The summer drawdown belongs to a book I no longer trade

−29.4%. That is what the NDX Rotate + X certificate on wikifolio lost between its close on 22 June and its close on 29 July 2026. On eToro, over the same weeks, the number was −28.2%. Both figures are in the permanent record of those accounts, and they will stay there (a maximum drawdown doesn’t fade with time, it just gets older).

This post explains three things: which version of the strategy produced that fall, what has changed on all three accounts since, and why I think the drawdowns you should expect from here are meaningfully smaller. The last one is a claim about a backtest, so I’ll show the working.

The book that fell

Until September, both the original wikifolio and the eToro account ran the same book: half the portfolio in five Nasdaq 100 momentum stocks, the other half in gold and Bitcoin. Concentrated and, above all, ungated: the stock half stayed invested whenever the monthly rotation found five names, no matter how violently the market underneath was moving.

That book has a backtest going back to 1995, and its worst fall in that backtest is −35.9%. So −29.4% was not a malfunction. It was the ungated book doing, at full volume, exactly what its own history said it could do. Run through the same five weeks, the old book’s backtest loses −28.0%, with the same peak and trough dates and within a point and a half of the live accounts (the gap is certificate fees and execution). Nothing was “broken”, but that doesn’t make the ride any more pleasant.

What changed

There are now three live books on three accounts, and all of them run a rule the old book didn’t have:

	NDX Rotate + X	eToro	Meridian Systematic
Account	wikifolio (the original)	eToro	wikifolio (the new one)
Current book since	10 Sep 2026	1 Sep 2026	1 Sep 2026
Stocks	5 (Nasdaq 100)	13 (Nasdaq 100 + Russell 1000)	21 (Nasdaq 100 + S&P 100 + Russell 1000)
Stocks / gold + Bitcoin	50% / 50%	55% / 45%	70% / 30%

The eToro book is the one you can copy, and it has 13 names for a very practical reason: eToro doesn’t open a copied position worth less than $1, so a copier with, say, $100 would silently miss the smallest positions of a wider book. Every position in the eToro book is sized to stay above that line. Meridian Systematic is a second, separate wikifolio that launched on 3 August, after the fall, and runs the widest version of the engine. Its goal is different from the original’s: lower overall risk. In the backtest it gets there on every measure that counts for that goal (worst fall since 1995 −21.1% instead of −23.0%, worst calendar year −3.5% instead of −9.2%, and growth per unit of worst fall, the MAR, 1.13 instead of 1.06), and it pays for it with somewhat less growth: 23.8% instead of 24.4% a year since 1995.

The rule they share is a volatility gate. Each stock part of the portfolio only holds stocks while its index is in a long-term uptrend and its recent volatility sits below a multiple of its own one-year average. When the market gets jumpy, that stock part goes to cash at the next month-end and waits. Gold and Bitcoin are never switched off.

How often does that happen? Far more than you’d guess. Over 31 years, the Nasdaq part of the book sits in cash 34% of the time, roughly one month in three, and the longest stretch was 33 months, from November 2000 to July 2003. This is not a crash detector that fires once a decade. The rule is closer to a thermostat: it’s boring, it’s frequent, and the protection comes from exactly that boredom.

The same five weeks, with the gate

Here is the comparison that matters, because it is the one you actually lived through:

22 June → 29 July 2026	Result
Live wikifolio, old ungated book	−29.4%
Live eToro, old ungated book	−28.2%
Old ungated book, backtest	−28.0%
NDX Rotate + X today, backtest	−9.7%
eToro book today, backtest	−7.1%
Meridian Systematic today, backtest	−9.1%

The worst fall any of today’s books would have taken in the whole summer was 11 to 12%, and it came in early June, not in July.

Now the honest part. Adopting a rule right after the loss it would have prevented is exactly what curve-fitting looks like, and a sceptical reader should be suspicious (I would be too). So the question I had to answer before switching the original wikifolio on 10 September was not “would the gate have saved July?” but “what does it cost when nothing is burning?”. The answer, measured over 2016 to 2026 (a window with no dot-com bust and no financial crisis in it, so the one that flatters a defensive rule the least), was 0.03 percentage points of growth a year. In exchange, the worst fall in that window came down by 8.5 points and the longest wait below a previous high shrank by 6.6 months. A sweep across different gate settings also showed the chosen one sitting on a broad plateau rather than on a lucky spike, which is what you want to see before trusting a parameter with real money.

The new numbers

Here are the backtests of the books as they are configured today, with costs, through 14 September 2026. The last ten years come first; the full story since 1995 comes right after, because one without the other is misleading.

Since 2016:

	NDX Rotate + X	eToro	Meridian
Growth per year	38.2%	34.1%	34.4%
Sharpe ratio	1.64	1.54	1.54
Positive years (2016–2025)	8 of 10	8 of 10	8 of 10

Since 1995 (31 full years):

	NDX Rotate + X	eToro	Meridian
Growth per year	24.4%	22.0%	23.8%
Worst fall	−23.0%	−21.1%	−21.1%
Worst calendar year	−9.2%	−5.0%	−3.5%

Why is the recent decade so much stronger? Three reasons, and none of them is “the strategy got smarter”: the window contains no real bear market, it ends on a partial 2026 that is already up more than 60%, and Bitcoin, which drove a good part of the growth, has no data before it. That’s why the 31-year line sits right next to it. If you only ever remember one number from this post, make it the 22.0% (eToro, since 1995), not the 34.1%.

One small piece of housekeeping, since you might notice the figures moved. On 27 September I corrected how the once-a-year rebalance between stocks, gold and Bitcoin is valued in all three books: it now uses the opening price it actually trades at instead of the previous evening’s close. That moved every long-term figure by a few tenths of a percentage point (the eToro book from 22.3% to 22.0% a year, for example). In live trading it only nudges the size of that one yearly transfer; the monthly rotation and the gate are untouched.

What this does not mean

It doesn’t mean the next fall will be 9% instead of 29%. The backtested worst for today’s eToro book is −21.1%, the public target is a maximum drawdown of roughly 25%, and I will not promise anything below what has already happened. Markets are perfectly capable of finding a path that the last 31 years didn’t show. What changed is the shape of the risk: the stock part now steps aside when volatility spikes, instead of riding every spike to the bottom.

The first four weeks under the new rules have been good (September so far: +9.5% on eToro, +8.9% on the original wikifolio). Four weeks, however, are noise, in both directions, and I’d say the same if they had been bad.

What I keep coming back to is the equity curve. The only way to tell a normal drawdown from a broken strategy is to know the history well enough to see where you are on it. The −29% of this summer is on the old curve, and the old curve said it could happen. The new curve says the same five weeks would have cost about a third of that. Both are just maps, not promises, so read them before the next storm, not during it!

Past performance and backtests are not a reliable indicator of future results. The figures for today’s books are simulations (historical back-calculation with costs), not a live record; nobody traded these exact books before September 2026. Copy trading involves risk, and the value of your investment can go down as well as up.

---