A configurable opening-range breakout EA, and the tooling to backtest it. Any session, any range length, any signal timeframe, any symbol, with correct daylight-saving handling.
python3 orb.py compile # build the EA
python3 orb.py start # control panel at http://127.0.0.1:8765Rules, results and limits live in each configuration's own report, generated from the run:
| Report | |
|---|---|
| ORB + NQ | the two that carry the result |
| ORB Asia | gold, 00:00 UTC opening range |
| Previous-day fade | gold, Asia session |
| NQ range fade | US100, New York session |
| All three together | pooled stats |
Account rules behind the pass rates:
fundingpips-1step-flex.md.
Copy mql5/ into your terminal's data folder (File → Open Data Folder),
compile, attach to an XAUUSD M1 chart. The panel starts OFF.
| From the zip | Goes to |
|---|---|
mql5/ORB.mq5 |
MQL5/Experts/ |
mql5/CheckBrokerOffset.mq5 |
MQL5/Scripts/ |
mql5/TimeZones.mqh |
MQL5/Include/ |
mql5/Panel.mqh |
MQL5/Include/ |
python3 orb.py compile # or press F7 in MetaEditorReleases ship source, not a compiled .ex5.
python instead of python3 on Windows. That is the only difference.
python3 orb.py start # control panel at http://127.0.0.1:8765
python3 orb.py stop
python3 orb.py status
python3 orb.py restart
python3 orb.py open
python3 orb.py logs
python3 orb.py start --local # use this machine's MetaTrader, no Docker
python3 orb.py start --slots 4 --port 8080
python3 orb.py run strategies/asia-gold.toml 2026.01.01 2026.09.11
python3 orb.py runs # what MetaTrader did on each run
python3 orb.py compile
bash update.sh # new days only: test, rebuild, audit, commit
bash update.sh --account accounts/ftmo.envNothing to copy or paste: the panel's token is generated per start and the page
fetches it itself. It is there because a site you visit can POST to localhost.
| Panel tab | |
|---|---|
| Backtests | every strategy with its last result; run, edit, duplicate, delete |
| Accounts | add or remove a broker login |
| Runs | the queue, and what MetaTrader did |
Accounts are written to accounts/<label>.env, mode 600, ignored by git. New
strategies land in strategies/generated/.
Reach the panel from a phone over Tailscale. Never a port forward.
python3 orb.py start --local
python3 orb.py run strategies/asia-gold.toml 2026.01.01 2026.09.11| Terminal found at | |
|---|---|
| Windows | C:\Program Files\MetaTrader 5\terminal64.exe |
| Linux / Wine | ~/.wine_mt5/drive_c/Program Files/MetaTrader 5/terminal64.exe |
| Override | MT5_TERMINAL=... |
Close the terminal first — it holds the tester lock.
docker build -f docker/Dockerfile -t orb-mt5 . # ~15 min cold
python3 orb.py start # containers are the default
bash research/farm.sh --slots 3 --env accounts/ftmo.env strategies/*.toml
bash research/farm.sh --dry-run --slots 3 strategies/*.toml| Volume, per slot | |
|---|---|
orb-bases-N |
price history, 2.1 GB, seeded once from this machine |
orb-tester-N |
the tester lock and the agent's history copy |
orb-config-N |
account and tester.ini |
orb-common-N |
where the run writes its CSVs |
docker volume rm orb-bases-1 orb-tester-1 orb-config-1 orb-common-1 # reset a slotThree slots on 8 cores with 5 GB free — each tester agent wants a core and a
gigabyte. tester not started because the account is not specified means the
image works and only the login is missing. Rebuild the image only for the
Dockerfile or the MetaTrader binaries; the repo is bind-mounted.
| Control | What it does |
|---|---|
| TRADING ON / OFF | Master switch. Starts OFF. Off stops new entries but keeps managing an open trade |
| Risk per trade | The number, plus a % / currency button: percent of balance or fixed cash |
| Reward : risk | The target, in R |
| Session start | HH:MM UTC |
| Range length | Minutes |
| Break window | Minutes after the range closes that a break still counts |
| Stop move | ON / OFF, with its two levels beneath |
| Yesterday filter | ON / OFF. Skips a break against the previous daily candle when its body is at least Min body. Off by default, not in the published numbers |
Settings lock while trading is on, and persist across restarts.
Defaults are the tested configuration.
| Input | Default | Meaning |
|---|---|---|
InpStartHour / Minute |
0 / 0 |
Session start, 00:00 UTC |
InpRangeMinutes |
15 |
Range length |
InpNoEntryAfterMin |
15 |
Stop looking at 00:29 |
InpMaxHoldMinutes |
90 |
Flat after 90 minutes |
InpMinClosePos |
0.50 |
The half-of-the-range rule |
InpTradeFri |
false |
Friday off |
InpYdayFilter / MinBody |
false / 30 |
Skip breaks against yesterday's daily candle. Measured, not published |
InpSLPercentOfRange |
50 |
Stop at the midpoint |
InpRR |
2.0 |
Target at 2R |
InpStopMoveAtR / ToR |
0.5 / -0.5 |
At +0.5R, move to −0.5R |
InpLotMode |
percent | percent of balance, or fixed cash |
InpRiskPercent |
2.0 |
Risk per trade |
InpWinterOffset |
2 |
Your broker's winter offset from UTC |
Check for your own broker: InpWinterOffset and InpFollowsUSDST. The
CheckBrokerOffset script measures both. Lots round to the nearest step, so
realised risk lands within ~9% of target.
- MetaTrader 5 build 6000 or newer, and M1 real-tick history for your symbol
- Windows, or Linux with Wine
- The tester cannot see today. It clamps the range you asked for and says so
in the run record.
SyncDump.mq5+research/sim_offline.pycover the current session. - One terminal at a time — it holds the tester lock. Containers get a slot each; local runs need the terminal closed.
- A new symbol's first run imports every month of ticks before it tests.
InpFollowsUSDSTis unverified; it matters for about a week each October.
Not financial advice. Test on demo first.
MIT — see LICENSE.