KB ALGO LAB

Stop trusting a flattering backtest.

Replay your EA or cBot against genuine broker bid/ask tick history — not 1-minute bars pretending to be ticks — then charge the real commission and overnight swap, and stress-test what's left.

Download for Windows See pricing
5–10s
per backtest
2,772
backtests behind our numbers
MT5·MT4
+ cTrader / cBot
Tick
real bid/ask spread
kbalgolab — run EURUSD H1
› 4,812,553 real bid/ask ticks, replayed — drag to rotate
Net profit
+18.4%
Max DD
7.1%
MC 5th pct
+6.2%
Profit factor
1.42
Watch

Ninety seconds on why your backtest lied to you.

Then a four-minute walkthrough of the whole tool, if you want the detail.

The case, in 1:26
The full walkthrough, 4:01 — setup to reading a result honestly

Short cuts if you're in a hurry: 30 seconds · 15 seconds

The problem

Same strategy. Two very different stories.

Most backtests flatter you in three specific ways. Each one is the difference between a strategy that looks tradeable and one that actually is.

01
Fake ticks
Most testers interpolate 1-minute bars into pretend ticks, so your fills land at prices that never traded. We replay the broker's own bid/ask history, spread and all.
02
Costs left out
Spread alone is not the cost. Adding real commission and measured overnight swap removed 203 of 678 profitable results from our own sweep — and changed the ranking, not just the totals.
03
Tuned on the answer
Optimise on the whole window and you have measured your own hindsight. Every run here holds back the last 30% as an exam it never saw.
What the data says

We ran 2,772 backtests so you don't have to learn this the expensive way.

Every strategy across 9 pairs and 7 timeframes, two years of real ticks, with real commission and real swap charged. Share of combinations still profitable after costs:

H4
34%
D1
32%
H1
24%
M30
16%
M15
10%
M5
2%
M1
2%

M5 managed 8 out of 396. H4 is seventeen times better. Costs are charged per trade, so the faster you trade the more of your edge they eat — which is invisible in a cost-free backtest, and expensive to discover live.

How it works

Four steps to an honest number.

01
Point it at your terminal
Select your MT5, MT4 or cTrader install. Your EA and settings stay on your machine.
02
Pick pair and timeframe
One combination or every combination — the app queues them and works through the list.
03
Replay on real ticks
Genuine broker bid/ask history, with commission and swap charged exactly as your broker does.
04
Read the honest result
Out-of-sample score, drawdown, and a Monte Carlo range — not one lucky run.
Strategy Lab

No EA? Describe one.

44 ready-made strategies you can run instantly, or write your idea in plain English and have it built, checked and tested for you — on the same real tick data, with no MetaTrader involved.

  • Marked on data it never saw — the last 30% is held back as an exam, and the leaderboard ranks by that score.
  • Stops, targets and real position sizing — and if a bar could have hit both your stop and your target, we assume the stop.
  • Check our numbers elsewhere — export any strategy as Pine Script and run TradingView's own backtest against it.
You type
“buy pullbacks to the 50 EMA in an uptrend when RSI dips below 40, 2 ATR stop and 4 ATR target”
↓You get a tested strategy
✓ rules generated & checked against the engine
✓ test-run on real ticks before you see it
✓ saved to your library, ready to compare
New

Find a setup, then trade the whole library at once.

Two tools that build on the same real tick data — one to spot what's coiling, one to turn everything you've saved into a single EA.

Consolidation/breakout scanner
Scores every pair on Choppiness Index, ADX and a Bollinger-inside-Keltner squeeze to flag what's genuinely ranging, then reads the Donchian bands, band position and volume slope to call a breakout direction. Sorted so the tightest coils surface first — labelled as the heuristic it is, not a promise.
Combined portfolio EA
Pick your best strategies from the library and export them as one MQL5 EA — each leg risking its own fixed slice of the account, so the combined risk stays bounded no matter how many you bundle. Correlation-checked before export, so you're not stacking five versions of the same idea.
PROAssistant

It runs the test instead of guessing.

Ask why a backtest didn't survive going live, or what actually works on gold. It can run backtests on your own tick data and search all 2,772 reference results — so you get numbers, not opinions.

  • It won't invent a figure — every number comes from a real backtest or our measured findings. When it doesn't know, it says so.
  • Your data stays yours — backtests run locally; only the conversation leaves your machine.
  • Not a financial adviser — it analyses strategies. It won't tell you what to buy.
You ask
“my EA made 150% in backtest — is it any good?”
↓It checks, then tells you straight
⚙ backtesting donchian_breakout on XAUUSD H1
It made +155% — but the Monte Carlo puts the probability of ruin at 8.3%, and the worst 5% of orderings end at zero. Max drawdown 44.6%. Not a bad strategy, but at this size roughly one run in twelve wipes the account.
PROLive account report

The real numbers, day by day.

A calendar of what your account actually made or lost, synced straight from MT5, MT4 or cTrader. Attach one EA to any chart and it keeps itself current.

  • Daily P&L calendar — every trading day, green or red, at a glance
  • Win rate, profit factor, drawdown — the real numbers, not a highlight reel
  • Your password never leaves your machine — nothing stored or transmitted but results
+£3,412
this month, synced live
+412
−180
+905
+301
−94

Illustrative layout — your own report shows your account's real figures.

Publish it. Prove it.

Every backtest can be published to a shared, always-live leaderboard — real results from real traders, ranked by out-of-sample score. Nothing typed in by hand.

Download for Windows View the live leaderboard