KoinlyticsA 10% position with a 50% stop does not risk 10% of the portfolio. It risks 5%. That gap, between the size of a position and the size of the risk taken on it, is where most portfolios bleed out, one reasonable-looking trade at a time.
Position size is not risk size
Position size is how much capital goes into a trade. Risk size is how much capital is actually lost if the trade goes wrong, measured from entry to stop. They are only the same number if the stop sits at zero, which almost never happens. Put $1,000 of a $10,000 portfolio into a token (a 10% position) with a stop 50% below entry, and the maximum loss on that trade is $500, or 5% of the portfolio. Put the same $1,000 into a position with a 20% stop and the risk drops to $200, or 2%. Two trades of identical dollar size can carry risk that differs by 2.5x depending on where the stop sits relative to entry.
This matters because most sizing mistakes come from thinking in position terms, deciding to put in 10%, instead of risk terms, deciding to lose no more than 2% of the portfolio on an idea that is wrong. Risk-based sizing makes position size a function of stop distance: a tighter stop allows a larger position for the same dollar risk, a wider stop forces a smaller one.
Fixed fractional sizing
Fixed fractional sizing sets a constant percentage of the portfolio to risk per trade, commonly between 0.5% and 2% for anyone trading with any frequency, and derives position size from it: position size = (portfolio value x risk fraction) / stop distance as a percentage of entry price. A $50,000 portfolio risking 1% per trade ($500) on a position with a 25% stop can hold a position of $2,000. The same portfolio, on a position with a 5% stop, can hold $10,000, because the tighter stop means less capital is exposed per dollar of position size.
The dollar amount at risk should shrink after losses, not grow. A fixed percentage already does this automatically, since it is a percentage of a smaller balance each time. Raising the fraction after a loss to recover it faster is one of the more reliable ways to turn a small loss into an account-ending one.
Why oversized risk guarantees ruin, even with a positive edge
Assume a trader has a genuine edge: a 55% win rate on 1:1 payoff trades, for an expected value of 0.55 x 1 minus 0.45 x 1, or plus 0.10 per unit risked. Positive expectancy, the kind that should compound into a growing account over time. What actually happens to the account depends heavily on what fraction of capital is risked per trade, because compounding is geometric, not additive.
The geometric growth rate per trade for a fraction f risked is approximately: win rate x ln(1 + f) + loss rate x ln(1 - f). At f = 10%, that is 0.55 x ln(1.10) + 0.45 x ln(0.90) = 0.55 x 0.0953 + 0.45 x (-0.1054), or about +0.005 per trade. Small, but positive and compounding. At f = 50%, the identical 55/45 edge produces 0.55 x ln(1.50) + 0.45 x ln(0.50) = 0.55 x 0.405 + 0.45 x (-0.693), or about -0.089 per trade. Negative. Same edge, same win rate, same payoff, and the account is now mathematically headed toward zero over enough trades, purely because too large a fraction was risked each time.
This is the logic behind the Kelly criterion, a published framework for the theoretically optimal risk fraction given a known edge and payoff. It is presented here to show how sizing interacts with edge, not as a sizing recommendation: real edges and win rates are estimated, not known with certainty, and full Kelly sizing is aggressive enough that most people who use the framework at all use a fraction of what it calculates.
The mechanism worth internalizing: a large loss does not just subtract from the account, it shrinks the base the next gain has to compound from. A 50% loss needs a 100% gain to recover, a relationship covered in full in the chapter on drawdowns. Risking a large fraction per trade means an occasional large loss is not a remote possibility, it is close to a certainty over enough trades, and each one pushes the compounding base further back.
Stops do not cover the failures that actually blow up crypto accounts
A stop-loss order protects against a token's price falling while there is still a liquid, functioning market to sell into. It does nothing for the failure modes that have caused the largest single losses in crypto, because none of them leave a market to exit through:
- A chain halt: the network stops producing blocks, or a bridge is paused, and there is no transaction to send, stop order or otherwise.
- A depeg: a stablecoin or wrapped asset breaks its reference value faster than an order can fill, often trading through several price levels within a single block, as happened when UST unwound in the Terra collapse.
- An exchange freeze: withdrawals are suspended while the balance shown on screen keeps updating, so the position exists on a screen but not in a wallet, the pattern seen in the FTX collapse.
- A contract exploit: a smart contract is drained, and the tokens it held or issued go to zero, or close to it, in the time it takes to read the news.
These are not tail risks in the statistical sense, priced into volatility and expected to mean-revert. They are binary events: the capital is either recoverable or it is not, and a stop order sitting on an exchange's order book is irrelevant to all four. This differs from a leveraged position being force-closed by an exchange, covered in the chapter on liquidations, where a market still exists and the mechanism is at least visible in advance. The only real mitigation for chain, bridge, exchange and contract failure is sizing: capital that is never concentrated in a single point of failure is capital that a halt or an exploit cannot take all at once.
Illiquid tokens need smaller size, because the exit is part of the risk
A stop order assumes an exit exists at approximately the stop price. In a thin market it does not. Selling a position that is large relative to order book depth or pool depth moves the price against the seller as the order fills, and an illiquid token can gap through several price levels on a single sell with no buyer in between. The effective loss on the trade is the price move plus the slippage, and slippage scales with position size relative to available liquidity, not with the dollar amount in isolation.
This means the same 2% portfolio risk rule produces a smaller dollar position in an illiquid token than in a liquid one, even at an identical stop distance, because the realistic exit price is worse. Checking depth before sizing a position, not after deciding to sell, is the only way to know whether the stop distance assumed on paper is the stop distance available in practice.
Working with sizing in practice
- Size every position from the dollar risk to the stop, not from a round percentage of the portfolio.
- Pick a fixed risk fraction per trade and hold it constant regardless of a recent run of wins or losses.
- Treat chain, bridge, exchange and contract risk as reasons to cap total exposure to any single point of failure, since no stop order reaches them.
- Check order book or pool depth before sizing into a low-liquidity token; the stop distance only holds if the exit does.
- Koinlytics shows current exposure by chain and by asset across connected wallets, the input needed to check concentration before adding to a position.