Skip to content
QuantCalc Tools
Directory

Kelly Margin Leverage Sizing

Drawdown-Constrained Fractional Kelly: Exact Probability of X% Loss Math

Trading at Full Kelly carries an exact 50% probability of losing half your capital and a 20% probability of an -80% crash; derive exact fractional Kelly drawdown probability laws to cap tail pain.

By Elena Rostova, FRM, CQFReview not recorded · Updated
Drawdown-Constrained Fractional Kelly: Exact Probability of X% Loss Math — QuantCalc Tools 2026 quantitative & supervisory data card
QuantCalc Tools (Kelly Criterion, Initial/Variation Margin Simulators & Liquidation Math): Exact martingale fractional Kelly drawdown power law and Grossman-Zhou HWM floors.Source: QuantCalc Tools Institutional Research DeskLicense: CC-BY-4.0

Every institutional fund mandate specifies an explicit maximum tolerable drawdown limit—for instance, "Never exceed a 15% probability of suffering a -25% peak-to-trough drawdown" or "Enforce a hard dynamic floor at 80% of High-Water Mark." Unconstrained Full-Kelly ($c = 1.0$) fails every institutional drawdown mandate in existence because its infinite-horizon probability of dropping to fraction $z \in (0, 1)$ of peak equity is literally linear: $\mathbb{P}(\mint Wt / M_t \le z) = z$ (50% chance of a -50% crash, 30% chance of a -70% crash!). Using martingale stopping-time theory (Thorp 2006; MacLean, Ziemba & Li 2005) and Grossman-Zhou (1993) Dynamic Drawdown Control, you can invert that exact power law to solve the precise *Fractional Kelly multiplier $c^$** guaranteed to respect your committee's drawdown ceiling! Test fractional multipliers in our Kelly Criterion Calculator and anchor in our Kelly, Margin & Liquidation Hub.

Exact Martingale Power-Law Proof: $\mathbb{P}(\text{Drawdown} \ge D) = (1 - D)^{\frac{2}{c} - 1}$#

Under zero-interest excess log-wealth diffusion $dWt / Wt = \mu f\,dt + \sigma f\,dWt$ where $f = c \cdot f^* = c(\mu / \sigma^2)$ with $c \in (0, 2)$, apply Doob's Optional Stopping Theorem to the exponential martingale $Zt = (Wt / W0)^{-\theta}$ where scale exponent $\theta = \frac{2\mu f}{\sigma^2 f^2} - 1 = \frac{2}{c} - 1 > 0$. As upper target $B \to \infty$, the exact infinite-horizon probability of ever experiencing a peak-to-trough fractional drawdown of depth $D \in (0, 1)$ (hitting $z = 1 - D$ of High-Water Mark) collapses to a jaw-droppingly clean closed-form power law—which inverts directly for *$c^$**:

Formula or model from the articleNot independently verified
Exact Fractional c-Kelly Drawdown Probability Power Law & Inverse Sizing Formula:
  • Let f = c * f* (0 < c < 2), Drawdown Depth D in (0, 1), Remaining Equity Ratio z = 1 - D:
  • 1. Exact Infinite-Horizon Drawdown Probability:
  • P( Max_Drawdown >= D ) = (1 - D)^( (2 / c) - 1 ) = z^( (2 / c) - 1 )
  • 2. Exact Inverse Formula — Solve Max Safe Kelly Multiplier c* Given Allowed (Dmax, alphaprob):
  • Require (1 - Dmax)^(2/c* - 1) <= alphaprob
  • => c* = 2 / [ 1 + ln(alphaprob) / ln(1 - Dmax) ]
  • 3. Grossman & Zhou (1993) Dynamic CPPI-Style Drawdown Floor (Hard 0% Breach Probability at Floor D_hard!):
  • fdynamic(t) = [ 1 / (1 - gamma) ] * [ (mu - rf) / sigma^2 ] * [ (Wt - (1 - Dhard) * HWMt) / Wt ]
  • (Scales exposure linearly with cushion above hard HWM floor Dhard — mathematically CANNOT breach Dhard in diffusion!)
  • Worked Institutional Mandate Calibration:
  • Mandate: "Cap the probability of ever hitting a D_max = 30.0% drawdown (z = 0.70) at alpha = 5.0% (1-in-20 chance)!"
  • 1. Solve c* = 2 / [ 1 + ln(0.05) / ln(0.70) ] = 2 / [ 1 + (-2.99573) / (-0.35667) ] = 2 / [ 1 + 8.3991 ] = 0.2128x Kelly!
  • 2. Check Growth Retained at c* = 0.213x Kelly: c(2 - c) = 0.2128 * (1.7872) = 38.0% of Theoretical Max Growth!
  • 3. Contrast Unconstrained Full Kelly (c = 1.0) on Same 30% Drawdown: P(DD >= 30%) = (0.70)^(2/1 - 1) = 70.0% CERTAINTY!

