ORB — Opening Range Breakout for MetaTrader 5

A single configurable MQL5 expert advisor that trades the breakout of a session's opening range. Every parameter is an input — session, timezone, range length, signal timeframe, stop placement, target, stop management, risk — so one binary covers any session on any symbol without recompiling.

Runs identically on Windows and on Linux under Wine. The backtests in research/ are reproducible on either.

00:00 UTC ─── mark the 15-minute range ─── 00:15 ─── watch for an M1 close
              outside it, up to 00:29 ─── enter at market ─── stop at the range
              midpoint, target 2R ─── at +0.5R pull the stop to −0.5R ───
              close after 60 minutes if neither level is hit

Results in one table

XAUUSD, Asia session, MT5 real ticks, 2% risk per trade, one trade per day. Full analysis: research/report.html · research/FINDINGS.md

Period	Trades	EV / trade	Win rate	Median range
2026	88	+0.580 R	51.1%	1 108 pts

Results are regime-dependent. Gold's realised volatility ran above 50% in 2026 against a long-run norm near 15–16%, and the Asia opening range widened 6.4×. The full three-year breakdown, including the periods this configuration loses, is in research/report.html.

Also tested and rejected: London loses in every configuration; New York is flat at every range length despite having the widest ranges of the three sessions.

Install

Clone anywhere, then link or copy the sources into your terminal's MQL5 tree:

MQL5/Include/TimeZones.mqh
MQL5/Experts/ORB.mq5
MQL5/Scripts/TestTimeZones.mq5
MQL5/Scripts/CheckBrokerOffset.mq5
MQL5/Experts/BarDump.mq5          # research tool, optional

Linux/Wine — symlinks work and keep git as the single source of truth:

MT5=~/.wine_mt5/drive_c/"Program Files"/"MetaTrader 5"
ln -sf "$PWD/TimeZones.mqh"         "$MT5/MQL5/Include/"
ln -sf "$PWD/ORB.mq5"               "$MT5/MQL5/Experts/"
ln -sf "$PWD/TestTimeZones.mq5"     "$MT5/MQL5/Scripts/"
ln -sf "$PWD/CheckBrokerOffset.mq5" "$MT5/MQL5/Scripts/"

Windows — copy, or use mklink:

set MT5=C:\Program Files\MetaTrader 5
copy TimeZones.mqh "%MT5%\MQL5\Include\"
copy ORB.mq5 "%MT5%\MQL5\Experts\"
copy TestTimeZones.mq5 CheckBrokerOffset.mq5 "%MT5%\MQL5\Scripts\"

Compile

REM Windows
scripts\compile.bat

# Linux / Wine
scripts/compile.sh

Two things that will waste your afternoon otherwise:

• /compile: must be a path relative to the terminal directory, and you must be standing in that directory. An absolute C:\... path makes MetaEditor exit silently with no .ex5 and no log — it looks like a hang, not an error.
• MetaEditor returns exit code 1 on a clean compile. Judge success by the log's 0 errors, 0 warnings line, never by the exit code.

MT5 writes every log as UTF-16LE, so on Linux use iconv -f UTF-16LE -t UTF-8 before grepping.

Verify before trading

TestTimeZones      → expect "18 passed, 0 failed"
CheckBrokerOffset  → prints your broker's UTC offset and DST ruleset

TestTimeZones asserts every session open against a known UTC answer across every daylight-saving switch — Tokyo 09:00 must be 00:00 UTC on all dates, New York 09:30 must be 14:30 UTC on EST and 13:30 on EDT, and so on. It needs no broker connection. Run it first. If the clock layer is wrong, every range opens at the wrong minute and the backtest is meaningless.

Reproduce the backtest

Both platforms read the same tester.ini.

REM Windows — close the GUI terminal first, MT5 refuses two instances
scripts\backtest.bat

# Linux / Wine
scripts/backtest.sh

Set MT5_DIR if your terminal is not at the default path. Requires real tick history for the symbol — download it in the terminal before the first run.

Each closed trade appends a row to Common\Files\ORB_<symbol>_<magic>_tester.csv with the range size, spread, minutes-after-range, realised R and exit reason. That file is what the analysis scripts consume.

Inputs

Session

Input	Default	Meaning
InpTimeZone	TZ_UTC	which clock the session follows — UTC / London / NewYork / Tokyo / Sydney / Broker
InpStartHour, InpStartMinute	0, 0	session open, in that zone's local time
InpRangeMinutes	15	range length
InpSignalTF	PERIOD_M1	confirmation candle
InpNoEntryAfterMin	15	stop looking this many minutes after the range closes
InpForceCloseMin	360	flatten everything this long after the range closes (backstop)
InpMaxHoldMinutes	60	per-position cap, measured from its own fill

InpTimeZone answers "which clock does this session follow?", not "where is this asset from". Tokyo never observes DST, so Tokyo 09:00 and UTC 00:00 are the same instant forever — while NewYork 19:00 would drift an hour for eight months of the year.

