Prop Firm Live Daily Breach Line is a lightweight MT5 chart indicator for prop firm evaluations and funded accounts. It tracks the remaining daily drawdown buffer in account currency and plots a dynamic Daily Breach Level on the active symbol, showing the exact price where current net exposure reaches the daily loss limit.

Accuracy is maintained by reconstructing the 00:00 server-time baseline from today’s deal history, reversing realized P/L, swaps, fees, and commissions for both entry and exit deals. A short history cache reduces load during rapid ticks. The reset rule supports Balance Only or the higher of Balance/Equity at 00:00, with a single GlobalVariable snapshot to prevent post-rollover equity spikes from shifting the daily floor.

The breach price uses net open volume on the chart symbol, then validates the estimate with a two-step OrderC...

👉 Read | Docs | @mql5dev

An MT4 indicator identifies bullish and bearish fair value gaps using three closed candles only, preventing repaint. Detection excludes bar 0 for both creation and mitigation, and each zone is tracked until mitigation.

A bullish gap forms when candle 1 high is below candle 3 low; bearish is the inverse. Zones are drawn as rectangles extending right until mitigated, with optional retention and recoloring after mitigation. Mitigation supports touch-based or full-fill rules; “touched” is only used under full-fill mode.

Zone state and prices are exposed via iCustom buffers, written on the candle 3 bar: state (0–3), top, bottom, and direction (+1/-1). Gaps can be filtered by minimum size in points or ATR fraction, with configurable history scan and maximum tracked zones. No alerts, trading logic, DLL, or WebRequest are included.

👉 Read | AlgoBook | @mql5dev

EMA trend-following system for XAUUSD on M30, built around a 20/50 EMA crossover filtered by a 200 EMA trend line. Signals are evaluated once per new bar using only closed candles via indicator buffer shifts 1 and 2, preventing repainting. Execution uses market orders at current bid/ask.

Trade management is single-position only: any open position or pending order blocks new entries. Trading is limited to a server-time window (default 07:00–19:00). Stop-loss is fixed as a percentage of entry (default 1%), with no take-profit and no signal-based exit. A daily schedule closes all strategy positions and deletes pending orders at a configured time (default 20:00).

Risk control uses balance-based sizing (default 1% risk per trade), deriving volume from SL distance with tick size/value and lot step, clamped to broker limits. Every order is routed through a ...

👉 Read | CodeBase | @mql5dev

A post-exit trade analysis script extends reporting beyond the close. For each position closed in the selected period, it scans subsequent M1 bars from the exit price and calculates the best move in the trade direction, the worst move against it, and whether the take-profit level was hit later, including minutes to reach it.

Stop-loss exits are summarized by how often price later reached TP, median time to do so, and median move after the stop. Take-profit exits report how far price continued after TP and how often it ran another 1R or more. Other exit types show median post-exit move and how often the original target was still reached. Values are expressed in R when SL exists, otherwise in points.

Output includes an on-chart summary panel, a per-position table in Experts (newest first), and an optional CSV (AfterExit_<login>.csv). The script reads hist...

👉 Read | Signals | @mql5dev

A regime indicator classifies every closed bar as trend up, trend down, or sideways, and shows the answer directly on the price chart. Each regime is drawn as a colored box spanning from the first bar of the stretch to the next regime change, and from the lowest low to the highest high in that interval. Blue marks up, red down, gray ranging; fills are blended with the chart background and user-defined colors.

A minimum of 3 bars is required before a box is printed, avoiding one- or two-bar noise. Short flips still appear in the lower window as ADX-based columns: up and sideways values plot above zero, down values below, with optional reference levels (25 by default). A status line reports regime, bar count, and ADX; ranging can also report weak ADX or quiet ATR.

Closed-bar history is stable: boxes and values do not repaint after a bar closes. For automati...

👉 Read | Forum | @mql5dev

Fixed-interval models force tick data into uniform time buckets, which removes microstructure transitions and compresses liquidity shifts into averages. Event-driven pipelines treat each tick as a first-class signal and only react on change.

