MetaTrader 5 Build 6140: Critical Fixes for Systematic Backtesting and Math Libraries

Per the release notes, three changes carry weight for systematic traders: a correction to triple-swap calculations in the strategy tester, a restructured OpenBLAS linear algebra dependency, and tightened proxy protocol support. The AI Assistant enhancements and mobile UI updates get the marketing spotlight. Skip those — here's what actually moves your backtest numbers.

The triple-swap fix is the only item that matters this week

Build 6140 fixes an error in triple-swap calculations in the strategy tester, triggered under certain combinations of testing conditions. MetaQuotes does not enumerate which combinations in the notes I have. For anyone running carry strategies, grid systems on swap-positive instruments, or multi-day optimization passes, the fix is direct: prior equity curves may have been quietly mispriced. Run your historical tests against this build and diff the equity curves. Any drift beyond broker spread noise is signal. Re-optimization is not warranted on its own — you need a quantified drift before you touch a parameter.

OpenBLAS is now a separate DLL

The OpenBLAS linear algebra library is now distributed as openblas.dll rather than bundled inside the terminal binary. The retail impact is zero. The systematic impact is significant. A separate dependency means you can pin versions, swap implementations if you need MKL or a tuned build, and isolate numerical drift between MT5 updates. For anyone running matrix-heavy indicator code or custom MQL5 math, this is infrastructure cleanup that has been overdue. A segfault in your matrix routines now points to a specific DLL instead of the terminal — small thing, real diagnostic value.

Proxy hardening and dropped SOCKS4

HTTP and SOCKS5 proxy support is improved. SOCKS4 is removed entirely. If you run MT5 instances behind a VPS proxy, a corporate firewall, or a redundancy layer that fans connections across endpoints, expect fewer reconnect events. Fewer reconnect events during volatility windows means fewer missed fills. It is plumbing, not strategy — but plumbing at the execution layer compounds over a year of live runs.

Algo Forge and the mobile features

The MQL5 Algo Forge user guide is published, with a Git-style fork and clone workflow for shared MQL5 projects. Useful if you collaborate on EA code. The mobile iOS app got additional chart features and stability fixes. Neither affects backtest integrity or execution quality.

What I'm watching

Two things. First, I want to see community reports isolating which instrument and timeframe combinations triggered the swap bug. If you are already running optimization jobs on Build 6140, document the build version in your result headers. Second, the tester now auto-shifts the actual start date forward when available history is insufficient for the configuration you specified, and logs the adjustment. That is a behavior change, not a bug. If your optimization runs assume the requested start date was honored, audit your tester logs. The tester will tell you — but only if you are reading them.