About this site

How the portfolios were tested

The main question is whether a change improves a whole portfolio once you count what paid for it. Selling stocks, selling bonds and borrowing inside a fund are three different choices. We compare each against a named alternative, and look at the growth, the losses, and the long stretches spent behind that alternative.

The headline charts use simulated histories from 1990 to 2025 and 1932 to 2025. They apply market and style returns to the displayed fund weights, with modeled fees and trading costs. They are not the funds' actual records. In the long history, US data stand in for foreign stocks, a nominal government bond stands in for inflation-protected bonds, and a simple trend rule stands in for a modern trend-following fund. The long window starts in 1932 after the trend rule gathers its initial sizing history; it excludes the 1929–1931 decline. These substitutions limit what the charts can establish.

Each experiment writes down what it is testing, what it is testing against and what it assumes, before it runs. Many reuse data already looked at, so writing the plan down first does not make the answer independent. Recent tests cut the assumed return from a strategy and keep its past swings. The earlier chart study's half-return test cut the risk too, which made it a weaker stress than the way it was first described.

The published portfolios are working choices for different needs. Other combinations may improve them. A stronger claim needs evidence about the actual funds, realistic costs and risks, and results that do not depend on one favorable sample. A gain in a simulation is not a chance of future success.

What the confidence words mean

The pages use these six words to say how sure we are about one particular use of a holding. They are not numbers, and they are not probabilities.

• Confident The stated mechanism or fact is well established. This does not promise an investment return.
• Probably The evidence favors a benefit for this use. How big it is, and whether it lasts, are uncertain.
• Maybe A plausible benefit, but the evidence, or the fund that would deliver it, is too uncertain to favor it yet.
• Unlikely The evidence leans against this use. Different costs, conditions or evidence could change that.
• Avoid The costs or risks make this a poor choice for the use described on the page.
• Not tested This site has not measured the proposed use. Evidence may exist elsewhere.

“Confident” is never used for a measured return. A return is at best “Probably”.

The idea behind the site

Several good strategies can work better together when they earn their returns for different reasons and take their worst losses at different times. How often each one beats an index cannot by itself tell you how the combination behaves. The size of the gains and the losses, the risks they share, the costs and what pays for them all matter. The stacking page explains the comparison.

Why Kelly

The Kelly criterion chooses a bet size to maximize long-run compounded growth when the return distribution is known. Investing is harder, because you have to estimate those returns and risks yourself. The site uses compounded growth as one measure alongside losses, liquidity, taxes and the ability to stay invested. The name does not mean these portfolios are known to be mathematically optimal.

Corrections

The site went live on 6 September 2026. When a number turns out to be wrong it will be corrected everywhere it appears, and an entry will land here saying what a reader would have done differently, newest first.

Disclaimer

None of this is advice, and none of it takes account of your situation. The site will never answer a question about your own money. The full disclaimer says the rest.