In short. The MetaTrader 5 strategy tester shows how an EA would have traded on history — and how close that result is to live trading depends on the test settings. The key one is the tick generation mode: on the same history the breakout EA in a MetaQuotes example made +169 in «1 minute OHLC» mode and −97 on real ticks. Test on real ticks, with delays and your broker's commission, keep a forward period and confirm the result on a demo or a small live account.
Tick generation modes
| Mode | Where prices come from | Use it for |
|---|---|---|
| «Every tick based on real ticks» | ticks collected by the broker — from exchanges and liquidity providers; if a minute has no ticks, the tester generates them as in «Every tick» mode | the final check, scalping, EAs with stops of a few pips |
| «Every tick» | ticks are generated inside each minute bar: OHLC prices and intermediate ones, by tick volume | when the broker has few real ticks |
| «1 minute OHLC» | only the four prices of each minute bar | quick draft runs and optimization |
| «Open prices only» | only bar open prices | EAs that make decisions once per bar |
| «Math calculations» | no history is used | calculations without trading |
How far the modes diverge: a MetaQuotes example
In the MetaQuotes article on testing strategies on real ticks, the SimpleBreakOut breakout EA with default settings was run on EURUSD H1 from 1 January to 30 June 2016:
| Mode | Net profit | Trades |
|---|---|---|
| «1 minute OHLC» | +169.46 | 96 |
| «Every tick» | −466.81 | 158 |
| «Every tick based on real ticks» | −97.24 | 156 |
A trade is an entry and an exit together, so these reports show twice as many deals: 192, 316 and 312. One EA, one set of settings, one period — and results from profit to loss. The more precisely an EA works inside a bar (short stops, trailing, breakout entries), the more it depends on the mode. Base your decision on the real-tick run.
What else separates a test from a live account
- Spread. When ticks are generated inside a minute, the tester uses the spread fixed in the minute bar. On real ticks the spread changes within the minute too — closer to what you will see on news.
- Execution delay. The settings offer «No Delay», «Random Delay» and «Fixed Delay». Random Delay holds execution by 0–8 seconds with 90% probability and by 9–18 seconds with 10%: this shows whether the EA survives slow execution.
- Commissions. Turn on «Use predefined commissions» and the tester takes the commission settings of your trading account.
- History quality in the report is the share of correct minute data. Segments below 50% are marked red: the result there is unreliable and the history is worth reloading.
Curve fitting and the forward test
Optimization tries thousands of setting combinations and almost always finds one that passed the past perfectly. To check whether it is curve fitting, keep part of the history unseen by the EA: in the «Forward» field choose «1/2», «1/3», «1/4» or «Custom» with a start date. The tester picks the settings on the first part and runs them on the forward period. If the forward result is much worse, the settings are fitted to the past. How to judge the result — in «How to evaluate a trading strategy».
Test checklist
- «Every tick based on real ticks» mode, history downloaded, no red segments in history quality.
- The tester account has the same deposit, leverage and currency you will use.
- «Random Delay» and «Use predefined commissions».
- A period of several years, with trends, ranges and major news days.
- A forward period of at least a third of the history.
- Equity drawdown, not only balance drawdown: for grids and martingale the gap is huge — «Grid EAs and martingale: how to calculate the risk».
- More than a hundred trades: on 20–30 trades luck cannot be told from a system.
The last step — a live account
Even the best test is history. The final verdict comes from an account where the EA trades now: a demo or a small live account with the same settings.
How Davinchu helps
Davinchu turns such a run into a ready report:
- Account analytics calculates itself: profit factor, expectancy, winrate, Sharpe ratio, Z-score, AHPR and GHPR, best and worst trade, commissions and swaps. Compare profit factor, expectancy and the number of trades with the tester report directly; services calculate Sharpe and Z-score differently — how to read them.
- Equity drawdown — from hourly snapshots including floating losses, like «Equity drawdown» in the tester report.
- The monitor watches the EA: «The EA is silent» if there are no trades for longer than your threshold, and «Account connection lost» after 10 minutes without a connection — in Telegram, by email or in Discord.
- The Davinchu EA only reads the account — it opens no trades and does not interfere with your EA.
An account history HTML report can be quickly broken down in our MT5 report analyzer — the file is parsed in the browser and sent nowhere. The Free plan connects two accounts at no cost. What else Davinchu does for robots — on «Davinchu for algo traders».