Honest Backtest Engine
Most backtests flatter you — good numbers in, good numbers out. This one runs your strategy on data it has never seen, charges real costs, stress-tests the result, and scores how much of the edge is likely real — in one run, without a subscription, without your strategy ever leaving your machine.
Not ready to buy? Try the lite engine free, right on this page ↓
The gap
This is a real run of a plain two-line moving-average crossover — nothing exotic, nothing cherry-picked. Watch what happens to the numbers once the strategy has to trade on data it never got to see first.
The strategy didn't change. The market wasn't unusual. This is what overfitting actually looks like from the inside — not broken, just quietly shaped around its own training data, right up until it meets data that isn't. Honest Backtest Engine headlines the right-hand number, by default, and there's no setting that turns it off.
Already see the problem? Get it — €149 →
See it in action
Opens on your machine as a local web app. Click any screenshot below to see it full-size.
01 Choose data → 02 Build signal → 03 Stack filters → 04 Run it
swipe through every screen →
Yahoo Finance, your own CSV, Interactive Brokers, or bundled sample data. Then build the entry signal by mechanism — cross, threshold, band or direction — not a black box.
Trend, momentum, oscillator, volatility, volume, band and statistical — 28 indicators, each with its own plain-English description, not a cryptic dropdown.
Optional filters by category, plus a support/resistance module with level source and price relationship. Every control explained, nothing hidden behind a toggle.
Use classic patterns as your entry signal or as a filter. And a live cost breakdown — entry fee, exit fee, slippage, borrow — on a sample trade, before you ever click run.
Position sizing, stop loss and take-profit — capped by what the instrument can really support, not a number you can quietly ignore.
Every check that passed, every one that didn't, and why — so a 70/100 is exactly as trustworthy as it should be, no more.
The full per-trade ledger and monthly profit and loss — nothing hidden behind the headline number.
Try it — free
One signal — a moving-average crossover — on three synthetic sample markets, scored by the same rules as the full engine: a fixed 70/30 out-of-sample split, real costs on every trade, and checks that get harder the more combinations you try. Pick parameters, hit Run, and watch what the score does.
Costs: 1 bp commission + 2 bps slippage, every side, every trade — always on. There's no off switch here either. Signals fill on the next bar, never the bar they were computed on.
Synthetic sample markets — deliberately not real data, same policy as the sample set bundled with the engine. This lite version is the flow, not the product: the full engine runs your data, locally, with 28 indicators, 39 candlestick patterns, filters, portfolios, position sizing, and custom Python strategies.
Like what it refuses to tell you? Get the full engine — €149 →
The problem
Give a strategy enough parameters and enough history to tune against, and it will show you a beautiful equity curve — then fall apart the moment real money is on it.
Overfitting. Look-ahead bias. Costs quietly assumed away. One lucky run dressed up as an edge. These are the ways a backtest fools the person who built it — and usually that person only finds out after the losses are real. The engine is built to find the flaw first.
What makes it honest
Every run ends in a single number — how believable the edge is, from 0 to 100 — and the exact checks behind it. Including the ones it didn't pass.
A real result from the bundled sample book. Two checks failed — so the score is 70, not 100. No other tool shows you the red marks. This one leads with them.
The lifetime license
€149 once — not per month
Pay once. Runs locally, forever. 30-day guarantee.
What it actually does
Scored on the slice of history it never tuned on — the only part that predicts anything.
Commission, slippage, borrow — charged on every trade. No frictionless fantasy.
Discounts the Sharpe ratio for how many combinations you searched to find it.
Reshuffles your trades thousands of times to separate a real edge from good timing.
Every signal uses only data that existed at that moment. No peeking at the future.
Scans for gaps, duplicates and bad bars before they distort the result.
Tamper-evident export. Verifiable by anyone — without trusting you, or us.
Your data
No data wrangling to get started. Three ways in, and sample data bundled so you can try everything the moment it opens.
Type a ticker — AAPL, BTC-USD, EURUSD=X — and pull years of history. No account, no API key, nothing to set up.
Already have price data? Point it at any CSV of bars and go. Your files never leave your machine.
Trade through IBKR? Pull history straight from your own feed for the instruments you actually trade.
A synthetic dataset ships inside, clearly marked as not-real, so you can explore every feature before touching your own data.
How it runs
On Windows it's a single download you double-click — nothing to install. It opens in your browser at a local address on your own computer. Your data, strategies and results stay entirely local.
Why it exists
Because a good-looking backtest is what sells the strategy, the course, the signal group.
We built this one to do the opposite: to find the flaw before the market does, and tell you what's likely true rather than what you want to hear. We sell exactly one thing — the backtester. No strategies, no signals, no subscription tied to keeping you hopeful — so there's no reason to make your backtest look better than it is. The honesty isn't a feature we added; it's the only reason this exists.