Broker clock

Input	Default	Meaning
InpWinterOffset	2	hours the broker's server sits ahead of UTC in winter
InpFollowsUSDST	true	broker switches on US dates; false for EU dates

Find yours with CheckBrokerOffset, or compare a D1 candle's start time to UTC.

Entry

Input	Default	Meaning
InpEntryMode	market on close	market / stop resting at the level / limit retest
InpMaxTradesPerDay	1	counted from deal history, so it survives a reload
InpMaxSpreadPoints	0	skip if the spread is wider (0 = off)
InpMinRangePoints, InpMaxRangePoints	0, 0	absolute range filters (0 = off)
InpRangeLookback	0	rolling filter: sessions to compare against (0 = off)
InpMinRangeRatio	1.25	rolling filter: range must be this multiple of their median
InpMinClosePos	0.25	skip reversal breaks — see below (0 = off)
InpTradeMon … InpTradeThu	true	day-of-week filter
InpTradeFri	false	Friday is the only losing day — 20.8% win rate

InpMinClosePos scores how near price was, when the range closed, to the side it then broke — 1.0 means the final range bar closed right against that boundary, 0.0 means it closed at the opposite one and had to cross the whole range first. The second case is a reversal wearing a breakout's clothes: measured on real ticks, the five worst such trades in 119 averaged −0.94 R. Cutting everything under 0.25 removes 7% of setups and lifts pass rate from 81.6% to 86.6%. Full sweep in §12 of the report.

Stops, targets, risk

Input	Default	Meaning
InpSLMode	percent of range	percent of range / fixed points
InpSLPercentOfRange	50	50 = midpoint, 100 = opposite boundary
InpStopMoveAtR	0.5	move the stop once the trade reaches this many R (0 = off)
InpStopMoveToR	-0.5	where it goes, signed: 0 = entry, −0.5 = half the risk still on
InpTPMode	RR	RR / fixed points / multiple of range / none
InpRR	2.0	reward-to-risk multiple
InpLotMode	risk percent	fixed lots / percent of balance
InpRiskPercent	2.0	risk per trade
InpMaxDailyLossPct	3.5	stop opening trades once the account is down this much today (0 = off)
InpMagic	20260821	give each chart its own value when running several instances
InpWriteCsv	true	write the per-trade research log

InpMaxDailyLossPct sums every deal on the account since broker midnight, not just this EA's, plus open floating P&L. Running one instance per session means several EAs that cannot see each other — this is what stops the third trade of the day breaching a prop firm's daily limit.

What the study found

Ten rules, each measured rather than assumed. Detail and numbers in research/report.html.

1. Only the Asia session works, and only in the 2026 volatility regime.
2. Shortest range, fastest candle. 15 min / M1 beats 30 / M3 and 60 / M5.
3. Never tighten the stop inside noise. A quarter-range stop loses 0.27 R per trade; a breakeven stop-move turns a profitable configuration into a losing one.
4. RR 2, never RR 1. Roughly five times the expectancy.
5. A late breakout is a bad breakout. Breaks after minute 15 drag EV from +0.375 to +0.181.
6. Optimise pass rate, not profit factor — for a funded-account challenge they choose different parameters, and the highest-EV setup ranks eleventh.
7. Win rate is worth paying for, but only if it comes free — from better entries, not from a nearer target.
8. Range size is necessary, not sufficient. New York has the widest ranges of any session and no edge.
9. Speed is nearly free; risk sizing is not. More trades per day barely dents pass rate; changing risk per trade moves it 15–30 points.
10. Distrust anything inside ±0.05 R — the simulator's measured error against MT5.

Repository layout

ORB.mq5                    the expert advisor
TimeZones.mqh              broker time ↔ UTC ↔ session time, DST-aware
TestTimeZones.mq5          18 assertions over the time model
CheckBrokerOffset.mq5      measures your broker's offset and DST ruleset
BarDump.mq5                records M1 bars for offline research
tester.ini                 backtest configuration, both platforms
dump.ini                   bar-recorder configuration
scripts/                   compile and backtest, .sh and .bat
research/                  the analysis harness and write-ups
  report.html              full research report
  FINDINGS.md              same content as markdown
  sweep.py                 offline replay engine, validated against MT5
  mt5paths.py              locates Common\Files; override with MT5_COMMON

The offline harness exists because the Strategy Tester needs ~90 seconds per real-tick pass over 2.6 years, so 600 configurations would take hours. sweep.py replays recorded M1 bars instead, and agrees with MT5 to within 0.03 R. Use it to shortlist, MT5 to confirm.

On Windows, point the analysis scripts at your data folder:

set MT5_COMMON=%APPDATA%\MetaQuotes\Terminal\Common\Files

Requirements

• MetaTrader 5 build 4000+ (developed on 6090)
• Real tick history for your symbol
• Python 3.9+ for the research scripts — standard library only, no dependencies

Not financial advice

This is a research artefact. The headline result rests on 119 trades in a single year on a single symbol, with no out-of-sample data left. Demo test thoroughly, and read §13 of the report on where the numbers are soft before risking anything.

Licence

MIT