Backtest your trading idea with ChatGPT or Claude. No coding.
Explain your idea in plain English. Your AI writes the code. Then a 10-point check tells you if the result is real, or if your AI is fooling you.
- 38-page playbook
- AI skills
- Copy-paste prompts
- Sample code
An illustration, not real numbers. Hypothetical backtest. Education only, not financial advice.
You've got a strategy you've never really tested. Or you don't know where to start.
- ✕"I don't know any programming language."→ The prompts tell your AI what to write. You paste and click.
- ✕"It ran. The chart looks amazing. Is it real?"→ The 10-point check. Ten things to look at in the output. If any one fails, the run is no good.
- ✕"Where do I even get clean price data?"→ It's built into the free account. Stocks, ETFs and crypto. You download nothing.
- ✕"Anything I find will just be a fluke."→ The backtest gate: first half, second half, double the fees, nudge the settings. Real edges survive.
- ✕"I don't even know what to test."→ Start with the worked sample and 4 ready-made templates. Change one rule. Run it.
If you nodded at even one of those, keep reading. The rest of this page is for you.
The chart looked amazing. 12,622 of its orders never happened.
While building this kit, we had an AI write a crypto backtest. The chart looked great.
Then we ran the 10-point check. Check 8 reads the order list. 12,622 orders had been rejected. A fee setting left the account a hair short, so the sell orders failed. Nothing on the chart showed it.
The strategy on the screen wasn't the strategy in the code. One check, two minutes, and a fake result became a known bug with a fix instead of a bad trade.
Your AI is fast and sure of itself. It won't tell you when it's wrong. The 10-point check will.
Why I built this
I've spent 300+ hours fixing backtest errors. Rejected orders, fee settings, data gaps, the same crash over and over. Every one of the 41 bugs in the playbook is one I hit, with the fix that worked. I'm hoping it saves you most of those hours.
Backtest Lab: the playbook, AI skills and prompts that turn your AI into a backtesting helper
A 38-page step-by-step playbook plus AI skills and copy-paste prompts for ChatGPT, Claude, Codex or Gemini. Same AI you already pay for. The skills give it a plan to follow.
- Your idea becomes exact rules before any code gets written.
- Your AI writes the backtest. It runs on a free QuantConnect account, data included.
- The 10-point check tells you if the result is real. You never read a line of code.
- 41 known bugs with fixes end the crash, fix, crash again loop.
Three steps
Turn the idea into rules
Prompt 1 asks you one question at a time until "buy the dip" is a set of exact rules.
Your AI writes the code
Run it on a free QuantConnect account, with the data already there. The AI skills keep it on track.
Check the result
Order list, fees, dates, first trades. The 10-point check tells you what each one should look like.
The real thing, sped up. Recreated from a real run of the included sample strategy. Hypothetical backtest. Education only, not financial advice.
Setup is about 10 minutes of sign-ups. The included sample strategy runs in a few minutes after that.
15 chapters and a folder full of tools. Tap any part to look inside.
The playbook teaches you. The AI skills and files do the heavy lifting inside your AI.
The playbook, in 4 parts
Part 1 · Night oneGet it running
4 chapters
- StartHand the kit to your AITen minutes of sign-ups. Then your AI takes it from there.
- 1Why most beginner backtests lieThe 3 stages every new tester hits, and how to skip the painful middle one.
- 2Day 1: get set upOne AI plan, one free account. That's the whole shopping list.
- 3Your first backtestRun a real strategy from our own research. Not a toy made up for the book.
Part 2 · Know if it's realTrust the result
4 chapters
- 4Read results like a researcherWhat to look at first, what to ignore, and every word on the screen.
- 5The 10-point checkTen questions that catch a backtest that's lying to you. Then the gate: halves, double fees, nudged settings.
- 6Known bugs and fixes41 errors we already hit, with the fix for each. So you don't pay to find them.
- 7Save your AI planStop paying for the same bug twenty times. Six habits that keep chats short.
Part 3 · Your ideas, testedBuild your own
4 chapters
- 8Your research systemIdea cards, test records and a weekly routine that fits around a job.
- 9Where ideas come fromA two-minute filter that throws out most ideas fast. Plus free research papers.
- 10Avoiding overfittingThe lucky-coin problem, and seven ways to keep it from fooling you.
- 11Risk and position sizingThe least exciting chapter. Maybe the most important one.
Part 4 · Real money, lastGo live slowly
4 chapters
- 12The going-live ladderFive rungs from backtest to real money. Paper first, then small. No skipping.
