This article analyzes protective-stop design as an engineering problem, using published numbers from our own failed experiment. It does not recommend any specific stop, distance, or strategy. Nothing here is investment advice. Decisions are yours.
The single bloodiest finding in our founding experiment was not about entries, ranking, or regimes. It was the stops: over half of all exits in a five-year crypto test were stop-outs, many on moves that meant nothing about the trend, and that one mechanism converted a strategy family with decades of equity pedigree into a losing system. Stop placement in a 24/7 market is a different engineering problem, and this is its anatomy.
The wick problem, with numbers
A protective stop makes one promise: to sit outside the noise and inside the disaster. The design question is where noise ends, and crypto answers differently than equities. Our transplant experiment placed stops at 1.25 times the 14-day ATR below entry, clamped between 2% and 15%, geometry that clears the daily noise floor of most equities comfortably. On crypto alts, a typical 8% ATR put stops roughly 10% below entry, and anything wilder hit the 15% clamp. Then the market did what it routinely does: intraday excursions of 10-15% that resolve within hours, wicks, in the chart vernacular, tagged those levels constantly, and 293 of 517 exits came at the stop. The stop's promise inverted: it sat inside the noise, converting ordinary volatility into realized losses, and the 15% cap, an equity safety feature, guaranteed the placement stayed wrong by forbidding the distance the volatility demanded.
Why the wicks exist: cascades, not news
Equity intraday extremes usually trace to information. Crypto wicks are frequently mechanical: the market carries a permanent overhang of leveraged perpetual positions, each with a liquidation price, and when price reaches a cluster of them, forced market orders fire into whatever depth exists, pushing price into the next cluster, a chain reaction that ends only where liquidity finally absorbs it, often several percent beyond where any voluntary seller lived. Books are thinnest on weekends and off-hours, so cascades travel furthest exactly when attention is lowest. The design consequence is profound: crypto's extreme intraday moves are not distributed like equity moves plus a scale factor, they have a fat mechanical tail that mean-reverts once the forced flow exhausts, meaning a stop in cascade range does not just realize noise, it sells into the single worst prices of the day with mechanical reliability.
The design space, honestly surveyed
Distance-based stops scaled by ATR remain the baseline, and the transplant's lesson is not that ATR fails but that equity multiples fail: the multiple must be derived from crypto's own excursion distribution, asking directly, what distance survives ninety-something percent of adverse wicks that precede continuation, asset by asset, regime by regime. That distributional derivation is entry one on our research docket. Structure-based stops, below the level that defined the trade, respect the market's logic but sit exactly where cascade clusters and stop-hunts concentrate: the crowding problem, moved from entry to exit.
Time-based invalidation, exit if the thesis has not paid within N days, kills the slow bleeds that distance stops never catch and pairs naturally with wide disaster stops. And close-confirmed exits, acting on the daily close rather than the intraday touch, are the wick problem's most direct answer, filtering the mechanical tail at the cost of giving back more when a move is real, a trade-off, not a free lunch. Serious designs combine layers: a wide distributional disaster stop, a time invalidation, and a close-confirmed trend exit each guard a different failure, and the combination is testable, which is the point.
The gift that offsets the tax
One property of 24/7 markets genuinely favors stop-based risk management: no gaps. An equity stop can be jumped overnight, a position opening 30% through its level having never offered the exit, which is why our equity sibling's worst single losses came from gaps, not stops. Crypto's continuous tape means a resting stop executes near its price, cascade slippage permitting, so the stop you design is approximately the stop you get. The honest framing: crypto taxes stop placement, via wicks and cascades, and subsidizes stop execution, via gaplessness. A design that pays the tax, distances derived from real excursion distributions, and collects the subsidy, tight execution once genuinely wrong, is playing the board as it lies, which is all good risk engineering ever is.
What we are testing, and the reader's takeaway
The derivation program's stop-geometry study measures wick distributions across the point-in-time universe, majors and alts separately, and grades candidate geometries by one criterion: survival-adjusted expectancy through the validation years that walk-forward validation holds back, under the pre-registered gates. Results, including failures, publish to the journal. For the reader who trades manually, the portable lessons are three: measure your asset's actual daily excursions before choosing any distance, treat equity-derived rules of thumb as untested imports, and decide in advance which failure each layer of your protection is for, noise, bleed, or disaster, because a single stop asked to guard all three will guard none well. The market does not care where your stop is. The cascade, unfortunately, does. Decisions are yours.