HFEA stands for HEDGEFUNDIE's Excellent Adventure.
Read from the start: chapter 1 The RuleLeveraged ETFs can lose a large part of their value in a short time.
Backtest on historical data: shows what the rule would have done, not what it will do. Not investment advice.
The backtest puts the strategy against holding without rebalancing. It computes on the closing prices of the two blocks, since both exist, and therefore shows only the recent years: at most as many as there are real prices.
What is computed is the classic version with the two US blocks, in their currency. A combination of European blocks is not computed by this page; it can be set up as its own version in the Tracker.
One curve is the strategy with rebalancing, the other the same two blocks without it; both start at the same value.
The value axis is on a log scale so that equal percentage steps look equally large, and the markers sit on the dates.
If holding is ahead in this period, one side ran away and the rule kept selling it; why rebalancing then falls behind is in the chapter Derivation.
A few years are a single market environment. The curves show what the rule would have done in exactly these years; what it does next and which setting is the better one in the long run cannot be read off them.
Missing from both curves are the trading costs of the round trips, that is bid-ask spread and order fee, and the tax on the gains that rebalancing realises; they would hit the strategy above all, because holding does not rebalance after the purchase. The ongoing product costs are inside the price.
What applies to you is in the chapter Implementation and Taxes further down.
The same rule can be viewed over longer periods on reconstructed series:
Here the same rule with quarterly rebalancing runs on reconstructed series of US equities and long-dated US Treasuries, with assumed costs; these are not fund prices. One curve shows the strategy, the other the same blocks without rebalancing; this includes the long period of rising interest rates up to 1981.
How the rule would have fared across long stretches and many starting points has been recomputed by others, with reconstructed series for the time before the products existed. The following analyses name their data and their assumptions.
Community backtests
Someone else's backtest is usable when it states where its data comes from, how it computes and which costs and taxes are included. The numbers of the following analyses are there to be read, not here.
HEDGEFUNDIE himself (2019) puts a backtest from 1955 into the original thread, that is across an environment of rising rates. The data series and the procedure are in the thread, and the author names the uncomfortable result himself: the bond leg does badly in the decades of rising rates from the 1950s to the early 1980s. 1986/87 is only the start of the first widely circulated backtest and not the edge of that weak phase. Taxes do not appear in it, because the strategy runs inside a tax-advantaged account. The tax on rebalancing costs the community's backtests roughly one to two percentage points of return per year. In those backtests the larger hurdle in Europe is availability: no threefold leveraged funds and no US ETFs.
modern_football (2021) tests not one backtest but all of them: the result is reported for every possible starting point rather than for the one that looks best. The data series and the procedure are in the post, together with its limits: the analysis starts in 1986, computes 50/50 rather than 55/45 and with ongoing monthly contributions, so it does not cover the decades of rising rates to the early 1980s at all. The end date is another limit: the analysis predates 2022 and does not know the year in which both blocks lost together. Like the US analyses in general it computes without tax.
ZahlGraf, parts 5 and 9 (2022) recompute the strategy for Europe, with the products available here and on data series from 1943. Part 5 derives the European variants, part 9 sets the combination with trend following beside it, and part 10 handles tax separately. The limit lies with the constructed series and with the product differences, which the series sets out rather than smoothing over. Part 5, part 9, part 10.
ChemicalStats, Vectis Mundi (2025) tests leveraged holdings across broad indices and includes German tax and the advance lump sum, which is what makes the difference for a German account. The two-block strategy is not the subject there; the leveraged block is. The limit lies in the dependence on index choice, which the second part sets out. Part I, part II.
What follows from that. Weights, rhythm and band width were selected against history; the version that is called HFEA today is already an adjusted version. The long history behind that comes from the community backtests further up and sits largely in a period of falling interest rates, which is a tailwind for the bond block.