Risk management basics

A gain has no fixed limit in principle, but the possible loss can be known in advance for some products and can exceed the amount put in for others. Risk is usually described with a maximum loss (the worst case), an expected loss under assumptions, and a tail (rare, large events). Estimates rest on models, and models can be wrong, so the maximum loss is the most solid number to know.

Fixed fractional: size equals capital times the risk fraction, divided by the loss per unit if the plan fails. Fixed amount: the same money at risk each time, regardless of capital. Volatility-scaled: size is inversely proportional to a measure of how much the price typically moves, so quieter things get larger positions and wilder things smaller ones. Kelly-style formulas relate size to an assumed edge and odds; they depend on inputs that cannot be known precisely and are very sensitive to errors, so practitioners often discuss using only a fraction of the result. These are descriptions of concepts, not recommendations.

A stop order becomes a market order when the trigger level is touched, so the fill can be well beyond the trigger when prices jump or liquidity is thin. A stop-limit order sets a worst acceptable price, but then may not fill at all. Neither can protect against a market that is closed overnight. The distance between the trigger and the actual fill is slippage, and it tends to be largest exactly when markets move fastest.

If the fall is d (as a fraction), the rise r needed to recover satisfies (1 minus d) times (1 plus r) equals 1, so r equals d divided by (1 minus d). Maximum drawdown is the largest peak-to-trough fall in a period; drawdown duration is how long it takes to regain the peak, and can matter as much as depth. Consecutive percentage changes multiply, they do not add, which is why a 10 percent loss followed by a 10 percent gain does not return to the start.

Diversification works to the extent that the parts are not perfectly correlated. Correlation is measured from past data and is not constant: it often rises in market stress, so the diversification benefit measured in calm periods can shrink when it is most needed. Concentration risk is the opposite problem: one position, sector, counterparty or idea making up a large share of the total, so a single event has an outsized effect. Hidden concentration (different positions driven by the same underlying factor) is common.

Pre-trade checks are deterministic tests applied to a proposed order (size, price band, limits, duplicates) and can reject it. Post-trade checks monitor what actually happened (positions, exposure, unexpected fills) and raise alarms or trigger reductions. Both are needed: pre-trade stops known problems, post-trade catches the surprises the first set did not foresee.

Risk of ruin is the chance that losses reach a level from which recovery is practically impossible. Intuitively it grows with larger position sizes, with higher variability of outcomes, and with a lower chance of favourable outcomes; it falls when each position is small relative to capital. It is a modelling concept: real markets violate the neat assumptions behind simple formulas, so treat any quoted figure as rough.