Advanced Risk Models

• The 1% rule is a deliberately conservative baseline that protects capital while an edge is being established. Advanced risk models allow position size to scale as the journal confirms a specific setup type's win rate and reward-to-risk ratio. They never replace the 1% rule on unproven setups.

• The Kelly Criterion is the mathematical formula that connects win rate and reward-to-risk to optimal position size. Full Kelly is almost always too aggressive for retail trading. Fractional Kelly, typically one-quarter to one-half of the Kelly output, is the professional implementation.

• The most practical advanced risk model for SMC traders groups trades by setup type, calculates each type's statistics separately, and applies fractional Kelly to each. Stop hunt setups and fair value gap entries have different statistical profiles and should have different risk allocations.

Risk management does not end with the one percent rule. Once a trader has developed a strategy with a proven edge, more advanced risk models allow position sizes to grow in line with that edge. The goal is to allocate more capital to setups with stronger statistical profiles and less to those with weaker ones, while maintaining the account-level protections built in the foundational framework.

If you have not yet read Market Structure, start there before continuing.

A risk model is a set of rules that determines how much capital to commit to each trade based on factors such as the strength of the edge, the size of the stop, recent performance, and current market conditions. The basic risk model in this series, the 1% rule combined with daily loss limits and drawdown thresholds, is the correct framework for all unproven setups. Advanced models build on that foundation once the statistical basis for a larger size has been established through a sufficient sample in the journal.

Advanced risk models do not override the discipline of the 1% baseline. They graduate from it as the evidence base grows.

Starting with the basics

Before applying any advanced model, every element of the basic risk framework should be working correctly. Advanced models amplify results in both directions. If the basic framework is inconsistent, a more complex model will make performance worse, not better. This means the Rule Adherence Score is above 90%, the daily loss limit and cooldown rules are applied consistently, and the journal contains at least 50 completed trades of the specific setup type before any Kelly calculation is run.

The Kelly Criterion: the mathematical foundation

The Kelly Criterion is the formula that connects a trading strategy's statistical profile to the optimal fraction of capital to risk on each trade. The formula is: f equals W minus (L divided by R). Where f is the optimal fraction of capital to risk, W is the win rate as a decimal, L is the loss rate (1 minus W), and R is the average reward-to-risk ratio.

For a setup with a 52% win rate and a 3.8:1 average reward-to-risk ratio, the Kelly output is: f = 0.52 minus (0.48 divided by 3.8) = 0.52 minus 0.126 = 0.394, or 39.4% of account per trade. Full Kelly at 39.4% is dramatically too aggressive for any retail forex account. QuantStrategy.io (January 2026) confirms: the full Kelly fraction assumes perfect statistical knowledge, and in real markets where win rates and R-multiples fluctuate, using the full percentage leads to rapid ruin.

Fractional Kelly: the professional implementation

Most professional traders apply Kelly at one-quarter to one-half of the formula's output. For the 39.4% example, quarter-Kelly is 9.9% and half-Kelly is 19.7%. FXNX (February 2026) provides the calibration guidance: feed the formula the median outcomes rather than the best month, then apply the fractional multiplier. If the journal shows a median win rate of 48% and a median R-multiple of 3.2 across 80 trades, that enters the formula, not the optimistic figures from the best 20 trades.

The practical implementation: start at 1% on any new setup type. After 50 trades, run the Kelly calculation on that specific setup's median outcomes. Scale toward quarter-Kelly gradually, adding 0.25% per 20 confirmed trades, stopping if the drawdown metrics deteriorate.

Setup-specific risk allocation

FXNX (February 2026) identifies the most important refinement: stop treating all setups as one strategy. A stop hunt entry on the daily chart during the London open has a different statistical profile than a fair value gap entry on the four-hour chart during the New York session. Running a single Kelly calculation across all SMC setups blends the profiles and produces an allocation that is optimal for neither.

The professional approach groups trades by setup type, calculates each type's win rate and average R-multiple from at least 50 data points in the journal, and applies fractional Kelly separately to each. The Kelly input uses only the Rule Adherence Score filter: trades where every rule was followed.