Questions before you buy? Email us — a person who works on the engine answers.
Where it stops
Stated plainly, before you pay for it.
The offer
| Three tiers — every one a single payment | Basic€79 | Most popularPro€149 | Institutional€499 |
|---|---|---|---|
| Honest core: out-of-sample split, real costs, trust score | ✓ | ✓ | ✓ |
| Built-in strategies | All | All | All |
| Open signal builder, filters, support/resistance | ✓ | ✓ | ✓ |
| Clean reports (no watermark) | ✓ | ✓ | ✓ |
| Walk-forward validation + optimizer | ✗ | ✓ | ✓ |
| Multi-asset portfolios & risk sizing | ✗ | ✓ | ✓ |
| Signed, tamper-evident reports | ✗ | ✓ | ✓ |
| Interactive Brokers data | ✗ | ✓ | ✓ |
| Your own Python strategies (look-ahead audited) | ✗ | ✗ | ✓ |
| Futures with margin-based sizing | ✗ | ✗ | ✓ |
| Scan & rank across whole watchlists | ✗ | ✗ | ✓ |
| Institutional-depth metrics: VaR/CVaR, drawdown anatomy, rolling consistency | ✗ | ✗ | ✓ |
| Trade-log CSV + full JSON exports | ✗ | ✗ | ✓ |
| Every future update included | ✗ | ✗ | ✓ |
| 30-day refund — keep your reports | ✓ | ✓ | ✓ |
| Get Basic | Get Pro | Get Institutional |
Updates, honestly: Basic and Pro are the version you buy — it runs offline, forever, and never stops working. Institutional includes every future version: when a new build ships, email your license key to [email protected] and we send it over. No auto-billing anywhere, ever. Prefer to feel it first? The lite demo runs free in your browser ↓
Learn more
Short answers, and the full reasoning if you want it.
Check whether it holds up on out-of-sample data, whether it survives realistic trading costs, how many parameter combinations you tried to get there, and whether the result survives resampling the trade order. Read the full breakdown of all 7 checks →
Usually one of five specific gaps: underestimated slippage, missing commission or borrow costs, a strategy tuned to fit the past rather than a repeatable pattern, market conditions that changed, or a lucky run mistaken for an edge. Read the full explanation →
Above 1 is generally acceptable, above 2 is strong, above 3 is excellent — but only if it's out-of-sample, net of real costs, and not just the best of many attempts. The number is easy to inflate. Read why the number can lie →
A statistic that discounts a strategy's Sharpe ratio for how many parameter combinations were searched to find it — correcting for the chance that the best of many tries just got lucky. Read how the correction works →
Building and tuning a strategy on one part of your data, then testing it unchanged on a separate part it never saw. It's the closest thing backtesting has to simulating not knowing the future. Read how to set it up properly →
A stricter version of out-of-sample testing that repeats the tune-then-test process across many rolling windows through time, instead of just once — closer to how a strategy is actually re-tuned while trading it live. Read how the windows work →
They scale directly with how often a strategy trades — a frequent strategy that looks great gross of costs can be a reliable loser net of realistic commission, slippage and borrow. Read the full breakdown →
No — your backtest and your live execution code are usually two separate implementations, free to quietly disagree on indicator timing, warm-up, or rounding even on identical data. Read how to verify they actually match →
Before you buy
No. Everything runs locally on your computer. Your price data, your strategies and your results never leave your machine. The only time it touches the internet is if you choose to pull data from Yahoo Finance or Interactive Brokers.
On Windows, download the file and double-click it — nothing to install. Mac and Linux are included too. It opens in your web browser at an address on your own machine. The first time on Windows you'll see the standard "unknown publisher" prompt (More info → Run anyway). That's Windows flagging that the binary isn't signed with a paid publisher certificate — it appears for most independent software. The app itself runs entirely on your machine and sends nothing anywhere, and if anything about that makes you hesitate, the 30-day refund exists for exactly that reason.
Any CSV of price bars you already have, plus Yahoo Finance by ticker (no account needed) and Interactive Brokers if you trade through them. Sample data is bundled so you can try everything immediately.
One payment, a license key that never expires, and a copy of the engine that runs offline forever — no subscription, no re-billing, no kill switch. Future versions: Basic and Pro are the version you buy; the Institutional tier includes every future update — when a new build ships, email your license key to [email protected] and we send it over.
30 days, no hoops. If it doesn't earn its place in your process, email [email protected] and you'll get a full refund — and you keep any reports you made.
Free checklist
The same checks the engine runs automatically — condensed into a short, practical checklist you can run against any backtest you've already built, by hand, today.
One email to confirm, then the checklist. Unsubscribe anytime.