Ask anyone what a leveraged ETF position needs and you get the same answer: a stop-loss. It sounds obviously right. TQQQ can fall 20% in a week, so cap the damage and re-enter later. I believed it too. Then I tested it, because my own money runs on this strategy and “sounds obviously right” is not a risk model.

The setup

The base is a rules-based 3x Nasdaq-100 strategy I trade with real money and document publicly. The backtest runs on synthetic 3x data from 1999 to 2026, built from the underlying index, so it includes the dot-com crash, 2008, the 2020 crash and the 2022 bear market, not just the friendly decade TQQQ has actually existed for. On top of that base we tested 54 stop-loss configurations: trailing stops and stop-limits, tight to wide, evaluated separately across market regimes.

The question was simple. Does any stop-loss configuration improve the strategy’s risk-adjusted outcome across regimes, rather than in one lucky window?

What happened

Three patterns, and they killed every candidate:

Tight stops died by whipsaw. Anything tight enough to matter in a real crash triggered constantly in normal 3x volatility. A 3x fund moving 4% on an ordinary Tuesday hits a tight stop, sells, and buys back higher after the rebound. Each round trip is small. A hundred of them are not. Death by a thousand paper cuts, compounded.

Wide stops did not protect. Set the stop wide enough to survive normal volatility and it only triggers after the damage you were trying to avoid has already happened. You lock in the loss near the bottom, then face the re-entry problem with less capital.

The lone survivor was a fluke. Exactly one configuration looked like an improvement. We attacked it: shifted the regime windows, perturbed the parameters, re-ran it on adjacent configurations. It fell apart. A robust edge survives its neighbors. This one was a coincidence with good marketing.

Why stops and 3x funds do not mix

Leveraged ETFs reset daily. Their volatility is not just larger, it clusters: calm for months, then violent in both directions. A stop-loss is a bet that prices move smoothly enough to exit near your trigger, and that big down moves keep going. In a 3x instrument, both assumptions fail regularly. The biggest up days live right next to the biggest down days, and a stop guarantees you hold through the former and miss the latter.

I did not only learn this in a backtest. My live account ran a trailing stop in an earlier version of the system. It executed at an awkward moment for a technical reason no backtest would ever show. Real trading finds failure modes on its own.

What works instead (in the data)

The strategy that survived all of this does something less satisfying than a stop: it decides, on published rules, when the environment justifies holding a 3x position at all, and goes to cash when it does not. Over 1999–2026 that turns an unmanaged 3x position (27.3% CAGR with drawdowns beyond 80%) into 30.7% CAGR with a worst drawdown of 48.3%. Backtested numbers, labeled as such; the live account has been documented publicly since 2025.

A 48% drawdown is still brutal. I publish it because it is true, and because everyone promising leveraged returns with 15% drawdowns is selling you the thing this article just tested.

The honest summary

• 54 stop-loss variants, tested across regimes: zero robust survivors.

• Tight stops bleed you dry, wide stops arrive late, the one “winner” was noise.

• If your leveraged ETF plan is “buy and use a stop-loss”, the data says you have a plan for feeling safe, not for being safe.

Every trade of the real account, and the rules it follows, are public here: https://www.tradingautomation.de/live. The Weekly Ledger documents it every week.

The Honest Curve is a publication that documents my own real-money account and my research. It is general information addressed to all readers, not personalized investment advice, and no recommendation to buy or sell any security. Leveraged ETFs are volatile and can lose most of their value; my worst backtested drawdown exceeded 48%, and an unmanaged 3x position would have lost more than 80% at times. I hold positions in the instruments I write about and I trade them according to the published rules. Past performance, live or backtested, does not predict future results.

The leveraged ETF corner of the internet has a credibility problem. It runs on backtest screenshots. A backtest is a claim about the past made by someone who chose the start date, the parameters and which failed attempts you never got to see. I know exactly how flattering those choices can be, because building backtests is half of what I do.

So I chose the one format that cannot be retouched: a real account, real money, rules published in advance, every trade on a public page with a timestamp. If the strategy degrades, that page degrades with it, in public. That is the deal, and it is the entire reason this newsletter exists.

What “documented” means here

• The rules are published. A dual moving average regime filter on the Nasdaq-100 decides when the account holds TQQQ and when it holds cash. No discretion, no “I had a feeling”. A bot executes the rules.

• Every trade is public. Order, date, and the rule that triggered it, on the live track record page: https://www.tradingautomation.de/live

• Live and backtest are never mixed. Backtested numbers are labeled backtested. The live record starts in 2025 and stands on its own.

• Losing weeks get published like winning weeks. The Weekly Ledger ships every week, on schedule, whether the account is up, down or asleep in cash.

The part almost nobody publishes

Research that failed. Publishing it is uncomfortable and it is the strongest honesty signal I know, so here is a sample from my own archive:

• We tested 54 stop-loss variants on the strategy. None survived robustness testing.

• We tested rotating into long-term treasuries (TLT) during bear regimes instead of cash. Looked clever for two decades of the data. 2022 destroyed it: the “safe” leg turned a 44.5% drawdown into 61%. Idea rejected.

• We tested pushing leverage beyond 3x. Every step made returns and drawdowns worse at the same time. Rejected.

Three ideas that sounded great, tested honestly, thrown away in public. The strategy you can watch trade today is what survived that process, and the ledger of rejected ideas is part of the product.

What this is not

I do not give personal advice. Not in issues, not in comments, not in DMs. This is a publication that documents my own account; whether anything here fits your situation is a question I am neither allowed nor able to answer, and the people who answer it anyway over the internet should worry you.

There are also no paid promotions here. Nobody pays me to mention any security, and my only positions are the documented ones.

Why you might want to follow along

If you are interested in leveraged ETFs, you have two options for learning. Option one: screenshots from strangers. Option two: watching one fully documented system operate in real time, crashes included, with the research trail attached. This newsletter is option two, once a week, free.

The Honest Curve is a publication that documents my own real-money account and my research. It is general information addressed to all readers, not personalized investment advice, and no recommendation to buy or sell any security. Leveraged ETFs are volatile and can lose most of their value; my worst backtested drawdown exceeded 48%, and an unmanaged 3x position would have lost more than 80% at times. I hold positions in the instruments I write about and I trade them according to the published rules. Past performance, live or backtested, does not predict future results.