SDformerFlow applies spiking dynamics and window-based attention to sparse event streams, aiming to recover price microvelocity and short-lived liquidity regimes. The design targets short-horizon forecasting, noise filtering, and order-flow feature extraction.

The architecture follows an encoder–decoder layout. A Spiking Feature Generator and Shortcut Patch Embedding convert sparse ticks into dense spatiotemporal tokens. STSF encoders stack spiking Swin-style blocks with 3D window attention plus patch merging for multiscale context, while preserving the time axis.

Training stability is addressed via membrane-...

👉 Read | Signals | @mql5dev

MetaTrader 5 subwindows behave differently from the main chart regarding indicator instance de-duplication.

Setting indicator_chart_window to a subwindow can allow multiple instances, creating cascaded subwindows. MT5’s usual “same settings, block duplicates” logic often does not apply when a new subwindow is generated.

A practical control is to set a unique indicator short name (IndicatorSetString with INDICATOR_SHORTNAME) and then scan chart windows/indicators via MQL5 to detect whether an instance already exists before creating another.

Subwindow management is scriptable: MQL5 can enumerate chart windows, list indicators per window, and enforce fixed subwindow height via indicator properties to prevent user resizing.

👉 Read | Signals | @mql5dev

An indicator is implemented to run exclusively via OnChartEvent, avoiding periodic recalculation through OnTick or OnCalculate.

Chart zoom is controlled through mouse wheel input, enabling zoom in and zoom out actions without additional UI elements.

The codebase is kept minimal, focusing on event handling and basic chart state updates for responsive interaction and low overhead.

👉 Read | AlgoBook | @mql5dev

Work continues on a minimal neuron class intended for building small neural network layouts without external ML libraries. Current implementation supports variable input counts and basic convergence tests, but lacks activation functions and chaining across layers.

Activation functions are framed as a practical constraint on gradient-based learning, with emphasis on how derivatives during backprop can stall training if the function is mismatched. ReLU is used as the first case: it zeroes negative values, while the derivative at x=0 is undefined and typically handled with an explicit convention in code.

An MQL5 indicator is outlined to plot activation curves on a chart. The implementation relies on CCanvas and chart change events to recreate a bitmap with correct dimensions, plus simple keyboard input to shift the view. Only a few offset parameters are int...

👉 Read | Docs | @mql5dev

Prop Firm Guard Panel is an on-chart monitor for prop firm challenges and funded accounts. It tracks rule headroom in real time: daily loss limit, maximum loss limit, and profit target. The tool is display-only and does not place, modify, or close orders.

The panel shows equity and balance, daily loss used vs limit with remaining room (based on start-of-day balance), max loss usage with remaining room and the equity floor, plus profit target progress. It also displays the server-time day start and the reconstructed start-of-day balance using today’s closed deals, keeping values consistent after terminal restarts.

Colors transition from green to orange to red as limits approach. Optional popup and push alerts trigger once when a configurable warning threshold is reached. Inputs cover initial balance, limit percentages, reset hour, warning level, layou...

👉 Read | Calendar | @mql5dev

This EA adds a clean measurement layer to M1-confirmed signals by separating what was known at confirmation from what happened afterward, eliminating look-ahead bias in scoring.

At the confirming M1 close, it freezes three normalized features scaled by candle C’s range (body fraction, breakout magnitude with a saturation cap, and elapsed time within D), then stores an equal-weight descriptive score with an explicit availability time and a validity flag.

Post-confirmation behavior is tracked independently: MFE/MAE are computed only from later M1 bars, excluding the signal candle itself to avoid counting pre-close movement as “after” information.

CSV exports distinguish snapshot vs completed records via a record_kind field, enabling reliable filtering (completed + Triggered + score_valid) for hypothesis tests. The design also documents practical pitfalls: hi...

👉 Read | NeuroBook | @mql5dev