TrueCostReport - MetaTrader 5용 expert

Cost is the one input to a trading system that is knowable exactly, and it is almost always estimated wrongly — always in the direction that flatters the strategy.

This Expert Advisor measures what one round trip actually costs on your broker and your symbol: the live spread sampled from ticks, the swap for both sides including the triple-charge day, and the commission read from your own closed deals. Then it puts the number your tools are probably using next to it.

It is strictly read-only. It never opens, closes or modifies a position.

The measurement it exists for

The usual source of a spread figure is the spread field of a price bar, because that is the one thing easy to reach from a script or a backtest. That field is a per-bar summary, not the quote at any instant.

Measured on XAUUSD: the live spread was 20 points across 60 consecutive readings while the M1 bar field had a median of 6. A factor of 3.3.

Anything built on that field inherits the error:

• A spread filter set to "skip the entry if spread > 15" never fires, because the number it reads is a third of the truth. It lets through precisely the moments it exists to block.
• A backtest charging that spread reports a system cheaper to run than the real one, and the gap grows with the number of trades — so it flatters high-frequency strategies most.

The half that most simulators do not charge at all

Swap. It is asymmetric — one side usually pays while the other collects — and one weekday is charged three times. On gold, per 0.01 lot per night, a real account gave −0.5842 long and +0.4098 short, with Wednesday tripled.

On a system that holds positions overnight this is not a detail. A basket of four legs each way costs roughly 1.9 a night simply to exist, which over a year exceeds most of the edges people are trying to capture. MT5's own Strategy Tester charges swap; a great many custom simulators do not, and their results are wrong by the whole of it.

Three things it does that a quick version would get wrong

• It counts distinct quotes, not just ticks. A feed that repeats the same value hundreds of times has been decimated somewhere, and a median over those repeats describes the decimation rather than the market. If fewer than five distinct quotes are seen, it says so.
• It refuses to measure a closed market. With the market shut the terminal keeps handing out the last tick it saw, so a sampling loop happily collects hundreds of identical readings of the frozen weekend quote — on the symbol this was written for, 50 points against a working-hours median of 20. The test uses the trap it documents: TimeCurrent() is the time of the last tick and freezes with the market, while TimeTradeServer() keeps running, so the gap between them is exactly how stale the quote is.
• Commission comes from your deals, not from a symbol property. Commission is an account term, not a symbol one: two accounts on the same broker and the same symbol can pay differently, and there is no field to read it from. Zero is a legitimate answer on a spread-only account, and is reported as zero rather than as unknown.

Inputs

On the swap mode. SYMBOL_SWAP_LONG and SYMBOL_SWAP_SHORT are quoted in whatever SYMBOL_SWAP_MODE says, and the modes are not interchangeable — reading a points value as if it were money produces a confident wrong number. Points and the deposit-currency modes are converted exactly; the interest and reopen modes depend on the position price and on the broker's year convention, so they are reported as not converted rather than guessed.

What it does not do

It does not model slippage, which cannot be measured without trading, and it says nothing about whether a strategy is viable. It measures the toll. What you carry over it is your business.