KB ALGO LAB

Step-by-step guide

Every section of the app, in the order you'll actually use them. If you're starting from scratch, do sections 1 to 4 in order — after that, dip in wherever you need.

1. Settings 2. Tick Data 3. New Test 4. Reading a result 5. Strategy Lab 6. Scanner 7. Optimize 8. Cloud Runner 9. History 10. Leaderboard 11. Reports 12. Account Report 13. Assistant
Sections 1–4 are the first-run path. Nothing else works until section 1 is done.

First run — do these in order

01

Settings

Where you point the app at MetaTrader and sign in. Nothing else in the app works until this is done, so start here.

  1. Open the Settings tab (under Settings in the sidebar).
  2. Under MetaTrader 5 folder, click Browse… and select the folder containing terminal64.exe — for example C:\MetaTrader 5.
  3. Click Save & test. The app checks it can actually launch and write to that install, and tells you straight away if it can't.
  4. Under Your account, sign in with your email and password (or create an account on the website first).
  5. Your plan appears next to the heading. Free shows FREE; once you upgrade it shows Pro or Elite, which is what unlocks real tick data.
  6. If you want your finished runs to appear on the public board, set Your name and the Publish token under Publish to the Hub. Skip this if you'd rather keep results private.
If Save & test fails — the usual cause is MetaTrader installed under C:\Program Files, where Windows blocks writes. Version 1.0.15 and later handles that automatically; if it persists, reinstall MetaTrader to a normal folder like C:\MetaTrader 5. See the FAQ.
02

Tick Data

Gets the real bid/ask history onto your machine. This is what makes a backtest worth trusting — and it's also where you check which dates you actually have.

  1. Open the Tick Data tab.
  2. Pick a Pair, then set From and To. Leaving the full range selected is fine.
  3. Start the download. On Pro and Elite it streams straight from our server — already built, nothing to assemble yourself. If your connection drops it resumes rather than starting over.
  4. Watch Pairs with real ticks installed. Each pair appears there with its exact date coverage once it's ready.
  5. Note the coverage dates before you set up a test. All 18 hosted symbols run from 15 July 2024 to yesterday.
This is the fix for most failed runs. A test window outside your downloaded coverage is the single most common reason a backtest won't start. Check this tab first.

Running a test

03

New Test

The main event: run one EA across as many pairs and timeframes as you like, in a single batch, on real tick data.

  1. Open New Test.
  2. Under 1 Your files, add your Expert Advisor — drag the .ex5 onto the box or use browse files. Use the compiled .ex5, not the .mq5 source.
  3. Add a .set file if you have specific settings to test. Without one the EA's own defaults are used.
  4. Show installed EAs lists what's already in your terminal, if you'd rather pick from there.
  5. Under 2 What to test, click + Add pair and choose your symbols. Free is limited to one pair; Pro and Elite are unlimited.
  6. Choose Timeframes — M1 through D1, or select all. Every pair × timeframe combination becomes its own run.
  7. Set From and To inside the coverage you saw in section 2, plus your starting Deposit.
  8. Set Data quality to real ticks for a trustworthy result. Faster modes exist for a quick look, and results carry an honesty tag showing which you used.
  9. Start the batch and watch Running. Close MetaTrader first — the app launches its own instance, and one already open on the same folder will block it.
  10. Finished runs land in Results. If any failed, the reason is shown per run rather than just a count.
Start with one pair on H1 or H4 to confirm everything works end to end, then widen the batch. A 12-timeframe × 8-pair batch is 96 runs and will take a while.
04

Reading a result honestly

Not a tab — the skill that makes the rest worthwhile. Open any finished run and work down in this order.

  1. Check the honesty tag first. Real ticks or a faster modelled run? A modelled run is a rough indication, not evidence.
  2. Look at drawdown before profit. Anything above 100% means the account was mathematically wiped out — treat it as a failure no matter what the profit says.
  3. Check the trade count. Under about 50 trades, the result is mostly luck and won't survive.
  4. Compare in-sample against out-of-sample. Strong on the first period and weak on the later one is the classic signature of curve-fitting.
  5. Read the Monte Carlo range. It reshuffles your trades thousands of times. If the bad end of that range would have stopped you out, the headline figure is not the number that matters.
  6. Look at the costs broken out. Spread, commission and swap are itemised — that gap is the usual reason a vendor's backtest looks better than yours.
  7. Use the equity curve and trade markers on the chart to sanity-check that it traded the way you expected.
A strategy that only looks good on one ordering, on one period, with a handful of trades, is not an edge. Most published EA results fail at least one of steps 2 to 5.

Going further

05

Strategy Lab

Build and test strategies without owning an EA at all. Uses the built-in engine rather than MetaTrader, so it's much faster for comparing many ideas.

