The backtest beats the index by 40 points. Move the rotation one day later and it falls below the index. Add one realistic crash gap and it nearly halves. This is what fragility looks like, measured.
A headline backtest number means nothing until you know how it behaves when reality is slightly less kind than the simulation. Real execution is late. Real fills cost more than the close. Real crashes gap through your exit before you can act. So we re-ran the leveraged-index strategy under each of these stresses, one at a time, and measured what survived. The answer is sobering: the edge is real but thin, and small, plausible frictions erase it.
With idealized execution, the strategy returned +124.8% over five years, a Sharpe of 0.66, and a 52% max drawdown. The S&P 500 buy-and-hold returned +84.9% with a Sharpe of 0.77. The strategy's only clear win is the 40-point return premium — and a Sharpe that is already below the index.
The base case assumes you act on the same day the index crosses its 200-day line. In practice a daily-bar system often executes on the next session's open. So we delayed every rotation by a single trading day.
A single day of execution lag dropped the return from +124.8% to +74.2% — below the S&P 500's +84.9% — while the drawdown got worse (56%). That is the entire edge, erased by one day of slippage in timing. The cross days cluster around volatile turning points, so being late on those exact days is expensive.
Rotations are rare (33 in five years), so trading cost is not the main risk here — but we measured it anyway. Charging an extra 25 basis points per rotation trimmed the return modestly and left the Sharpe essentially unchanged.
This is the one piece of good news: because the strategy trades so seldom, transaction costs are a minor drag. The danger is not friction per trade — it is timing and tail risk.
The most important stress is the one no timing rule can dodge. We injected a single realistic overnight crash gap — the kind of move where the index falls hard before the 200-day signal can react and you cannot exit at yesterday's price.
A single bad gap cut the five-year return nearly in half, from +124.8% to +47.3%, and pushed the Sharpe down to 0.41. Because the position is 3x leveraged, an overnight move that would dent an index fund mauls this strategy. The 200-day filter cannot help: it reads closes, and the damage is already done by the time the close prints.
| Scenario | Total return | Sharpe | Max drawdown |
|---|---|---|---|
| Base case (idealized) | +124.8% | 0.66 | -52.2% |
| Rotate 1 day late | +74.2% | 0.51 | -56.3% |
| +25bp slippage | +110.9% | 0.66 | -54.1% |
| One crash gap | +47.3% | 0.41 | -54.3% |
| S&P 500 (buy & hold) | +84.9% | 0.77 | -25.4% |
Put the rows side by side and the conclusion is unavoidable. The strategy's return advantage over a plain index fund is fragile: a one-day lag or a single crash gap is enough to wipe it out, and in every scenario the Sharpe stays below the index's 0.77. The leverage that produces the upside is the same leverage that makes the strategy hypersensitive to timing and tails.
It is tempting to imagine an intelligent layer that 'predicts' the crash and exits early. It cannot. Crashes move faster than information; an overnight gap is, by definition, a price you never had the chance to trade at. The honest role for AI here is to describe the current risk, not to promise protection from the tail. Anyone claiming a model that prevents leveraged-crash losses is, at best, mistaken.
This is exactly why Mercurio runs on paper capital and frames the strategy as a specific, high-risk bet rather than a sure thing. The base-case numbers are real, but they sit on a knife edge. Read the full methodology in the backtesting deep dive and the rule itself in the strategy.
Disclaimer. Every figure here comes from historical simulation on paper capital, including the stress scenarios. Nothing here is financial advice. Leveraged ETFs carry substantial risk of large, rapid loss. Past performance does not guarantee future results.
