AlphaStrat 1.0

AlphaStrat 1.0 is a set of rules for buying stocks. The rules are boring on purpose. Every quarter we rank the US large-cap universe on two questions: How good is this business? and Are we paying a fair price? We buy the thirty names that score best on both, equal-weight, and we hold them until the next quarter. We do this in a computer because computers don’t fall in love with their stocks and don’t panic when the market does.

Two homemade tools answer those two questions. AlphaQuality scores a business across six things any sensible owner cares about—whether the company makes money, what it earns on the capital invested, whether growth is real or borrowed, what the cash flow looks like, what the balance sheet looks like, and whether any of it is stable. AlphaVal compares the price to what we think the business is actually worth, using the right math for the kind of business (discounted cash flow for compounders, book value times return on equity for financials, FFO multiples for REITs, the long-run commodity price deck for oil and gas). Lower combined rank wins.

Here is the scorecard. Over 27.4 years from January 1999 to May 2026—through the dot-com crash, the great financial crisis, the COVID crash, and three roaring bull markets—the rules would have compounded at 14.57% a year. The S&P 500 did 8.17%. The Nasdaq-100 did 10.67%. The worst drawdown was -34.8%, which sounds bad until you see that SPY lost -55.2%, QQQ lost -83.0%, and RSP lost -59.9% in the same window.

What we’re willing to own

Plain US common stocks worth at least $1 billion at the time we would buy them, with at least one earnings report on file in the past three years. No foreign listings with a currency mismatch, no ETFs, no companies too small to trade without moving the price. That cuts the universe to roughly 1,500–2,000 names depending on the year.

No peeking at the future

At every quarterly rebalance from 1999 forward, the ranker sees only the financial statements that had actually been filed by that date. No restated numbers, no benefit of hindsight. Market caps use the share count and price from that day, not today. If a company hadn’t filed its 10-K yet, we treat it as not filed. This is the part of a backtest that most people quietly cheat on, and it’s usually where the magic comes from.

Scoring the business (AlphaQuality)

AlphaQuality grades each company from 0 to 100 on six things any owner ought to care about: does it make money, what does it earn on each dollar invested, is the growth real, is the cash flow real, can it pay its bills, and are any of those things stable enough to bet on. We grade these differently depending on the kind of business—a bank is not Apple, a refiner is not a software company. Companies that lose money for years in a row or carry an impaired balance sheet get a failing grade regardless of what the formulas say.

Estimating what it’s worth (AlphaVal)

AlphaVal uses the right valuation math for each kind of business: discounted cash flow for compounders, book value times return on equity for financials, FFO multiples for REITs, the long-run commodity price deck for oil and gas. Margin of safety is the difference between our estimate and the current price, expressed as a percentage. Higher means cheaper. We won’t buy anything with less than a 10% cushion.

Putting them together

Rank every eligible name by quality (best to worst) and by margin of safety (cheapest to most expensive). Add the two ranks. The thirty lowest sums get bought, equal-weighted. This is the Greenblatt approach, with two of our own scoring systems doing the ranking. To be eligible at all a company needs a quality score of at least 70 and a margin of safety of at least 10%.

How we hold and trade them

Each new position is sized at 1/30 of the portfolio. Once we own it, we leave it alone—winners are allowed to grow into bigger positions, losers shrink. We rebalance on the first trading day of each calendar quarter. To keep us from being too twitchy, a name we already own that drifts to rank 31-40 stays in the portfolio rather than being swapped for a marginal improvement. No more than five names from any one sector, which caps any industry at about 17% of the book. Trading costs are modeled at 10 basis points one-way, which is generous on stocks this size. Cash we’re not currently invested in earns the Treasury bill rate.

A million dollars handed to these rules in January 1999 would have grown to $40.97M by May 2026. The same million in SPY would be $8.43M, and in QQQ $15.41M. Over this 27.4-year span, AlphaStrat 1.0 achieved outstanding relative performance metrics across all passive benchmarks.

Drawdown is the number that decides whether you can actually live with a strategy. A 50% loss feels nothing like a 30% loss when you’re in it—the first one keeps people up at night and gets them fired; the second one is uncomfortable but survivable. Across the full 27 years, these rules drew down less than every passive benchmark we measured. The bulk of that gap was earned in two episodes: the dot-com crash, when expensive tech stocks unwound and we weren’t in them, and the financial crisis, when high-quality businesses held up better than the market.

