What Is Backtesting? Definition, Limits and Live Data

Backtesting is the simulation of a trading strategy against historical price data to estimate how it would have performed. It produces an equity curve, a profit factor and a drawdown figure without a single order ever reaching a broker. Those outputs are a hypothesis about the strategy, not a record of it.

How it works

The tester replays historical bars or ticks, applies the strategy's entry and exit rules to each one, and records a simulated fill. Everything that separates the simulation from reality lives in the cost and fill model:

Net result per trade = Gross P&L − spread − commission − swap − slippage

Spread cost  = spread (in price) × contract size × lots
Swap         = swap rate × lots × nights held

Three settings decide how honest the output is:

• Modelling quality. MT4 and MT5 report it as a percentage. "Every tick" interpolates from 1-minute bars unless you supply real tick data; "control points" and "open prices only" are approximations that flatter intrabar strategies badly.
• Spread assumption. Most testers default to a fixed spread. Live spread widens at the rollover, on news and at the session open — exactly the moments many systems trade. A spread modelled at 0.8 pips and paid at 4.0 pips is a different strategy.
• Slippage and fill logic. The tester assumes your order is filled at the requested price. Slippage, rejected orders, requotes and partial fills do not exist in it.

There is a fourth problem the software does not report: the parameter set itself survived a search. If you ran 400 variants and kept the best one, the result you are reading was selected by hindsight. That is curve fitting, and it is the single largest gap between tested and live performance.

Why it matters

A backtest is the cheapest way to reject a bad idea and the most expensive way to confirm a good one. It can establish that a rule set is mechanically coherent, that it survives a rough cost assumption, and that its drawdown profile is something you could sit through. It cannot establish that the edge exists, because every input it uses is a reconstruction and every parameter it uses was chosen after the data was seen.

Anyone deciding where to put money should treat a backtest as a screening tool with a veto, never as evidence.

What the data shows

ShowMyTrades publishes 10,000+ accounts as public pages, resting on 15,436,464 synchronised trades (August 2026). Not one of them is simulated: every trade in that count was filled by a broker and read back from it.

That published population is overwhelmingly automated — median autotrading share 99%, with 53.9% of accounts running above 90% automation and 42.2% below 10%. The automated portion consists, in the main, of strategies that were backtested before they were funded. Across the whole published set the median profit factor is 1.28 and the median Sharpe ratio is 0.05. Those are the figures the tester reports were competing against.

Costs are the line a simulation most often understates, and here they are billed rather than assumed: $4,782,670 charged in commission and $862,547 in swap, with 86.3% of the accounts that carry any swap paying it net negative. A backtest set to zero swap on a strategy that holds positions overnight is not slightly optimistic; it is missing a recurring charge.

A tester with a fixed spread and no swap has left out two of the three.

Where you see it on ShowMyTrades

Nowhere, deliberately. ShowMyTrades does not host backtests, upload simulations or accept a strategy report as a track record. Every account page is synced read-only from the broker, which is why the numbers above describe executed trades rather than modelled ones.

The modules that expose what a backtest hides are: Total Commissions and Total Swap Paid in the Advanced Statistics block, the real cost of the strategy in account currency; the Terminal panel, which reports Ping and Retransmission to the trade server, the conditions your fills were actually obtained under; and Trade History & Balance Progression, whose Closed Trades tab lists each executed order at the price it was genuinely filled at rather than the price a tester would have granted it.

The Track Record Verified badge, held by 65 accounts, and Trading Privileges Verified, held by 264, mark the records where that broker-to-page chain has been checked rather than asserted.

Common misunderstandings

• "It used every tick, so it is accurate." Tick modelling improves the price path, not the fill. Your broker's spread, execution and rejections are still absent.
• "It was tested over ten years." Length is not independence. A decade that contained one dominant regime is one observation, not 2,500.
• "Only three parameters were optimised." Three parameters at twenty values each is 8,000 experiments. The winner of 8,000 experiments is partly lucky by construction.
• "The demo forward test confirmed it." Demo servers fill optimistically and never reject. A demo result is a second simulation, not a first live test.

Before you trust any performance figure, tested or live, work through how to verify trading performance claims.

Related terms