The screen you'd build by hand — run without ever reading a filing.

The value engine is Greenblatt-adjacent, not P/E-driven

Ranking is on EV/EBIT, which is the right denominator if you care about capital structure — you're paying for the whole enterprise, debt included, and comparing against unlevered operating profit. Cheapest 30% within country, ranked separately for the UK and the US, so you're not arbitraging Sterling against Dollar multiples.

The quality overlay is deliberately blunt, but it screens out the usual traps

Positive trailing free cash flow cuts the accrual-driven "profitable on paper" names. Diluted EPS positive and growing year-on-year — the growth condition is what separates this from a pure deep-value screen; you're not buying melting ice cubes at 4× EBIT. Net debt to EBITDA under 3× is the leverage discipline, waived only if the company is net cash.

The universe exclusions are structural, not tactical

All financials and REITs are out permanently — not a view on the sector, an admission that EV/EBIT and net-debt/EBITDA are meaningless for a balance-sheet business. Same reason it excludes trusts, ADRs, and ambiguous dual-class structures. That's a known, accepted blind spot rather than a bug.

Where it will feel wrong to you: the momentum gates

Price above the 200-day average, positive 12-1 return, positive six-month relative strength. This is the part that stops it being a value screen. Its function is not alpha generation — it's a value-trap filter. A name has to be cheap and already re-rating. In practice that means the system never buys the thing at its actual low. It buys the second-cheapest version of it, three months after the bottom, once the tape has confirmed. If your edge is being early and sitting through the drawdown, this screen is designed to do the opposite of what you do.

The 200-day average is also the entire sell discipline

There's no thesis review, no "the story's broken" call. Three exits exist: composite rank falls out of the top 25, price closes below its 200-day average at screen time, or a corporate action forces it. Checked once a month, executed the following Monday, nothing in between.

The part that will actually bother you

There is no override, in either direction. If a holding puts out an accounting restatement on the Tuesday after rebalance, it is held until the next screen date. If a name you know is a fraud passes all four gates, it is bought. The system has no mechanism for judgement, because the whole premise is that the human's judgement is the failure mode being controlled for — this is a discipline record first and a performance record second.

And the sizing is the risk control: twelve equal-weight positions, ~8.3% each, no stops. At this account size a stop protects less than it costs in spread and whipsaw, so the position cap does the work — a total single-name zero costs about 8% of the book. Concentration risk is capped structurally rather than managed actively.