Four questions, twelve companies, once a month.

If you pick companies yourself, you'll pick badly. Not because you're stupid — because everyone does. You'll fall in love with one, you'll refuse to sell a loser because selling makes the loss feel real, and you'll buy something on a Tuesday because someone convincing said it was going up. So instead, the rules are written down once, in advance, and the choosing is handed to a computer program that has no feelings about it.

Once a month — the first Saturday — the program runs. It looks at every company listed in London and New York and asks four questions in order.

One: is this an ordinary company I can actually buy?

It skips anything unusual — funds, property trusts, banks — not because they're bad but because their accounts don't compare fairly against normal businesses. It also skips anything very small, anything barely traded, and anything too new to have a track record. Then a practical check, done by hand: can it actually be bought in the account? If not, skip it, take the next one.

Two: is the business genuinely making money?

Real cash coming in. Profits bigger than last year's. Not drowning in debt. This is the "is it a real business" test.

Three: has the share price been going up?

Specifically: higher than it has averaged over the last ten months or so, and up over the past year. This sounds obvious but it does a specific job — it stops you buying a cheap company that's cheap because it's quietly dying. Cheap and falling is a trap. Cheap and recovering is the thing you want.

Four: is it cheap?

Of everything still standing, keep only the cheapest third — measured roughly as what you'd pay for the whole business versus the profit it produces.

Then it scores them

Each survivor gets marks out of 100 for being cheap and marks out of 100 for going up, added together. The top twelve win. The money is split into twelve equal piles, one per company.

The next month it runs again. A company already owned is kept if it's still scoring well and still rising; otherwise it's sold. If fewer than twelve companies pass, the leftover money just sits in cash — that's on purpose. In a bad market almost nothing passes the price test, so the account naturally ends up mostly in cash. That's the system protecting itself, not breaking.

The part that matters most

The list the computer prints on Saturday is bought on Monday exactly as printed. Every single line. Nothing skipped because it looks scary. Nothing added because it looks exciting. And in between the monthly runs, the prices aren't looked at at all — no checking, no worrying, no "just a quick look".

That last rule is genuinely the whole project. The rules above are ordinary — plenty of people use similar ones. What's unusual is committing in advance to follow them without exception, including in the month when following them feels obviously wrong. The first year isn't really trying to make money. It's trying to prove the discipline holds.

And the safety net: it's £2,000, ring-fenced. If the whole account went to zero tomorrow, nothing in real life changes — no bill goes unpaid. That's the condition that makes it possible to leave the rules alone when they're losing.