Three of the four major crises were milder for the strategy than for the S&P. The COVID crash was the exception—a one-month liquidity panic where everything fell at roughly the same speed, and a quality-and-value tilt was useless because nobody was thinking about quality or value. We made it all back by the end of that year.

The most reassuring thing we can show you is that the rules win in the years when they should win and lose in the years when they should lose. They beat the index by the widest margin during the dot-com bust, when a bunch of expensive stocks with no earnings imploded and we were never in them. They trailed the index for most of the 2010s, when a small number of mega-cap tech companies (Apple, Microsoft, Google, Amazon, eventually Nvidia) were responsible for almost all of the S&P’s gains and our valuation discipline correctly noted that the prices were stretched—correctly, but unprofitably for us. This is the same pattern that the academic value literature has been documenting for forty years, and it is not going away.

An aggregate 14% a year hides a lot of texture. In the worst year the strategy lost about 20% of the portfolio; in the best year it made roughly 50%. The good years against the S&P clustered when value was in fashion—the early 2000s, 2022, 2023. The bad years against the S&P clustered when a tiny number of expensive growth stocks were doing all the heavy lifting—2017, 2020, 2021, and most recently 2025. If you owned the strategy through those years, you would have spent a lot of dinners explaining to people why you didn’t own whatever stock was on the cover of the magazine.

Every backtest is a small story told in a way that flatters its author. Here is everything we know we’re leaving out, and the rough size of each lie. Read this section twice.

1. The dead are missing

The universe is built from the US large-caps that exist today. Enron is not in it. Neither is Lehman Brothers, WorldCom, Bear Stearns, Washington Mutual, Countrywide, Sun Micro, or Pets.com. A correctly-built historical universe that included these tombstones would, in our estimate, knock 1 to 3 percentage points a year off the CAGR we show. We’re working on fixing this in a future version.

2. Yesterday is not tomorrow

These numbers are simulated. Nobody made any of this money. The next 27 years will look nothing like the last 27 years, and the particular pattern of value rotations and tech booms that this strategy navigated may not recur in the same way. The honest answer to "will this work going forward?" is "the principles are sound; the magnitude is unknowable." Anyone telling you otherwise is selling something.

3. Sector risk is reduced, not eliminated

No sector ever gets more than five of the thirty slots, which caps it at about 17% of the book. That stops us from being all in on banks the way an uncapped value strategy can be. But a credit shock that hits banks, REITs, and leveraged infrastructure all at once can still produce a correlated drawdown across positions that look diversified.

4. The trading costs are optimistic

10 basis points of one-way slippage. No commissions, no market impact. That is generous for $1B+ stocks and fine for an individual investor at this size. It would be wrong if a billion dollars tried to run this strategy.

5. Cash is treated as a free option

Uninvested cash is assumed to earn the day-over-day total return on short-term Treasuries. That overstates what you’d get on real cash sitting in a brokerage account, and understates the borrow cost if you ever ran the strategy on margin. The size of this lie is small but it’s nonzero.

6. The IRS gets paid first

This is the most important paragraph in the report and you should not skip it. Every number we’ve shown you so far is before tax. Rebalancing four times a year turns the portfolio over at roughly 75 to 90% a year, which means almost every gain we realize is held under twelve months. In a US taxable account that gets taxed as ordinary income. Your federal rate is whatever your last dollar of wages is taxed at, plus the 3.8% net investment income tax for high earners, plus your state. The S&P 500 in a buy-and-hold account has almost no realized gains until you sell. We have a lot of them, every quarter, forever. That is a structural tax disadvantage, and no amount of clever ranking fixes it.

Here is what the 14.57% gross CAGR turns into after the tax man takes his share, at different brackets, assuming 78% turnover:

The strategy belongs in a tax-advantaged account—an IRA, a 401(k), an endowment, a foundation. In a taxable account at the top California bracket the edge over SPY narrows to a sliver. Your real result will depend on your bracket, your state, whether you can harvest losses against the realized gains, and whether any positions happen to survive long enough to qualify for long-term treatment. Run the numbers with your accountant before you do anything.