The method behind the thirteen
How it works
Edge Thirteen runs on a nearly century-old idea, executed at a speed no human could match. Here's the whole thing in plain English: the philosophy it's built on, and exactly what the AI does with it every week.
The whole game, in one question
Every week the analysis is trying to answer one thing about a stock: is the price you'd pay today lower than what the business is actually worth? That gap — price below worth — is the entire strategy. It's called value investing, and it rests on a simple premise: a share of stock isn't a lottery ticket, it's a small piece of a real business, worth the earnings and assets behind it. Pay less than that, and time is on your side.
The foundation
Benjamin Graham — the rules of price
Graham is the father of value investing and the man who taught Warren Buffett. His great insight was psychological as much as financial. Picture the market as “Mr. Market”— a moody business partner who shows up every day and shouts a price at you. Some days he's euphoric and overpays; some days he's despairing and sells cheap. You're never obligated to trade. You just wait for the day he offers you a bargain, and ignore him the rest of the time.
To know a bargain when you see one, Graham insisted on a margin of safety — only buying when the price sits well below your estimate of value, so an error or a bad break still leaves you protected. And he made it concrete with quantitative tests. These are the gates the AI applies to every company:
Is it actually profitable?
At least eight of the last ten years in the black, and a return on equity above 12%. One good year proves nothing. Ten years of them is a business.
Can it pay its bills?
A healthy current ratio and a balance sheet that isn't buried in debt, so a bad quarter doesn't turn into a solvency problem.
What does it earn in a normal year?
Not last year — the average across the whole cycle. A company at a record high looks cheap on a P/E right up until earnings come back to earth, which is exactly when it stops being cheap.
Is there a margin of safety?
The price has to sit meaningfully below what the business looks worth. That gap is what protects you when the estimate turns out to be wrong, and sometimes it will be.
The evolution
Warren Buffett — cheap isn't enough
Buffett started as Graham's student and then upgraded the method. Graham would buy almost anything if it was statistically cheap. Buffett noticed the flaw: some things are cheap because they deserve to be. A shrinking business at a low price is a value trap— the earnings keep falling and the “bargain” evaporates. So he added a second test on top of price: the business itself has to be good.
That means a durable competitive advantage — a moat like a strong brand, pricing power, or high switching costs — and a high, consistent return on equity that lets the company compound over years. His famous line captures the shift: “It's far better to buy a wonderful company at a fair price than a fair company at a wonderful price.” The only way to judge whether a company is wonderful is to read what it actually files — its economics, its risks, its management — which is exactly the step most people skip.
The updates
Three places the textbook needed fixing
Graham wrote his rules for the companies in front of him — railroads, steel, textiles. Apply them unchanged to a market with software and semiconductors in it and they misfire in ways that are easy to miss, because the output still looks like a tidy list of cheap stocks.
A record year is not a discount
This is the one that costs people money. An investment bank comes off a blowout year, earnings triple, and suddenly it trades at four times earnings. The screen lights up. But you're not buying those earnings — you're buying whatever it makes next year, which is usually a lot closer to its average. So every company here is valued on what it earns across the cycle, not on its best year, and anything currently earning more than about 1.6 times its own normal gets held back from the value-pick tier no matter how good the multiple looks.
One yardstick for every industry doesn't work
Graham wanted a P/E under 15 and a price-to-book under 1.5 — one bar, whole market. The trouble is that a bank is mostly a balance sheet, so it sits near book value in good times and bad, and sails through that test on structure rather than cheapness. Run the screen that way long enough and you own a portfolio of banks and insurers without ever deciding to. Each company now gets measured against what's typical for its own industry instead, which is the only comparison that tells you anything.
Some assets never make it onto the balance sheet
Accounting rules say research and development is an expense, gone the moment it's spent. For a software or chip company that's the entire business — years of engineering that still earns money every day, recorded as worth nothing. Judge those firms on book value and they look permanently overpriced, so a pure Graham screen simply never buys one. The fix is to put that spending back on the books and let it wear off over a few years, the way the thing it built actually does.
None of this loosens the standard. It's the same question Graham asked — is the price below the worth — measured in a way that survives contact with a market he never saw.
The engine
How the AI puts it to work
No human reads 5,000 annual reports a week. A machine can. It runs Graham's discipline across the whole market for breadth, takes a view on where the year is heading, then applies Buffett's judgment to the survivors for depth — roughly what a good analyst would do, minus the sixty-hour week and the temptation to fall in love with a name.
Screen the entire market
Every week the AI pulls fundamentals for roughly 5,000 U.S. public companies straight from their SEC filings — the same EDGAR data professionals use — and runs Graham's quantitative gates on all of them: profitability, earnings stability, balance-sheet strength, and valuation. No watchlist, no favorites; the whole market takes the same test.
Read the year ahead
Before pricing anything, the AI takes the measure of the market itself — what every sector earns relative to its own normal, what the Treasury curve is doing, where the qualifying names actually are. Out of that comes a written view on which sectors the next twelve months should help and which it should hurt, each with the mechanism spelled out and, just as important, what would prove it wrong. That view nudges the bar. It never removes a name: a genuinely cheap, genuinely good business in an unfashionable sector still shows up, just further down.
Price each name against its own industry
Graham judged every company on the same yardstick — a P/E under 15, a price-to-book under 1.5. That was fine for 1949 and it quietly breaks today. Banks hold enormous balance sheets and trade near book by nature, so they clear that test whether or not they're cheap. Software firms expense their R&D the moment they spend it, so the thing actually earning the money never lands on the balance sheet at all, and they fail no matter how good the price gets. So each company is priced against what's normal for its own industry, with R&D put back on the books where it belongs.
Read the 10-Ks
A screen can tell you something is cheap; it can't tell a bargain from a trap. So the AI reads the actual 10-K and the latest quarterly report for every name that clears the screen — not just a favored few — the risk factors, the results, the direction of travel — applying Buffett's question: is this a solid business temporarily on sale, or a declining one that's cheap for a reason?
Make the call — three tiers
Every name lands in one of three tiers. A value pick passes both tests — the Graham math AND the Buffett filing read — a genuinely strong business that's temporarily cheap; the gold. Cheap for a reason passes the math but fails the read — statistically cheap, but the filings show a business in trouble (a value trap). Too expensive is priced well above what the business is worth, good company or not. Each call comes with a plain-English reason straight from the filings.
Value picks first
The list leads with every value pick the week turns up — the whole point, and genuinely rare, because markets are at least somewhat efficient. When there aren't thirteen of them, it fills with the most instructive 'cheap for a reason' traps to avoid, then a name or two of the most egregiously overhyped. It lands in your inbox Friday at 1 a.m. Mountain Time.
Where the honesty lives
This method tilts the odds; it doesn't remove risk. Cheap can get cheaper, a solid business can stumble, and any estimate of value can be wrong — that's the entire reason for the margin of safety. So the scorecard shows every value pick it has made and how it's done since — winners and laggards alike, priced automatically, never cherry-picked.
Edge Thirteen is a general-circulation educational publication offering impersonal analysis of publicly traded companies — not individualized investment advice. I am not a registered investment adviser, and nothing here is a recommendation to you to buy or sell any security. Investing involves risk of loss, and past performance does not guarantee future results.
See it every Friday
Thirteen stocks priced against their fundamentals, grounded in the filings, for $13/month. Cancel anytime.
Secure checkout via Stripe · cancel anytime · terms & refund policy.