What we found
Every idea tested here, the verdict, and how sure we are. If you read one page on this site, this is a reasonable one.
Two of the four verdicts below look similar and are not. No means we measured it and it went the wrong way. We can’t tell means the test could not have seen the effect even if it were there, so the honest answer is that we do not know. Treating the second as if it were the first is the most common way to misread evidence, and it is a mistake this project has made and corrected.
| Idea | Verdict | Why |
|---|---|---|
| Paying less for the funds you hold | Settled | It comes out of a fund contract, not a forecast. The largest single line available to most people. |
| Holding each fund where it is taxed least | Settled, in direction | Statute, not opinion. The size depends entirely on which accounts you have, and it is zero if you have one. |
| Holding a fund rather than an old-style mutual fund | Settled, in direction | In 2025, 7% of exchange-traded funds paid out a capital gain against 52% of mutual funds. |
| Not trading | Settled, in direction | The average dollar earned 8.7% a year over ten years while the funds it sat in earned 9.9%. That gap describes a population of investors rather than what any one of them can bank. |
| Leaning toward cheaper, smaller, more profitable companies | Might | It measured 0.79% a year ahead of a cheap global index fund, and it would take about thirty years of data to tell that from zero. |
| Adding a managed-futures holding | We can’t tell | Its behaviour in a crisis is real. Its average return does not resolve anywhere, and a version of it would need 244 years to prove. |
| A 200-day moving-average timing rule | We can’t tell | On US data the effect is smaller than the test could see. On a century of data across sixteen countries it is real. In a taxable account it is undone by tax, and since it became famous it has cost 6.5 points a year. |
| Holding gold | We can’t tell | Its worth flips sign depending on whether you sell stocks to buy it or borrow to hold it on top, and neither answer is large enough to be sure of. |
| Bitcoin as protection against a crash | No | It falls harder than it rises, it was positive in 2 of the 12 worst months for stocks, and it made the worst case worse at every weight tested. |
| Crash insurance | No | One such fund lost 8.35% a year since 2017 while the market gained 14.52%. Of 102 funds sold as giving you the market with a floor under it, 86% did worse than simply holding less. |
| Direct indexing | No | It costs about 0.09% a year net at a 0.09% fee, and still slightly negative at a zero fee, so no cheaper provider fixes it. |
| Rebalancing as a way of making money | No | It is a way of stopping your mix drifting. Whether it earns or costs a little depends on the stretch you measure. |
What this adds up to
The things whose sign is known before the fact are all boring, all administrative, and worth more than every clever idea on this site combined. The clever ideas are not worthless. They are unproven, and the honest size of the best of them is smaller than most people expect.
Against a cheap index fund, everything measured here nets out to somewhere around zero, with a range wide enough to cross it. We cannot tell you this beats an index fund. Anyone who tells you their portfolio does is either measuring against something easier or has not checked how long their own claim would take to prove.