ATR-based dynamic sizing

WeMasterTrade (May 2026) confirms the professional application: fixed technical stop-loss levels fail when market conditions shift from low-volatility to high-volatility environments. The ATR-based adjustment: calculate the Average True Range over the last 14 sessions. If the current ATR is more than 50% above its 60-session average, reduce the standard position size by 25 to 50%. Within normal range, apply standard sizing. The inverse does not apply automatically: below-average ATR does not justify larger positions without other confirming factors.

Drawdown-based continuous scaling

The graduated drawdown response from the Risk Management article, halving position size at 8 to 10% drawdown and stopping at 15 to 20%, remains the primary protection. The advanced version scales position size continuously rather than at discrete thresholds. For every 1% of drawdown from the 90-day account peak, reduce the standard position size by 5 to 10% of the baseline. This continuous scaling reaches the same floor as the fixed threshold but arrives there gradually, reducing abrupt shifts in exposure.

The data requirement

None of the above applies to unproven setups. The minimum data requirement before any Kelly-based scaling: 50 completed trades of the specific setup type, all recorded as rule-following in the journal. A 30-trade sample produces unreliable estimates. The 50-trade minimum is the starting point for a basic calculation. 80 to 100 trades produce a statistically stable estimate sufficient for gradual scaling.

The following shows the setup-specific Kelly calculation applied to two SMC setup types confirmed by a journal across 80 completed trades each.

The scaling is gradual because the journal's statistical estimates improve with each additional data point. The first 50 trades produce an estimate with wide uncertainty. By 100 trades, the estimate is more reliable. By 200 trades, it is stable enough to approach quarter-Kelly with confidence. The Kelly formula is only as reliable as the inputs. FXNX states it directly: Kelly is only as honest as the inputs you give it.

The 1% rule protects capital. Advanced risk models compound it. A trader running SMC stop hunt setups at a confirmed 52% win rate and 3.8R average across 200 trades, applying fractional Kelly at 5% rather than 1%, will compound account equity significantly faster than the same trader applying a flat 1% rule indefinitely. At 1% risk, 100 winners at 3.8R and 100 losers at 1R produce a net of 280R. At 5% risk, the same outcome produces 1,400R. The difference compounds only because the journal confirmed the edge first.

The risk of applying advanced models without sufficient data is symmetric: if the 80-trade sample overestimates the edge and the real edge at 300 trades is lower, the scaled position size will produce a deeper drawdown than the 1% baseline would have. This is why the scaling is gradual, why it resets if drawdown metrics deteriorate, and why the minimum data requirement is non-negotiable.

For the foundational framework that governs position sizing before any Kelly scaling is applied, read Risk Management alongside this topic.

• The Kelly Criterion connects win rate and reward-to-risk to optimal position size. Full Kelly is too aggressive for retail trading. Fractional Kelly at one-quarter of the output, applied separately to each setup type after a minimum of 50 confirmed data points, is the professional implementation.
• ATR-based dynamic sizing adjusts position size for current volatility relative to the 60-session average. Drawdown-based scaling reduces position size continuously for every 1% of drawdown from the 90-day account peak. Both operate within the 1% baseline floor.
• Advanced risk models amplify results in both directions. They work when the basic framework is already in place and performing consistently. They fail when applied before the basics are working. Risk models are tools. They work when you understand them and apply them consistently.

Advanced risk models applied at the individual setup level connect to the broader question of how multiple positions interact across a trading account.

• Portfolio approach - Learn how to manage multiple open positions simultaneously, how correlation between positions multiplies actual account risk beyond the individual trade level, and how a portfolio-level risk framework operates alongside individual position sizing.

• Risk Management - Revisit the foundational framework and see how the 1% rule, daily limits, and drawdown thresholds form the floor on which advanced risk models are built.

Before applying any advanced model, answer three questions: Is your basic risk framework working correctly, meaning are you applying the 1% rule, the daily loss limit, and the cooldown rule consistently above 90% rule adherence? Do you have at least 50 confirmed data points on the specific setup type you intend to scale, filtered to only include rule-following trades? Have you calculated the Kelly output and quarter-Kelly allocation on paper before changing any live trade size? If all three answers are yes, open your journal, filter by your most frequent setup type with the Rule Adherence Score filter applied, and run the Kelly calculation. Scale by 0.25% per 20 confirmed trades. Risk models are tools. They work when you understand them and apply them consistently. They fail when they are applied before the basics are in place. Practice the scaling process on a demo account for four weeks before applying it to a live account.