implementation is pure MQL5 with an internal socket API and no DLL dependency. It supports QoS 0/1/2 with automatic retry, delayed publish queues, topic aliases, and flow control for receive quotas. MQTT v5 features include CONNECT/CONNACK properties (session expiry, packet limits, topic alias limits) and subscription options (no_local, retain_as_published, retain_handling). TLS/SSL, binary payloads, and UTF-8 are included.
Installation is file-copy based into MQL5/Include/MQTTFive/ and inclusion via MQTTClient.mqh. Core API: Connect, Disconnect...
👉 Read | NeuroBook | @mql5dev
524 415
ACEFormer extends time-series forecasting for noisy markets by pairing ACEEMD denoising (explicitly dropping the first IMF to reduce high-frequency noise while keeping turning points) with a Transformer distillation stack that mixes probabilistic attention for informative regions and standard self-attention for global context.
The article focuses on integrating the OpenCL probabilistic-attention kernels into an MT5-side module: CNeuronMHProbAttention. It inherits a residual two-convolution backbone, keeps internal components statically allocated for predictable lifetime, and centralizes setup in Init, including Top-K queries and random key counts computed from sequence length.
Key handling is practical: a single index buffer stores both sampled keys and selected queries sized by max(randomKeys, topK)*heads. Random key selection uses uniform stratif...
👉 Read | Calendar | @mql5dev
524 415
Small tweaks in neural-network training can explode runtime in MT5. Dropping the target error from 1e-3 to 1e-4 pushed a single-neuron trainer from under a second to ~80 seconds, showing how tighter tolerances amplify repeated cost evaluations.
The fix wasn’t GPU code or parallelism, but restructuring. By specializing the neuron for a known shape (2 inputs, 1 output) and replacing a generic Cost routine with a tailored Cost_2, the code avoids rescanning the training set multiple times per step and computes related errors in one pass, cutting runtime to ~18 seconds at the same precision.
Key takeaway for traders and MQL5 developers: profile training loops, reduce redundant passes over data, and specialize only where the model’s input/output structure is stable.
👉 Read | Calendar | @mql5dev
524 415
The article extends an MQL5 technique for editing chart text in-place: any OBJ_LABEL can be clicked and temporarily converted into an OBJ_EDIT, letting users change text without opening Object Properties.
A key fix is refining event handling so only one edit control exists at a time, avoiding multiple OBJ_EDIT instances caused by selection/deselection edge cases.
It also tackles the “can’t move while editing” problem by toggling OBJPROP_SELECTABLE on the fly: one click enables MT5-driven dragging, the next disables selection to allow text input, and release events are used to finalize or revert back to OBJ_LABEL.
The next step previews mouse-based resizing without direct mouse hooks by steering standard MT5 object events.
👉 Read | AppStore | @mql5dev
