Top Ideas

How this list is built

• Universe: the 500 largest US-listed companies in our coverage whose price has updated within the last 7 days. Valuations our sanity checks flag as unreliable (cross-currency ADRs, extreme outliers) are excluded.
• Inclusion criteria (all must hold): margin of safety between +15% and +60% (margin of safety = (fair value − price) ÷ price — we cap at +60% because gaps wider than that are usually data or model artifacts, not bargains), a quality factor z-score above +0.5 (where the quality factor is missing, the overall composite stands in), and an overall factor-composite z-score above 0.3.
• Momentum filter: names whose momentum factor z-score is at or below -0.3 are excluded — when the market is voting that hard against a statistically cheap name, it is more often a deteriorating business than a bargain. Names without a momentum score are not excluded by this rule. Each card shows its momentum z-score as the “M” chip.
• Conviction marker: Strong setup = margin of safety between +20% and +50%, quality z-score above +1 and positive momentum — the band where the gap, the business quality and the tape all agree. Everything else that qualifies is a Solid setup.
• Value-trap penalty: names down more than 30% over the last twelve months stay in the pool but have their rank score multiplied by 0.7, and their caveat leads with the deterioration warning. A deep discount after a deep fall is often deserved.
• Sector diversification: at most 2 ideas per sector in the top 10. We walk the pool in rank order and skip lower-ranked names from an already-full sector in favor of the next name down; names with no usable sector label share a single capped bucket.
• Ranking:rank score = 0.5 × margin-of-safety percentile + 0.5 × quality z-score percentile, both computed within today's qualifying pool (6 names), then the value-trap penalty above where it applies. Up to 10 (after the sector cap) are shown — if fewer qualify, fewer are shown; the list is never padded. Each card prints its percentiles — and the ×0.7 multiplier when applied — so you can check the arithmetic.
• Update cadence: factor scores and valuations refresh on the daily data sync; this page recomputes at most once an hour.
• Conservative by design: our fair values are deliberately conservative, so a name has to look clearly mispriced to make this page. The risk line on each card states the specific weak spots visible in our data for that name — nothing generic.
• Limitations: fair values are model estimates with substantial error bars, scores are relative to our coverage rather than the whole market, and a high rank is not a prediction of returns. Screening well is a reason to research a company — not a reason to own it.