Algorithmic Skepticism: How to Trade Based on Mathematics, Not Emotions

Risk management is not a protective add-on to a trading strategy, but its load-bearing frame. Without it, any system of analysis, however sophisticated, turns into a lottery with a delayed explosion. In this chapter we will examine how position size is calculated, how volatility should determine the level of a stop-loss, how market liquidity intrudes on these calculations, and why correlation between open positions can imperceptibly transform a diversified portfolio into one enormous concentrated bet.

The Mathematics of Position Size: From Intuition to Formula

The first and most common mistake of a retail trader is determining position size "by eye," based on whatever sum feels psychologically comfortable to risk. This approach ignores a fundamental principle: position size must be a derivative of three variables—the size of the capital, the allowable risk per trade, and the distance from the entry point to the stop-loss. These three parameters form a rigid formula that leaves no room for emotional adjustments.

Allowable risk per trade is the percentage of capital a trader is willing to lose if the stop-loss is triggered. Standard practice limits this figure to one or two percent of the deposit. The number is not arbitrary: it is derived from the statistics of losing-trade sequences. Even a strategy with acceptable mathematical expectation is capable of generating five, seven, or sometimes ten losing trades in a row—this is not an anomaly but a normal manifestation of randomness within a limited sample. If risk per trade is two percent, a series of ten consecutive losses will reduce capital by roughly eighteen percent—painful, but not fatal. If risk per trade is ten percent, however, the same series will wipe out the deposit almost entirely, and recovery becomes mathematically all but impossible, since the percentage gain required to compensate for a deep drawdown grows nonlinearly relative to the drawdown itself.

The central lesson is this: a fifty percent loss requires a hundred percent gain to return to the starting point. This is not abstract arithmetic but the reason why excessive risk per trade amounts to slow suicide for capital even under a formally profitable strategy.

The calculation of position size proceeds in reverse—from risk to volume. First, the trader determines the sum they are willing to risk in absolute terms—a percentage of capital. Then the distance from the entry point to the stop-loss is determined in price units. Dividing the allowable risk by this distance gives us the quantity of the asset that can be purchased without exceeding the risk limit. This mechanism automatically adjusts position volume according to how wide the stop-loss must be: the farther the stop sits from the entry price, the smaller the position volume, and vice versa. This inverse relationship is not a technical detail but a fundamental principle that protects capital from the emotional urge to "take more" at moments when the chart looks especially convincing.

This calculation must be performed anew for each trade, not used as a fixed number of lots. The market is constantly reshaping its internal geometry, and what yesterday required a fifty-point stop may today require a stop of one hundred fifty points because of increased volatility. Ignoring this recalculation is one of the most common reasons formally disciplined traders still lose capital: they observe the rule of "no more than two percent risk" but forget that position size must adapt to the shifting price context.

Volatility as the Foundation for Stop-Loss Calculation

A stop-loss established without regard to the current volatility of the instrument is a stop-loss set blind. A fixed number of points, identical for a calm market and a market in a state of turbulence, is doomed either to trigger too frequently on normal price fluctuations or to leave excessive risk during moments of anomalous calm.

The Average True Range is a fundamental tool for calibrating the stop-loss to current market conditions. It shows the typical amplitude of price fluctuations over a given period, including gaps and breaks between sessions. The logic is simple and mathematically sound: a stop-loss set at a distance smaller than the average range will be systematically knocked out by ordinary market noise that has nothing to do with a trend reversal. A trader using such a tight stop will lose not because their hypothesis about the direction of movement is wrong, but because the very construction of the stop ignores the physics of market fluctuations.

The practical rule is to tie the stop-loss to a multiplier of the Average True Range—for example, one and a half or two ranges from the entry point. If an asset's range is fifty points, a stop-loss set at a distance of seventy-five points accounts for the natural amplitude of noise, leaving room for random fluctuations that do not invalidate the trading idea. But here it is important to return to the previous section: the wider the stop, the smaller the position size must be so that the risk per trade remains within the established percentage of capital. Volatility and position size are not two separate parameters but two sides of the same formula, which must be calculated in tandem.

The dynamic nature of volatility needs examination on its own terms. It is not static—periods of calm are replaced by periods of high turbulence, and an algorithm or trader using an outdated value of the average range risks applying yesterday's market parameters to today's conditions. Volatility has a tendency to cluster: a period of high amplitude is often followed by a continuation of elevated activity, not an instantaneous return to calm. This means that the range calculation must be updated on a rolling basis, not taken as a constant over an extended period.

A separate trap is the use of a percentage stop-loss without any tie to volatility at all. A fixed percentage—for example, a one percent stop from the entry price—seems like a universal solution, but in reality ignores the fact that the same percentage can be either too tight or excessively wide depending on how prone the asset is to fluctuations in the first place. An instrument with historically low volatility, trading in a narrow range, requires a tighter stop in absolute terms than an instrument prone to sharp moves. Mixing these approaches is a frequent reason why formally "sound" risk management still leads to uneven results across different assets in a portfolio.

Adaptation to Changing Liquidity

The calculation of position size and stop-loss rests on the assumption that the market will execute orders at the expected price. This assumption holds true only under conditions of sufficient liquidity. When market depth becomes depleted, all the risk mathematics built on theoretical calculations begins to diverge from the reality of order execution.

Slippage is the difference between the price a trader anticipated and the price at which an order actually executed. When the order book is deep, this difference is minimal and predictable. But in moments of liquidity depletion—during low-volume sessions, before the release of important economic data, during holiday trading, or during off-hours on cryptocurrency exchanges—that same order can execute at a price significantly worse than expected. A stop-loss set at a certain distance from the entry price may, in a thin order book, trigger at a price substantially different from the set level, turning a planned two-percent risk into an actual risk of five or seven percent.

