Who it is for
For traders and investors with their own rules
Traders and investors who have their own rules and want to know how they would have behaved on past data.
What we need from youHome / Strategy testing
Your trading rules, tested on past data with realistic costs
We write your rules down, code them and test them on past data. The report states every assumption and limit, including when the idea does not hold up.
The price is the same whatever the result.
Trading carries risk. The value of investments can go down as well as up, and you can lose money. Past or tested results are not a guide to future results. We sell software and testing only: no advice, no signals, no promise of profit. We are not authorised by the Financial Conduct Authority. Read the full notice.
In short
What you get
A written report and the code
A written report, your rule set written out in full, the test conditions, and full commented source code so anyone can re-run the test.
See the sample reportWhat a backtest can and cannot show
- It uses past data only, under the costs and assumptions the report states.
- It is not a forecast, and it does not say whether to trade.
- The report explains how costs, slippage and fitting to the past affect the result.
- Past or tested results are not a reliable guide to future results.
- The report states what the data showed, what it could not show, and where the data was thin.
See the sample report layout, with every figure blanked. It is a sample layout. It is not a real result.
How we test
Eight checks make up a full validation. A single strategy test includes all of them except walk-forward analysis and the price-path Monte Carlo, which come with full validation.
01
Out-of-sample data
Part of the history is set aside before anything is fitted. It is used once, at the end.
02
Walk-forward analysis
The rules are fitted on one window and tested on the period after it. Then the window moves on and the test repeats.
03
Realistic costs
Spread as recorded in the data where it exists, and a stated, cautious figure where it does not. Your own commission, a stated slippage figure, and a second run at higher costs.
04
Parameter sensitivity
Results are checked across a range of settings. Rules that only work at one exact value have been fitted to the past.
05
Monte Carlo analysis
Trade order and price paths are resampled separately. This shows how much of the result came down to luck.
06
Number of variants recorded
Test enough versions and one will look good by chance. So we report how many were tried.
07
Data integrity
No data that was not known at the time. For shares, the data includes companies later delisted, unless you choose otherwise, and the report says so on its first page. Gaps and bad ticks are handled, and every fix is recorded.
08
Stated assumptions
The report lists what the test assumed, what it could not account for, and where the data was thin.
We test your strategy with the same method we apply to the systematic strategies we develop: out-of-sample data, realistic costs and a written account of what the evidence does and does not show.
How a test runs
Seven stages, in this order.
01
Feasibility read
We read your rules and check the data exists for what you trade. You get a written answer on what can be tested, what it needs, and a price range. The fee is credited in full against work you commission within 60 days.
02
Your rules, pinned down
Your strategy is written out precisely enough to code from. You answer every awkward case, such as a gap, a partial fill, a stop and target hit in the same bar, or a lost connection. This starts once you accept the range and pay the deposit. You sign it off, and the fixed price follows.
03
Decided before we look
The test periods, the settings that may be tried, the most versions allowed and the measures you will judge it by are fixed in writing before we see any result. Nothing is changed afterwards to make a result look better.
04
Data and code checked
Every data set goes through integrity checks, and every fix is recorded. Trades are then checked by hand against your rules before any result is read.
05
Tested
The test runs the checks your package includes, under realistic costs.
06
Separate review
A review done apart from the build re-runs the test from the frozen inputs. It checks every number against your rules and the plan before the report reaches you.
07
Report and walkthrough
You receive the report, the rule set, the test conditions and the source code. We then go through it with you on a call.
What we need from you
A clear brief: your rules in any written form, the instrument, the timeframe, your entry and exit, your risk rules, and the platform and broker you use. Your commission schedule, if you know it. We never need your broker login, passwords, balances or statements.
What this service does not cover
- Finding or suggesting a strategy for you.
- Signals, tips or market opinions.
- Managing accounts or trading on your behalf.
- Advice on whether a strategy is worth trading.
What it costs
- Feasibility read
Your rules read and the data checked. A written answer on what can be tested, what it needs, and a price range. Credited in full against work you commission within 60 days of the written result, up to the price of that work. After that the credit lapses. It is never paid back as cash.£195 - Single strategy test
One instrument, one timeframe. Coded, tested on out-of-sample data with costs modelled, a parameter check, a trade-order Monte Carlo and a written report.From £450 - Full validation
Everything in a single test, plus walk-forward analysis, full parameter sensitivity, the price-path Monte Carlo and a full robustness report. Overfitting statistics where several variants were tried.From £1,450 - Multi-instrument or portfolio
The same rules across several instruments, or several rule sets tested together.From £1,950
Questions
What if the test shows my idea does not work?
Then the report says so, with the reason. It names the assumption the idea failed on, and whether it failed over the whole period or only in some conditions. The price is the same whatever the result.
How do you stop a test being fitted to the past?
The rules and the test design are fixed before any result is seen. The most recent data is kept back and used once, at the end. Every version tried is counted, and results are checked across a range of settings. If you built the rules on the same history, we tell you, because it cannot then count as unseen.
What data do you use?
It depends on the instrument and how precise the answer needs to be. FX and CFD work can often use your broker's own tick history, or licensed data. Shares usually need licensed data that includes delisted companies. We tell you what the job needs. Data is passed on at cost and agreed before it is bought.
Do I have to share my strategy with you?
Yes, in enough detail to code it. We offer a mutual non-disclosure agreement at no cost before you send anything. Our standard terms keep everything confidential either way. We do not trade, resell or reuse client strategies.
Do you use AI tools on my strategy?
Yes. We build and review with Claude, made by Anthropic, under Anthropic's commercial terms, which keep your material out of model training, and we tell you this in writing before you send anything.
Next step
Send us your rules
Send your rules with your enquiry, as clearly as you can, and we reply with the next step.
Contact us
Use the form or email us directly.
Tell us what you need and we will reply with the next step.