Skip to content

The prompt from the video

One strategy, 8 weekends

6 chapters · Video coming

The prompt

Get the weekend bot prompt

One prompt for Claude Code: it proves the rule on five years of daily prices, tries hard to break it, then installs the bot in practice mode, with a menu of settings to try.

Get the prompt

Your name and email, and the prompt appears straight away.

Your email goes on my list for notes on this test and the next videos. Unsubscribe in one click. Privacy policy.

weekend-prompt.md · the only thing you paste

The one-shot prompt

One file · 20 KB · Runs on your computer

The prompt

Fill in the form to unlock it.

A faceless clay figure typing at a laptop with a blue screen, a desk calendar beside itVideo coming

Become an AI trading wizard

“On Thursday, if a coin’s price closed above its average for the last 20 days, buy it, and sell it on Sunday.”

I had Claude run over 4 million backtests, and one rule survived: every Thursday, buy each of 19 coins whose price closed above its 20-day average, and sell on Sunday. On five years of daily prices, $1,000 became about $2,428, against about $650 for just holding the same coins. Then a bot traded it for 8 weekends: $500 became $567.64. One prompt proves the rule in Claude Code and installs the bot.

Your host

Lewis JacksonTests trading ideas with Claude. Founder of Part-Time Quant Academy.

Inside the test

What five years
of weekends showed.

4 charts

What $1,000 became, 23 Sep 2021 to 20 Sep 2026The rule, 19 potsThe rule, 19 pots: $2,428$2,428Whole balanceWhole balance: $720$720Just heldJust held: $650$650
01Five years of weekendsFrom 23 September 2021 to 20 September 2026, starting with $1,000: every Thursday, buy each of 19 coins whose close is above its 20-day average, one pot per coin, and sell on Sunday. It lost money in 2022 and 2025, and its worst drop was 33%, against 82% for just holding the coins. Putting the whole balance into the few coins that qualify did far worse.Backtest on Yahoo Finance daily closes, 0.12% charged on every buy and every sell. Backtest results are history, not a forecast.
27 coins it was never built on, $1,000Buy on ThursdayBuy on Thursday: about $3,000about $3,000Buy on any dayBuy on any day: $400 to $500$400 to $500
02Why ThursdayOn 27 coins the rule was never built on, buying on Thursday turned $1,000 into about $3,000. Entering on any day the rule fired made about $400 to $500. Same coins, same rule: only the day changed. Those coins tended to fall from Monday to Thursday and rise on Friday and Saturday.Thursday version made money on 22 of the 27 coins. Backtest on Yahoo Finance daily closes, 0.12% charged on every buy and every sell. Backtest results are history, not a forecast.
Top 100 coins from 2021, dead ones in, $1,000The ruleThe rule: about $2,600about $2,600Just heldJust held: about $430about $430
03With the coins that diedEvery coin above is still around today, so picking them quietly picks the winners. So I ran it on the top 100 coins from September 2021, dead ones included. The rule turned $1,000 into about $2,600, against about $430 for holding them.97 of the 100 have Yahoo prices. Backtest on Yahoo Finance daily closes, 0.12% charged on every buy and every sell. Backtest results are history, not a forecast.
$500 over 8 weekendsThe botThe bot: $567.64$567.64Same coins, heldSame coins, held: about $342about $342
048 weekends of the botOn 23 September 2026 I had Claude turn the rule into a bot and started it with $500. After 8 weekends it had $567.64. The same 19 coins, just held, fell to about $342, and Bitcoin on its own fell 23%. That’s my run, and 8 weekends is a short one.The bot’s own log, 24 September to 15 November 2026.

From the video

“Same coins, same rule. Only the day changed.”
Lewis Jackson
Share this video

Questions

Before you run it.

01What does the prompt do?

It downloads daily prices from Yahoo Finance, proves the rule on five years of them, tries it on coins it was never built on and on coins that died, then installs the bot on your computer in practice mode. It runs in Claude Code.

02Does it place any trades?

Not unless you choose to. The bot starts in practice mode: it writes down every buy and sell it would make and places none. Trading it for real is your call, with your own exchange key that can’t withdraw.

03Can I change the settings?

Yes. The prompt offers a menu: a 10 or 50-day average, a different buy day, selling on Monday, double the costs, or only the last two years. It runs each one next to the original and writes your results card. Paste it in the comments under the video, because I want to see what you find.

04Will I get the same numbers?

Close, not identical. Yahoo revises its prices, and your run may end on a later date than mine.

05Should I trade it?

That’s your call, and this isn’t advice. Know what it’s like first: it lost money in 2022 and 2025, and its worst drop was 33%. Run it in practice mode before anything else.

More videos · 5

LewisInvestingbyJackson

Become an AI trading wizard.

Legal

© 2026 Jackson. All rights reserved.Privacy · Terms · Contact