This means that position sizing cannot be limited to price volatility alone—it must also account for the volatility of liquidity itself. The practical approach consists of monitoring order book depth and trading volume across different times of day and market sessions. If current trading volume is substantially below the median value for the preceding period, this signals a need to reduce position size or to widen the planned stop-loss to account for expected slippage. This adjustment is not excessive caution—it is a direct consequence of the fact that order execution costs are part of the overall trade risk, not a separate, secondary quantity.

This problem is especially acute in assets with uneven liquidity distribution throughout the day. The Asian session in the foreign exchange market, off-hours on cryptocurrency exchanges, the period between the close of one national session and the opening of another—all these are moments when even a relatively modest market order can move the price by an amount that would be statistically insignificant during peak activity hours. An algorithm or trader ignoring this unevenness and applying identical risk parameters at any time of day will sooner or later encounter a situation where a formally correct position calculation proves practically impossible to execute at the expected price.

The liquidity problem also has an inverse side: the size of one's own position relative to the average trading volume of the instrument. If the volume of a planned trade constitutes a notable share of the asset's average daily turnover, the very opening or closing of the position can move the price against the trader, creating costs that were not factored into the initial risk calculation. This is especially relevant when working with assets of medium and low capitalization, where a position large relative to the market transforms the trader from an observer of market dynamics into a direct participant influencing their own execution outcome.

Correlation of Risk and Capital Allocation

Calculating risk for each individual trade is a necessary but insufficient condition for capital preservation. Danger emerges at the portfolio level, when several formally independent positions turn out to be secretly interconnected through the correlation of their underlying assets. A trader who has opened five positions with a two percent risk on each, confident that total portfolio risk amounts to ten percent, may discover that the actual risk is significantly higher if all five instruments move in sync at the moment of market stress.

Correlation between assets is not a constant quantity. During calm periods, different asset classes may demonstrate weak interdependence, creating an illusion of diversification. But in moments of market panic, liquidity crises, or large-scale sell-offs, correlation among most risky assets tends toward one: investors simultaneously reduce exposure across the entire portfolio, regardless of the fundamental differences between individual positions. In such moments, diversification built on historical correlation coefficients from calm periods ceases to perform its protective function—and this happens exactly when protection is needed most.

The practical takeaway from this observation is that open positions should be viewed not as isolated bets but as a unified system, where aggregate risk can substantially exceed the sum of the nominal risks of individual trades. Before opening a new position, it is reasonable to assess how it interacts with existing ones: if the new trade adds exposure in a direction correlated with already-open positions, the portfolio's effective risk grows disproportionately to the nominal risk of a single trade. This is especially important when working with multiple instruments within a single asset class—for example, several currency pairs pegged to one reserve currency, or several cryptocurrency assets moving in the wake of the market's dominant asset.

Capital allocation between strategies and instruments should account for this hidden interdependence. Formal diversification by the number of open positions is not the same as diversification by sources of risk. A portfolio of ten positions all responding to the same macroeconomic factor represents a concentrated bet disguised as a distributed one. True diversification requires seeking out assets and strategies that respond to different factors, different time horizons, and different sources of volatility—only then does the aggregate risk of the portfolio actually decrease, rather than simply appearing to be spread across a greater number of rows in the trading terminal.

Dynamic Risk Adjustment Based on Performance

Risk management does not end at the moment of opening a position—it must continue throughout an entire series of trades, responding to changes in the state of capital. A static approach, in which risk per trade remains unchanged regardless of current drawdown or a streak of successful results, ignores an important principle: the capacity of capital to withstand risk changes together with its state.

After a series of losing trades, sound practice consists in reducing risk per trade rather than maintaining it at the previous level in an attempt to recover losses more quickly. This is counterintuitive from a psychological standpoint—the desire to compensate for a loss faster pushes toward increasing stakes—but mathematically, it is the reduction of risk during a drawdown period that protects capital from the very nonlinear recovery effect discussed at the beginning of this chapter. The deeper the drawdown, the more conservative the approach to subsequent trades must become, since the margin of safety for capital shrinks while the required percentage gain for recovery grows.

Similarly, a series of successful trades should not be automatically interpreted as a signal to increase risk. Capital growth following a fortunate period is often interpreted as confirmation of system effectiveness, which provokes a gradual increase in position size. However, system effectiveness is measured not by a short streak of favorable outcomes but by statistics over a sufficiently long horizon that encompasses different market phases. Increasing risk at the peak of euphoria following a streak of successful trades is the classic mechanism by which traders give back to the market—in a matter of days during an unfavorable period—profits accumulated over months.

A more robust approach consists in tying the size of risk not to the emotional state following a series of trades but to objective metrics—current drawdown relative to the historical maximum of capital, volatility of recent results, changes in market regime. Such a system responds to data rather than to the trader's psychological state, which brings us back to the fundamental principle of risk management: capital must be managed as an engineering system with feedback, not as an object of intuitive decisions made under the influence of the last few trades.

Конец ознакомительного фрагмента.

Текст предоставлен ООО «Литрес».

Прочитайте эту книгу целиком, купив полную легальную версию на Литрес.

Безопасно оплатить книгу можно банковской картой Visa, MasterCard, Maestro, со счета мобильного телефона, с платежного терминала, в салоне МТС или Связной, через PayPal, WebMoney, Яндекс.Деньги, QIWI Кошелек, бонусными картами или другим удобным Вам способом.