Five ways of selecting funds that did not work
Everything on this site is built by rules, so every rule can be tested — and most of the ones we tried failed. These are five that looked sensible, cost real effort, and were beaten by the simple version. Publishing them is cheaper than letting someone else repeat them.
1. Combining three screeners concentrates instead of diversifying
The idea was tidy: take the three funds with the strongest recovery signals, the three with the highest Opportunity Score, and the three with the highest Quality Score. Three independent lenses, nine funds, one portfolio.
| Strategy | Net a year | Max drawdown | Rotations |
|---|---|---|---|
| Three Screeners | −6.4% | −46.5% | 34 |
| Income Conservative | +11.2% | −10.8% | 7 |
It was the worst result we have recorded, and the reason is that the three lenses are not independent. The Opportunity Score rewards what has fallen furthest; the momentum screener rewards what is bouncing after a fall; and in this universe the Quality Score tends to reward the same technology names. Instead of three different bets, the portfolio bought the same bet three times — and paid for it three times when the bet went wrong.
2. Income per unit of risk buys the wrong funds
Ranking funds by yield divided by volatility is a classic move: it should find the income that comes cheapest in risk. On our universe it did the opposite, selecting funds whose volatility was low because their price had already collapsed and settled.
| Selection criterion, 5 dividend funds | Gross | Net | Max drawdown |
|---|---|---|---|
| By quality | +11.6% | +8.3% | −26.5% |
| By dividend growth | +10.1% | +6.7% | −27.0% |
| By income per unit of risk | +6.5% | +2.7% | −38.4% |
A dead fund is a calm fund. Any ratio with volatility in the denominator will eventually find the instruments that have stopped moving because there is nothing left to move.
3. Picking the best ten within one issuer beats holding all of them — until you count the tax
Holding all 56 IncomeShares products seemed naive: surely selecting the ten with the best scores would do better. Gross, it did. Net of Italian taxes, it did not.
The reason is the offsetting of realised losses. A wide basket always contains something that has fallen, and those losses shelter the distributions of everything else. Select the ten best and you have selected away your own tax shelter — and added rotations, each with its commission and its realised gain.
4. You cannot pick your way out of an asset class
The last and most instructive failure. We tried to build a strategy from the classic dividend ETFs by selecting harder: by dividend growth, by quality, by coverage, by yield. Then by holding more of them. Then by forcing one fund per region.
| Variant, classic dividend funds only | Net a year | Max drawdown |
|---|---|---|
| 5 funds, by quality | +8.7% | −26.4% |
| 10 funds | +7.8% | −26.4% |
| 14 funds | +7.6% | −27.3% |
| 10 funds, one per region | +6.7% | −30.9% |
| 10 funds + 20% bonds | +6.5% | −23.2% |
| 12 funds, one per region + 30% bonds | +4.5% | −24.0% |
| Euro Income — mixed categories | +13.2% | −6.2% |
Every variant landed between +4.5% and +8.7% with a drawdown around 26%. The selection method made almost no difference, the number of funds made none, and forcing regional spread made things worse — because it obliges the portfolio to buy regions that returned less, without reducing the risk: dividend funds fell together everywhere in 2020 and 2022.
The strategies that do work — Euro Income, Income Conservative — get their advantage from mixing categories: currency, bonds, and a slice of option income, which behave differently when things go wrong. Not from selecting dividend funds better.
5. Defensive rules that defend against the wrong thing
Several attempts — exiting on a drawdown threshold, requiring a minimum yield to enter, rotating faster when volatility rises — shared one outcome: they cut the recovery rather than the fall. The rule fires after the drop has happened, sells at the bottom, and is not back in when the rebound comes.
What did reduce drawdowns was structural, not tactical: holding fewer option-income funds, adding bonds, and staying in one currency. Boring changes to the composition, decided in advance, instead of clever reactions to the market.
Why we publish the failures
A backtest that only shows what worked tells you nothing about the method, because the same effort applied to any dataset produces a handful of good-looking results by chance. What tells you something is how many attempts it took, and which ones failed.
Ours: four of the twelve published strategies came from ideas that worked first time, and the rest of the ideas are on this page. The five above are the ones worth knowing about; there were others that failed for less interesting reasons.
Method and limits
- Same engine as the published strategiesEvery variant was run through the same simulation, with the same commissions, the same tax profile and the same rules on entries and exits. Nothing was tuned afterwards.
- Short periodsMost tests start in 2023, some in 2016. A failure over three years is not a law of nature; it is what happened in those three years.
- Our universe148 European UCITS funds that exist today. A different universe could give different answers, particularly for the dividend results.