- 13Your first 30 daysA week-by-week plan. If your first idea fails the test, that's a win.
- 14–15Reading list and glossaryWhere the pros learned this, and every term in plain English.
- +ChecklistsOne-page checklists and a coach card for when you only have the PDF.
The files your AI uses
Inside your download folder8 items
- ✓The AI skills
AGENTS.md · SKILL.mdYour AI loads these and follows the plan, step by step. You never have to open it. - ✓The bug list
coach/known-bugs.md41 bugs, each with the exact error message and the fix. Your AI checks here before it guesses. - ✓The 10-point check
coach/verify.mdThe check, written for your AI. It fills in the table, you read the answers. - ✓The results reader
tools/qc_results.pyLets your AI pull the stats, orders and logs itself. No pasting walls of text. - ✓3 audit prompts
prompts/Code audit, result audit, experiment review. A tough second opinion from a fresh chat. - ✓Strategy templates
templates/algorithms/Trend crossover, RSI buy-the-dip, momentum rotation and a walk-forward tester. Start from one and change a rule. - ✓The sample strategy
Sample projectA full crypto momentum system: top 40 coins, weekly picks, a Bitcoin trend filter, real fees on. - ✓A research folder
templates/research-repo/Idea backlog, strategy list and error log, ready to fill in. Plus the go-live checklists.
Plus a library to grow into
You never have to read the code. But it's all there, plain and commented, if you get curious.
One download. Everything for your first and hundredth backtest.
- The 38-page step-by-step playbook (PDF)Setup, first run, reading results, going live slowly. Plain English.
- AI skills + copy-paste promptsIdea to rules, write the code, read the output, get a skeptical second opinion. ChatGPT, Claude, Codex or Gemini.
- The 10-point check + the backtest gateCatches the ten ways a backtest lies, then stress-tests what's left.
- 41 known bugs with fixesThe exact error, why it happens, the fix. Your AI reads this first.
- Worked sample strategy + 4 templatesA real crypto strategy, plus trend, buy-the-dip, rotation and walk-forward templates for stocks and ETFs.
- Research systemIdea cards, test records and a weekly routine that fits around a job.
Cheaper than one bad trade. Way cheaper than a coding bootcamp.
The First Backtest GuaranteeRun the sample strategy on your QuantConnect account. If it won't run cleanly (zero rejected orders, passes the 10-point check), you get every cent back. Ask any time in the first 30 days. No forms.
This is for you if…
Get it if you…
- have a day job and trade on the side, with a strategy you've never really tested
- don't code, or would rather not
- already tried "ChatGPT, backtest this" and hit a crash loop or a result you couldn't trust
- would rather find out an idea is broken in a backtest than in your account
- trade stocks, ETFs or crypto
Skip it if you…
- want signals, tips or a bot. There are none here, and no trade recommendations.
- want a strategy handed to you. Plenty of ideas should fail a good test. That's the point.
- already do this for a living with your own tools
- don't want to pay for an AI plan or sign up for a free QuantConnect account
Two things you should know
Your questions, answered
Will this make me money?
I can't promise that, and you should be wary of anyone who does. The playbook teaches you to test ideas. The biggest way it can save you money is by saying "no" before your account does. Education only, not financial advice.
Why not just ask ChatGPT for free?
You can, and you'll get code. What you won't get is the list of bugs it's about to make, a rule that stops the crash loop, or a way to check its answer without asking it. Same AI, better questions.
Is my strategy private?
Yes. Everything runs in your own AI chat and your own QuantConnect account. Nothing is sent to us.
Is it only for crypto?
No. The worked sample uses crypto data. The templates cover stocks and ETFs: trend following, momentum rotation and mean reversion.
How long until my first backtest?
About 10 minutes of sign-ups, then the sample runs in a few minutes. Your own idea takes as long as it takes you to answer the rule questions.
What if I get stuck?
Your AI checks the 41 known bugs first. For anything new, the two-strike rule stops it patching and makes it find the real cause. The playbook also has a prompt that gets a second opinion from a fresh chat.
Is your idea an edge or a guess? Find out tonight.
Not ready for $39? Take the free starter pack.
Prompt 1 with a 6-question worksheet, the 10-point check on one card, and 3 real bug fixes. A 4-page PDF.
Send me the free pack Enter your email at checkout and the PDF arrives right away. You'll get 3 short emails about backtesting, one about the full playbook. Unsubscribe anytime.