  1. Open Strategy Lab.
  2. Pick from the built-in strategy library, or describe your own idea in plain English and click Build it — it's turned into a real, testable configuration.
  3. Use Backtest it now for a single run, or Save to my library to keep it alongside the built-ins.
  4. To compare several at once, tick them under Select strategies to compare (or Select all).
  5. Under Where to test, choose Pairs — All pairs with data is the quickest option — plus Timeframe, From, To and Deposit.
  6. Open Advanced (optional) to set stop loss, take profit, trailing stop and position sizing.
  7. Click Run comparison. Stop halts it at any point.
  8. Results appear ranked by out-of-sample return — deliberately, so the ranking isn't led by curve-fitted in-sample numbers.
Anything you build here can be exported as Pine Script for TradingView, or as a real MQL5 EA you can run in MetaTrader.
06

Scanner

Scores every pair for how tightly it's ranging and leans toward a likely breakout direction — a heuristic to point you at what's worth a closer look, not a signal to trade blind.

  1. Open Scanner and pick a Timeframe.
  2. It scores every pair with built tick data on Choppiness Index, ADX, and whether Bollinger Bands have contracted inside Keltner Channels — three independent reads on whether a pair is genuinely coiling.
  3. Results are sorted by consolidation score, tightest first, so the coiled pairs surface at the top.
  4. Each row also shows a breakout bias — which Donchian band is closer, where price sits in the Bollinger band, and volume slope, combined into a leaning direction with a confidence score.
This is a scored heuristic built on existing indicators, not a trained or backtested predictive model — treat a high score as "worth watching," not a trade signal on its own.
07

Optimize

Searches your EA's input settings for better combinations, instead of you changing one number at a time.

  1. Open Optimize and select your EA and its Settings (.set) file.
  2. Under Parameters, choose which inputs to vary and their ranges. Fewer parameters with sensible ranges beats varying everything.
  3. Set Symbol, Timeframe, From, To and Deposit as usual.
  4. Set Search depth — deeper searches test more combinations and take longer.
  5. Click Run on my machine, or Run in the cloud to use our server instead and keep your PC free.
  6. Review All results, and use AI pick for a shortlist that favours robustness over the single highest number.
The top result of any optimisation is the most curve-fitted one by definition. Always re-check a chosen setting on out-of-sample data before trusting it.
08

Cloud Runner

Beta

Runs a backtest on our server rather than your machine — useful for long jobs, or if you don't want MetaTrader tied up locally.

  1. Open Cloud Runner and choose your Platform.
  2. Select the EA file to upload.
  3. Set Symbol, Timeframe, From, To and Deposit. The symbol box only offers what the server actually holds history for, so you can't queue a job that would fail.
  4. Click Upload & queue.
  5. Track progress under Your cloud runs; cancel from there if you change your mind.
Your EA file is uploaded to our server for this feature — unlike local backtests, where nothing leaves your machine. Elite accounts get priority in the queue.

Your results

09

History

Every run you've ever done, in one list.

  1. Open History to see Every run, newest first.
  2. Click any row to reopen the full report — charts, trade list, costs and Monte Carlo, exactly as when it finished.
  3. Use it to compare a re-run against the original after changing a setting.
10

Leaderboard

The shared board of results published by other users — and yours, if you choose to publish.

  1. Open Leaderboard for the All-time leaderboard.
  2. Tick only real-tick confirmed to hide anything from a faster modelled run.
  3. Tick hide < 50 trades to drop results with too few trades to mean anything.
  4. To publish your own, set your name and publish token in Settings first (section 1).
Use both filters. They remove the two easiest ways for a result to look better than it is.
11

Reports

Written research write-ups from our own testing — measured findings rather than opinion.

  1. Open Reports and pick any entry under Research reports.
  2. Use them for context on what's already been tested, so you don't spend an evening rediscovering it.
  3. Open the Portfolio report and click Build Combined EA (.mq5) to download the 8 selected, low-correlation legs bundled into a single Expert Advisor — each leg trading its own symbol and timeframe from one chart attach point, sized so no single leg risks more than its own slice of the account.
12

Account Report

Live sync in beta

Turns a real trading account into the same style of report as a backtest, so live results and tested results can be compared like for like.

  1. Open Account Report.
  2. For a one-off look, use Or import a one-off statement and load a statement exported from your terminal. This is the simplest route and needs nothing installed.
  3. For continuous updates, set up Live sync (beta) so the report refreshes as the account trades.
  4. On cTrader, use Copy cBot source under Using cTrader instead? and add that cBot to your platform.
  5. Read the result in Report, Equity Curve and Daily P&L. Use ← and → to move between periods.
  6. Keep anything worth returning to under Saved reports.
Live sync is still in beta. The one-off statement import is the dependable route today.
13

Assistant

Ask questions about strategies and results in plain English. It can run real backtests on your own tick data to answer, rather than guessing.

  1. Open Assistant.
  2. Type a question, or use a starter such as What's the best strategy on gold H4? or Is M5 scalping worth trying?, then click Send.
  3. Ask it to test something and it will run an actual backtest and answer from the result.
  4. New chat starts a fresh conversation when you change topic.
It works from measured findings and will say when it doesn't have the data rather than inventing a figure. It is not a financial adviser: it won't tell you what to buy or predict prices.
Stuck on something specific? The FAQ covers the exact errors people hit most, with the fix for each. Anything else — hello@kbalgolab.co.uk.