Some ideas help when added to stocks and hurt when they replace them. These cards separate what the portfolios hold, what was measured and set aside, and what still needs a proper test.
Each card carries one of six words, defined on the about page: Confident, Probably, Maybe, Unlikely, Avoid, or Not tested. Seven ideas earned a place in a portfolio and are explained on the portfolio pages. The practical choices about costs, accounts and trading are on how to hold it. The idea behind the whole site, that stacking several small edges should beat the market, gets its own page.
Two habits run through every verdict. First, how you pay for an idea matters as much as the idea. Trend following bought by selling stocks is a coin flip; added on top of the stocks, it gave about 1.7% a year more on the same data. Gold bought by selling stocks cost 0.4% a year. Bought out of the bond line, it cost nothing. Second, every result is rerun with the extra return cut in half, because ideas earn less once everyone knows them. Nothing gets “probably” unless it survives that.
Buying companies that are cheap for their profits beat a plain index by a little on both histories, and stayed ahead with the extra return cut in half. Needs decades of patience.
12.1% a year the full lean against 10.3% for a plain world stock index, 1990 to 2025, simulated
Momentum and value take turns. Moving five points of US stocks into a momentum fund added a little on a century of data and in the fund's own three years, and it is the one fund swap the long history could tell from zero. The lean portfolios now hold ten points; the second five added about as much again.
+0.16% a year from moving 5% of US stocks into a momentum fund, 1927 to 2026, simulated; range +0.10% to +0.22%
A strategy that follows markets that keep moving one way, held alongside stocks. The long simulated history is encouraging. Real funds can do something else.
13.3% a year Plus trend against 11.3% for all US stocks, 1932 to 2025, simulated
Five percent of gold in place of bonds cost nothing the tests can see and made the worst falls a little shallower. Its real return over a century is about zero; its job is to move when nothing else does.
1.75% a year over cash 1975 to mid-2026, after storage costs
These bonds raise what they owe you as prices rise. Holding fewer stocks cuts stock-market risk, though bond prices still fall when interest rates climb faster than inflation.
−18.1% worst fall of the cautious portfolio against −52.7% for a plain world stock index, 1990 to 2025
Five points in the cautious portfolio, paid for from the other bonds, not from stocks. They rise in a deflationary crash and fell 40% in the 1970s and again in 2022; the two bond lines cover different disasters.
−59.1% worst fall of long US government bonds against cash, 1926 to 2025
Forty of every hundred stock points, against the world market's thirty-seven. Foreign stocks are far cheaper than US ones, and most of the US lead since 1990 was rising prices rather than rising profits.
81% of the US lead over 35 years came from US stocks getting dearer, 1990 to 2024
A real idea, and a history too short or too noisy to tell it from zero.
Simulations that count what paid for it find plausible growth, bigger swings and an advantage that is not certain. They do not show that the timing is what created the gain. Still exploratory.
Moving half the stocks to cash whenever the market sits below its ten-month average cut the century's worst fall from 84% to 65%, for a return gap too noisy to read. The tax-free case avoids the modeled tax cost of repeated sales.
−65% worst fall of the full lean with half its stocks on the rule against −84% without, 1927 to 2026, simulated
A basket of oil, metals and grain futures earned about 3% a year over cash since 1877 and protects against inflation, but in a recession it falls with stocks. Not held; gold does the covering job more cheaply here.
36% of the worst stock months since 1926 in which commodities rose
NTSX in place of the plain stock funds in Plus trend added about half a percent a year on the 1990 history and a tenth on the century, at the same worst fall. It pays only while bonds earn more than the borrowing, which in 2026 they barely do, and it fell 26% in 2022.
+0.5% a year Plus trend with NTSX in place of VTI, VXUS and SCHP, 1990 to 2025, simulated; +0.1% on the century
Holding more stock when the last three years were calm and less when they were wild, up to 1.5 times, is a well-known idea. Here it moved the return by an amount too noisy to read and made the worst fall deeper on both histories. Capped at no borrowing, it simply lost a percent a year.
−3 points change in the full lean's worst fall on the century with exposure scaled by recent calm, cap 1.5, simulated in a tax-free account
A fund that stacks trend following on Treasuries instead of stocks. Ten points of it on the cautious portfolio's bond line added about 0.6% a year on 96 years, and nothing since 2009.
The one holding whose losses come from hurricanes rather than markets. The funds earned about 1% a year over cash, and the price is the thinnest on record.
2.2 times the expected loss, mid-2026; the question reopens at 3.5 times
Moving 10% of stocks into cash-like government bills added as much back in the worst months as gold did, for no fee.
+0.92% added back in the average worst stock month since 1926
Carry tries to earn from the gap between financing costs and futures prices. It can spread the risk in trend, though some carry trades lose when stocks do. The fund, and what pays for it, matter.
