Koryu · 黄龍
Koryu
← Research library
Research · 研究 · 47 · Foundations

Survival first. Compounding is for the survivors.

21 Aug 20265 min readFoundationsKoryu Research

This article works through survival arithmetic. The numbers are illustrations of the mathematics, not recommendations of any risk level. Nothing here is investment advice. Decisions are yours.

Risk of ruin is the probability that an account falls far enough that the game effectively ends. It is the statistic that matters more than expected return, because ruin is absorbing: from zero, no strategy, no edge, and no recovery curve applies. Every sizing rule in serious trading is downstream of this one number.

Why ruin is different from loss

Ordinary losses are recoverable states; ruin is an absorbing one. The classical gambler's-ruin result says a bettor facing even a small edge disadvantage loses the entire stake with probability approaching certainty if they keep betting fixed amounts against a deep-pocketed opponent, and the market is the deepest-pocketed opponent there is. Trading versions soften the cliff but keep the shape: the account does not have to reach literal zero for the game to end, only a level from which the recovery arithmetic becomes absurd. Crypto's drawdown base rates reach that level often enough to matter: -90% needs +900%. The practical definition of ruin is the drawdown from which your strategy, capital schedule, and psychology cannot plausibly return.

The streak arithmetic

Fixed-fractional risk makes ruin arithmetic computable on a napkin. Risking a fraction f per trade, a streak of n consecutive losses costs 1 minus (1-f) to the nth power. Ten straight losses: at 1% risk, -9.6%, an annoyance. At 2%, -18.3%. At 5%, -40.1%, deep in the punitive zone of the recovery curve. At 10%, -65.1%, functionally ruinous for most operators. The second half of the napkin is realizing that ten-loss streaks are not tail events. A strategy winning 40% of the time, respectable by crypto breakout base rates, has a 0.6% chance of opening any given trade with ten straight losses, which sounds safe until you run a few hundred trades: across 500, you should expect such a streak to occur, roughly a handful of times. Streaks are scheduled, not unlucky. The only free variable is whether the account's risk fraction was set so the scheduled streak is survivable.

Crypto's three accelerants

The same arithmetic runs hotter here for three structural reasons. Correlation: the sector-correlation evidence says multiple open positions fail together under crypto's one-factor stress behavior, so a five-position book at 2% each is not five independent 2% risks, it is one 10% risk wearing five names, and four concurrent positions in our transplant experiment drew down as a single leveraged BTC bet. Wicks and cascades: stop-loss arithmetic assumes stops fill near the stop, and the candles a liquidation cascade prints routinely fill them far beyond, so realized f exceeds planned f exactly on the worst days. Leverage: margin converts the drawdown-then-recover path into drawdown-then-liquidation, replacing a survivable state with an absorbing one at the exchange's chosen level. Each accelerant alone tightens the safe range of f; stacked, they are why risk fractions imported from equity practice are quietly lethal in this market.

Behavioral ruin arrives first

The mathematics understates the danger, because humans exit the game before the account does. The documented sequence: a deep drawdown triggers the get-back-to-even reflex, size doubles, a scheduled streak lands on the doubled size, and a recoverable -40% becomes a terminal -80%. Or the mirror failure: the operator survives financially but capitulates psychologically, abandoning the system at maximum drawdown, locking the loss precisely where the arithmetic says persistence matters most, which is why we sell discipline rather than predictions. A ruin threshold is therefore honest only if set where the operator breaks, not where the account hits zero, and for most people those levels are separated by a factor of two. Pre-registering the quit condition, in writing, before the drawdown, is the cheapest ruin insurance that exists, the same logic this site applies to itself in the derivation journal.

The synthesis

Put the pieces in one frame. Expected return decides how fast wealth grows IF the game continues; risk of ruin decides whether it continues; and only the second is fully under the operator's control, through the sizing arithmetic: the risk fraction, the heat cap read as one correlated bet, and the refusal of leverage that converts drawdowns into terminations. The order of operations is absolute: survival first, then compounding, because the reverse ordering is not a strategy, it is a countdown with interim gains. Any service, feed, or system that quotes you returns without quoting the risk fraction and streak behavior underneath has answered the second question with silence, and silence there is an answer, the kind seven checks on a signal provider are built to surface. Decisions are yours.

Related reading
FoundationsPerpetuals vs spot: what you actually own4 min readFoundationsToken unlocks and emissions: supply schedules that move prices4 min readFoundationsThe Kelly criterion in crypto: why fractional is the only way5 min read
Frequently asked

What is risk of ruin?

The probability an account falls far enough that the game effectively ends. Ruin is absorbing: from zero, no edge or recovery curve applies. Practically it is the drawdown from which your strategy, capital schedule, and psychology cannot plausibly return, not literal zero.

How much does a losing streak cost at different risk levels?

Risking a fraction f per trade, n straight losses cost 1-(1-f)^n. Ten straight: 1% risk loses 9.6%, 2% loses 18.3%, 5% loses 40.1%, 10% loses 65.1%. The fraction, set in advance, decides whether a scheduled streak is an annoyance or a termination.

How likely are long losing streaks?

Near-certain over real sample sizes. A 40%-win-rate system has a 0.6% chance that any given trade opens a ten-loss streak, which compounds to an expectation of several such streaks across a few hundred trades. Streaks are scheduled, not unlucky.

Why is risk of ruin worse in crypto?

Three accelerants: one-factor correlation makes concurrent positions fail together (five 2% risks behave as one 10% risk), cascade wicks fill stops far beyond their levels so realized loss exceeds planned loss on the worst days, and leverage replaces recoverable drawdowns with liquidation.

What is behavioral ruin?

The operator breaking before the account: doubling size mid-drawdown to get back to even, or abandoning the system at maximum drawdown and locking the loss. Honest ruin thresholds are set where the human quits, which for most people is around half the mathematical level.