Koryu · 黄龍
Koryu
← Research library
Research · 研究 · 28 · Craft

Three out of four fail. The mechanics explain which.

27 Jul 20265 min readMomentum craftKoryu Research

This article explains market mechanics and quotes base rates from our own published experiment. It does not recommend trading breakouts or anything else. Nothing here is investment advice. Decisions are yours.

A breakout is the market's most seductive pattern: price pushes through a level everyone was watching, and the chart seems to announce that something has begun. Sometimes it has. Our own data says the honest base rate is brutal, roughly three out of four crypto breakouts we tested failed, and understanding the mechanics of why is worth more than any pattern library.

What a breakout physically is

Strip the mythology and a breakout is an order-book event. A resistance level, say a prior high, accumulates resting sell orders: profit-takers from below, breakeven-sellers who bought the last visit, and market makers leaning on the level. Price breaking through means aggressive buyers consumed that entire shelf of supply, which is genuine information about demand. The pattern's premise is that whoever ate the shelf did not do so to sell it back one percent higher, so the path above, now empty of resting supply, invites continuation. That premise is sometimes true, and it is truest when the eating was done by patient size rather than by a crowd of small triggers all firing at once, a distinction the chart alone cannot show you.

Who is on the other side

Every breakout entry buys from someone choosing to sell at the exciting moment, and the seller census explains most failures. Above the level sit the stop-losses of short sellers, whose forced buying fuels the first thrust, briefly. Into that thrust sell three groups with better information than the breakout buyer: holders who accumulated below and planned this exact exit, market makers fading a move their flow data says is retail-driven, and, in crypto specifically, whoever engineered the level's approach in the first place. The 24/7, retail-heavy, leverage-saturated structure of this market means the third group is not paranoia: pushing price through a watched level to harvest the trigger-buyers is a standard game wherever books are thin, and thin books are crypto's default outside the majors, thinner still once wash-traded volume is discounted.

The base rates, from our own ledger

In the full transplant experiment, we ran a disciplined 20-day-high breakout system across the top 120 pairs for five years: the win rate was 25.5%, and over half of all exits were stop-outs. Read that correctly: it does not say breakouts cannot be traded, trend systems have compounded for decades at similar hit rates when winners dwarf losers. It says the popular image of breakouts, most of them working, is backwards, and any approach whose economics require a high hit rate dies here. The gated variant of the same system, allowed to act only in favorable regimes, kept a nearly identical win rate but let its rare winners run far enough to overcome the failures. The pattern did not improve; the context did, which is the entire lesson of the dial's published formulas.

The failure taxonomy

Crypto breakouts die in four recognizable ways. The instant fade: the thrust exhausts on short-covering alone, sellers overwhelm, and price is back inside the range within hours, the signature of a crowded trigger with no patient size behind it. The wick kill: the breakout holds, but a routine 10-15% intraday excursion tags equity-calibrated stops before the trend resumes, a mechanism so dominant here that stop placement gets an article of its own. The slow bleed: price hovers above the level for days, momentum never arrives, and the position dies of time and funding rather than violence, which is why time-based invalidation exists. And the regime rug: a perfectly formed breakout in a deteriorating market, where BTC's turn drags every open breakout down at once, correlation doing what correlation does, the one-factor problem this library keeps running into. Note what the taxonomy implies: three of the four deaths are about context and RISK geometry, not about the pattern's shape, which is where pattern-education obsessively focuses.

Confirmation without a closing bell

Equity breakout craft leans on the daily close as its judge: a close above the level, settled in the auction, filters the intraday head-fakes. Crypto has no auction and no close, so confirmation must be a chosen convention rather than a market event. The honest options each trade something: a UTC-close confirmation, our board's convention in a 365-day market, filters wicks at the cost of entering a day late; a time-above-level rule, hold N hours through the level, filters instant fades at the cost of some whipsaw; volume confirmation is the traditional third leg, and in crypto it is the weakest, because volume is the one input an adversary can manufacture on demand. Whatever the choice, the principle survives translation: a breakout is a hypothesis, confirmation is evidence, and the entry's job is to price the hypothesis cheaply enough that being wrong three times in four still leaves the expectancy arithmetic standing.

The honest summary

Breakouts in crypto are real order-book events with a real but minority success rate, traded profitably only by approaches that expect failure as the normal case: small defined risk per attempt, winners given room to become outliers, and context filters that refuse the pattern in hostile regimes. The pattern's popularity rests on its best examples, which are unrepresentative by construction, and on educators who sell shape recognition because shape is photogenic and risk geometry is not. Our research program's wager, the one the derivation journal keeps score of, is that the tradable part of the pattern lives almost entirely in the context and the geometry. The shape is just where the crowd gathers. Decisions are yours.

Related reading
FoundationsPosition sizing in crypto: surviving 8% daily noise5 min readMethodologyWash trading and fake volume: crypto's dirtiest number5 min readFoundationsCrypto market sectors: L1s, L2s, DeFi, memes, rotation4 min read
Frequently asked

What percentage of crypto breakouts fail?

In our published five-year test of a disciplined 20-day-high system across the top 120 pairs, the win rate was 25.5%: roughly three of four breakouts failed. Trend systems can still compound at such hit rates when winners dwarf losers, but approaches requiring high accuracy die here.

What actually happens in the order book during a breakout?

Aggressive buyers consume the shelf of resting supply at a watched level, briefly fueled by short-sellers' stops above it. Continuation depends on whether patient size did the eating or a crowd of small triggers, and the chart alone cannot show which.

Who sells into a breakout?

Holders who accumulated below and planned that exact exit, market makers fading retail-driven flow, and, in thin pairs, whoever engineered the approach to harvest trigger-buyers. Every breakout entry buys from someone choosing that exciting moment to sell.

How do you confirm a breakout without a market close?

By convention: a UTC-close confirmation filters wicks at the cost of late entry; time-above-level rules filter instant fades; volume confirmation is weakest in crypto because volume is the one input an adversary can manufacture on demand.

What makes a breakout tradable despite the failure rate?

Context and geometry rather than shape: small defined risk per attempt, winners allowed to become outliers, and regime filters that refuse the pattern in hostile tape. In our test the gated variant kept the same win rate and overcame failures purely through context.