TAIL helped in the 2020 crash but lagged a funded cash alternative over the full test. CAOS stayed close to cash in its shorter record.
About a third the cost of RSST per dollar of trend, but it pays for the trend by selling stocks, and on the histories here that funding gives back most of the gain.
On 1926 to 2014 the idea was the one bond-like bet that did not move with government bonds. The funds are twelve years old and the extra yield is the thinnest since 1990.
Keeps your mix from drifting, which is the point. As a source of return it came out a hair behind never trading.
−0.09% a year a yearly review against never trading, inside a range the test could not read
Measured, and the evidence leans against it.
Funds that buy the most profitable companies regardless of price, such as QUAL and DUHP, matched or trailed the market over their ten years. Profitability pays here only alongside a low price, which the value funds already look for.
14.2% a year QUAL over ten years to 2026, against 14.9% for an S&P 500 fund
Funds such as VUG and QQQ buy fast-growing companies whatever the price, which is the other side of the value lean held here. Over the long record expensive stocks earned less than cheap ones. The good part of growth, that rising shares keep rising for a while, is what the momentum funds hold.
−0.3% a year the full lean with a growth fund in place of its momentum fund, 1990 to 2025, simulated; −0.4% on the century, behind in four of five ten-year stretches
Small companies as a group have not beaten large ones once you strip out the cheap, profitable ones among them. The small-cap funds here are held for cheapness and profits, not size.
The only holding tested that made the portfolio's worst fall deeper at every weight. At most 2% as a bet, and zero is fine.
24.8% to 29.2% a portfolio's worst fall before and after a 10% bitcoin holding, 2015 to 2026
Interest-rate risk plus the risk that borrowers fail to pay, which arrives when stocks fall. In 2026 the extra yield over Treasuries is near its lowest since 1990, so the reward for that risk is thin.
0.8% extra yield of investment-grade corporate bonds over Treasuries, August 2026
The stock market with a screen and a tax bill. Six of them trailed the whole market at matched risk.
0.84 out of 1 how closely a dividend fund moves with the whole market
Property funds gave 112% of the market's falls for 80% of its rises. Stocks by another name.
0.82 out of 1 how closely a property fund moves with the whole market
Selling stocks when the market looks expensive lost money on 150 years of data even before tax. A high valuation is a reason to hold more foreign and cheap stocks for good, not a signal to trade.
The cost is certain and the benefit is not.
Halves the worst fall on a century of US stocks and earned less than holding, with most exits losing money and a tax bill of about 1.9% a year in a taxable account.
−43.1% worst fall under the rule against −83.7% for holding, 1926 to 2026, simulated
A 3x fund held since 1927 lost almost everything at its worst and spent 28 years getting back. The 200-day rule still left a loss of about 67% in the 1987 episode.
−99.9% worst fall of a 3x US stock fund, 1927 to 2026, simulated
A cap on your gains in exchange for a cushion on your losses. Priced on a century of twelve-month windows, the package costs you.
−2.4% to −4.1% a year what the cap-and-cushion package was worth on 1,183 twelve-month windows since 1926
Owning an index's stocks one by one to harvest tax losses cost money on every one of 400 simulated thirty-year paths.
0.2% of the losses it harvests ever save tax for someone who never sells
Named and priced, but no test has been run.
They sell the market's upside each month for cash, the mirror image of buying crash insurance. No whole-portfolio test has been run.
A fund that shifts its stock and bond mix as a target date approaches. Compare what it actually holds, what it charges and what it assumes about your withdrawals.
Nothing here separates a sector from the market it sits in. About half the US market's lead over the world is its larger tech sector.
A themed fund is a bet on one story. The nearest thing tested, an AI stock-picking fund, returned about half the S&P 500 since 2017.
The one form tested, owning an index's stocks one by one to harvest losses, lost money on every one of 400 paths.
Long bonds, gold and commodities held in fixed shares with borrowed money. The pieces are measured here one at a time; the funds themselves are not, and 2022 was their bad year.
Mostly a lean toward smaller companies, in a dearer form than the small-cap value funds held here. Not tested as a holding.
The same inflation-protected bonds in a different form. A ladder suits a dated spend; I-bonds are capped at $10,000 a year.
No fund isolates them, and the trend fund held already trades fifty-plus markets worldwide.
Ten-year returns of 2.9% to 4.3% a year in a decade whose cash rate was most of that. Too small to be worth resolving.
Close to a US bond fund with a currency hedge attached. Emerging-market bonds carry the currencies that fall hardest in a crisis.
Staking pays about 1.3% to 1.9% a year before a fee that can exceed it, on an asset that fell 67% from its 2025 high.