Regime-Conditional Position Sizing

Regime-Conditional Position Sizing

The single most underused lever in trading is varying size by regime. Most traders size by ticker conviction or by setup quality and treat regime as background. The flip is the entire game: regime sets the size envelope, conviction allocates within that envelope.

5.5.1Size by Regime, Not by Ticker

The traditional sizing question is “how confident am I in this trade?” The regime-conditional sizing question is “what is the maximum risk per trade allowed by the current regime, and within that ceiling, how does this trade rank?” The reframe is from a continuous individual-trade decision to a tiered system in which the regime sets the ceiling and the conviction sets the percentage of the ceiling claimed.

The reason is empirical. Across thousands of trades and multiple regime cycles, the variance in trade outcomes attributable to regime far exceeds the variance attributable to setup quality within a regime. A high-quality setup taken in a Crisis regime has a meaningfully lower expected return and a larger downside tail than a mediocre setup taken in a Bull regime. Sizing the high-quality Crisis trade as if it were a high-quality Bull trade is the most common way good frameworks produce bad outcomes.

The discipline is to set three risk envelopes — Bull, Neutral, Crisis — and live within whichever envelope the current regime dictates. The envelope governs every sizing decision until the regime changes; trade-by-trade conviction operates inside that envelope. The envelope is non-negotiable; the within-envelope allocation is where judgment lives.

Regime sets the ceiling. Conviction allocates under the ceiling. The trader who reverses this order eventually meets the regime that punishes the reversal.

5.5.2The Three Sizing Tiers

The AZTMM regime sizing structure uses three tiers anchored to the regime posterior and the current account size. These are starting points; individual books with different baseline volatility can scale the percentages, but the relative ratios should remain.

Bull regime (posterior > 0.75 in Bull state). Full risk per trade is 1% of account. Maximum gross heat is 5% across all open positions. New trades acceptable on any regime-matched setup that meets quality criteria. The market is in a state where mean reversion is shallow, drawdowns are narrow, and trends persist. The framework is designed to extract premium for that condition.

Neutral regime (posterior > 0.50 in Neutral state, or no state above 0.75). Risk per trade is reduced to 0.5%. Maximum gross heat is 2.5%. New trades require both quality and corroboration; mediocre setups are passed. The market is uncertain, mean reversion is more aggressive, and the framework cannot count on trends to bail out marginal entries. Cutting size in half is the cheapest insurance available.

Crisis regime (posterior > 0.50 in Crisis state). Risk per trade is reduced to 0.25%. Maximum gross heat is 1%. New entries are limited to the highest-quality setups only; reflexive directional trades are stopped entirely. The market is in a state where realized volatility can outrun your stop in a single session, and the asymmetric tail is meaningful. Sizing down by 4x preserves the framework for the regime cycle to come.

5.5.3Posterior-Weighted Risk

Within a sizing tier, the posterior probability of the leading regime acts as a continuous multiplier. The math is direct: effective risk = tier_risk × posterior_confidence. A 1% Bull-tier trade taken at 0.89 posterior confidence carries an effective risk of 1% × 0.89 = 0.89% of account. Taken at 0.78 posterior, the effective risk drops to 0.78%. The size on the trade is reduced proportionally.

This may sound like double-counting (the tier already discounts uncertain regimes), but it is the discipline that prevents the cliff effect. Without posterior weighting, a regime that drifts from 0.91 to 0.76 over a week stays at full Bull sizing all week, and then drops to half the moment it crosses below 0.75. With posterior weighting, the sizing scales smoothly from 0.91% effective risk down to 0.76% effective risk, and the cliff at the threshold becomes a step from 0.76% Bull to 0.5% Neutral — a third smaller, not a half smaller. The smoother sizing path means the framework does not over-react to small posterior wobbles around a tier boundary.

The posterior weighting also applies symmetrically on the Crisis side. A 0.25% Crisis-tier trade taken at 0.85 Crisis posterior confidence carries an effective risk of 0.21% of account; taken at 0.55 Crisis posterior, it carries 0.14% of effective risk. The downside is more conservatively sized when the model is more confident the regime is hostile.

5.5.4Compounding Across Positions

Per-trade sizing is half the discipline; the other half is the maximum gross heat across all open positions. Five 1% trades that all go against you on the same day — which happens, in any regime — produce a 5% loss in the account. The gross heat ceiling is the safety mechanism that limits the worst-case correlated draw to a recoverable percentage.

The Bull tier maximum heat of 5% means that if you are at 5 open positions of 1% each, the framework requires you to close existing risk before adding new. The Neutral tier maximum of 2.5% means you might be at 5 positions of 0.5% each at saturation, or 3 positions at near full size with room for one more partial. The Crisis tier maximum of 1% means you have meaningful room for only 2 to 4 positions at the reduced sizing.

Heat is computed as the sum of risked dollars at each position’s stop, not as the notional or as the option premium. A trade with a wider stop carries more heat than the same dollar trade with a tighter stop. This is the framework treating all positions as commensurable in risk space, regardless of how they are expressed instrument-by-instrument.

5.5.5Tying Kelly Fraction to Posterior

The Kelly criterion gives an optimal sizing fraction for a known edge with a known win/loss asymmetry. The classic full-Kelly fraction is aggressive enough that virtually no real trader runs it; quarter-Kelly or eighth-Kelly is more typical because the practical estimation error of the edge inputs makes full-Kelly behave too violently.

The regime overlay improves the Kelly framework substantially. The posterior is already a probability that the regime favorable to the strategy is in effect. Multiplying the Kelly fraction by the posterior gives a regime-adjusted Kelly: f_adj = f_kelly × posterior. A strategy with a quarter-Kelly fraction of 1.5% in its favorable regime, taken at a 0.78 posterior, sizes at 1.5% × 0.78 = 1.17%. The discipline is to use this only as a sanity-check on the tier-based size; if Kelly disagrees with the tier by more than a factor of two in either direction, investigate the inputs before sizing.

The framing matters. Kelly is a continuous instrument in a discrete-tier framework. The two should converge in steady-state regimes (Bull at 0.89 should produce a tier size and a Kelly size within 20% of each other). Divergence between the two is a signal that either the tier is wrong (regime is not what the posterior says) or the strategy edge inputs are wrong (the per-trade win/loss has shifted). Either way, it is a journal investigation, not a sizing override.

5.5.6Common Mistakes

• Sizing by ticker conviction without checking the regime ceiling. The ceiling is non-negotiable; ticker conviction is within-ceiling.
• Reading the regime label without checking posterior confidence. A 0.62 Bull posterior is not a Bull-tier sizing situation; it is Neutral-tier.
• Ignoring max gross heat. Per-trade sizing alone does not bound correlated drawdowns.
• Not posterior-weighting within a tier. A 0.91 Bull and a 0.77 Bull are different sizing situations even at the same tier.
• Treating Kelly as the answer rather than a corroboration check. Kelly works in known-edge worlds; trading is not one.
• Sizing down only after a loss. The regime sizing should adjust on the regime, not on yesterday’s P&L.