MT5 Test Engine

Exported code

Indicator lab

Selected indicator

Average move after the alert, from its close and before costs (ATR of the alert candle) · dashed: random entries

Generated indicators

Also show the indicators that do not survive the correction for everything tried (DSR below 95 %)

#	Alert	Exit	Val. trades	Val. R/trade	Val. PF	Val. t	DSR	q	Exp. R/trade	Val. alerts	Alert t	Robustness	Stability

Frozen for forward test

Segment	Alerts	Per day	Per week	Buys / sells	Target first	Random	Avg R	Edge (R)	t	t adj.	Direction (R)

Trades with the exit

Segment	Trades	Win rate	Avg R	90 % range	Resamples ≤ 0	Total R	PF	Max DD (R)	DD incl. open	t

Strategy factory

#	Rules	Tr. trades	Tr. PF	Tr. R	Val. trades	Val. PF	Val. R	Val. DD (R)	Robustness	Curve

Daily P&L calendar

Segment	Days	Active days	Trades/day	Green / red days	Avg week	Best week	Worst week	Weeks ≥ +20 %

Weekly target (%)

Each closed trade counts on the day it closes (server time); “tr” is the number of trades. A dashed day had candles but no trades, and an amber line marks the day validation or the final test starts. Hover a day for details.

Replay results

Equity curve (per closed trade)

#	Type	Entry	Price	Exit	Price	Reason	R	Result	Balance

Automatic search

The engine combines indicator rules, improves them through evolution and only keeps the ones that also win on candles it did not use to search.

Strategy style

CPU cores: 1 of 1

Min. trades (training)

Min. profit factor (validation)

Max. drawdown in validation (%)

Min. robustness (%)

Min. trades per day (average, 0 = off)

Statistical strictness

Import GPU results (.json)

How your candles are split

The search only sees training. Validation discards what does not repeat. The final test is not touched until you request it for a specific strategy. With many strategies tested, one always wins by chance: the final test is what exposes it.

System resources

Build an indicator

The alert fires when the trigger happens on a closed candle and every filter agrees. Sell alerts use the mirror conditions.

Exit

Alerts

Pause after an alert

From hour (server)

For

Measure and protection

Target and stop (× ATR)

Max. candles

Commission (pips per trade)

Slippage (pips per fill)

Alerts are measured with a target and a stop of this size, against the same trade entered on any candle. The trades with the exit keep it as their protective stop and close after the maximum candles; with very few candles the exit hardly matters. Every trade also pays the commission (round trip) and the slippage twice, on top of the spread: stops are filled exactly at their price here, which a fast market does not do.

Automatic generator

Trading type

Combines triggers, filters, hours and pauses, evolves them on training and keeps only the indicators whose alerts also beat random entries on validation, whose trades with their best exit make money on validation, and that still do when a parameter moves one step. Robustness is the share of those neighbouring settings that pass too: it is not a probability of making money.

Min. alerts (training)

Min. robustness (%)

Statistical strictness

The generator tries around a hundred thousand combinations, so some pass validation by pure luck: with no edge at all, the best of the thousands reviewed on validation can reach t ≥ 3. A higher t lets fewer of those through. No threshold alone corrects for the number of tries, so the list hides what does not survive the deflated Sharpe for everything reviewed, and shows the false discovery rate (q) and the edge to expect on new candles.

Search method

Luck check

Runs the same search, with the same settings and for as long as the last run, on a copy of the candles whose directions were flipped at random: each keeps its size, gap, time and spread. No rule can predict direction there, so whatever it accepts is what luck alone produces with this much searching.

Walk-forward of the generator

Checks the whole process instead of one indicator: the generator runs on the first 40 % of training + validation, its best indicator is frozen and traded on the next window it never saw, and so on over 5 windows. The final test is not used.

Seconds per window

Compare with random search (doubles the time)

Export for MetaTrader

Early alert while the candle forms (provisional)

Warns as soon as the condition holds inside the candle (grey arrow, “PROVISIONAL”), then “CONFIRMED” or “CANCELLED” when the candle closes. Only the confirmed alerts are backed by the statistics above: a provisional one can vanish before the close.

Code shown

MetaTrader 5: File → Open Data Folder → MQL5\Indicators, copy the .mq5 there, compile it in MetaEditor (F7) and drag it onto a chart of the same timeframe. MetaTrader 4: the same with the .mq4 in MQL4\Indicators. Arrows mark the closed candle of each alert; pop-up, sound, push and e-mail alerts are switched on in its inputs. It sends no orders: its trades are virtual, so place the stop at the broker when you enter (the indicator draws it and warns the moment it is touched).

Account

Initial balance ($)

Risk per trade (%)

Spread (pips)

Real spread of each candle (CSV Spread column)

Each trade risks that % of the balance down to the stop loss. The spread is charged as in MT5: buys enter at the Ask and sells exit at the Ask.

Test Engine

Exported code

Indicator lab

Selected indicator

Average move after the alert, from its close and before costs (ATR of the alert candle) · dashed: random entries

Generated indicators

The same search on luck candles

Walk-forward of the generator

Frozen for forward test

Trades with the exit

Strategy factory

Daily P&L calendar

Replay results

Equity curve (per closed trade)

Favorites