Worked Institutional Mandate Calibration: Why 5% Risk at -30% Demands Fifth-Kelly#

Look at that result: an allocator demanding < 5% probability of a -30% drawdown MUST trade at 0.213x Kelly (~1/5th Kelly), NOT Full Kelly (70.0% probability!) and NOT even Half-Kelly ((0.70)^3 = 34.3% probability!).

Comparison between static constant-fraction c-Kelly drawdown probabilities and Grossman-Zhou dynamic cushion-scaled HWM floor control
Comparison between static constant-fraction c-Kelly drawdown probabilities and Grossman-Zhou dynamic cushion-scaled HWM floor controlVerified Exhibit

Exact Probabilistic Drawdown Matrix Across Fractional Kelly Multipliers#

Verify how rapidly tail drawdown probabilities collapse as you step down from 1.00x Full Kelly to 0.20x Fifth-Kelly via $(1 - D)^{2/c - 1}$:

Quantitative Distribution · Exponent ($2/c - 1$)
Exponent ($2/c - 1$)P(DD >= 20%)
011.00x Full Kelly$1.00$020.50x Half-Kelly$3.00$030.33x Third-Kelly$5.00$040.25x Quarter-Kelly$7.00$050.20x Fifth-Kelly$9.00$
Fractional Kelly Tier ($c$)Exponent ($2/c - 1$)P(DD >= 20%)P(DD >= 30%)P(DD >= 50%)P(DD >= 80%)Excess Growth Kept $c(2-c)$
1.00x Full Kelly$1.00$$80.0\%$$70.0\%$$50.0\%$$20.0\%$$100.0\%$
0.50x Half-Kelly$3.00$$51.2\%$$34.3\%$$12.5\%$$0.80\%$$75.0\%$
0.33x Third-Kelly$5.00$$32.8\%$$16.8\%$$3.13\%$$0.03\%$$55.6\%$
0.25x Quarter-Kelly$7.00$$21.0\%$$8.24\%$$0.78\%$$< 0.002\%$$43.8\%$
0.20x Fifth-Kelly$9.00$$13.4\%$$4.04\%$$0.20\%$$\approx 0\%$$36.0\%$
Probability of suffering a 50 percent peak-to-trough drawdown collapsing from 50.0 percent at Full Kelly to 0.20 percent at One-Fifth Kelly
Probability of suffering a 50 percent peak-to-trough drawdown collapsing from 50.0 percent at Full Kelly to 0.20 percent at One-Fifth KellyVerified Exhibit

Static Fractional Kelly vs Dynamic Grossman-Zhou Cushion Scaling: Crucial Trade-Off!#

Compare the two philosophies of drawdown control:

  1. *Static Fractional Kelly ($c^$ constant)*: When your strategy drops $-15\%$, static $0.25\times$ Kelly keeps its exact `0.25 f*` leverage intact—participating $100\%$ in V-shaped mean-reversion rebounds (asymmetric drawdown recovery math), at the cost of a small non-zero tail probability ($\alpha = 5\%$) of crossing $D_{\max}$.
  2. Grossman-Zhou (1993) / Cvitanic-Karatzas Dynamic Cushion Control: Scales exposure proportionally to remaining drawdown cushion $(D{\text{hard}} - Dt) / D{\text{hard}}$. It guarantees $0\%$ diffusion breach of $D{\text{hard}}$, but behaves pro-cyclically (selling into drawdowns like CPPI!) and risks "cash lock" if an overnight jump wipes out the cushion! Institutional CTAs blend both: trade 0.25x static Kelly down to -10% drawdown, then engage linear cushion tapering between -10% and -20%!

Connecting Drawdown Caps to Exchange Close-Outs & Brownian Nulls#

Verify that your chosen $D_{\max}$ sits safely above statutory FCA 50% retail margin close-out thresholds and Magdon-Ismail Brownian Expected Maximum Drawdown bounds, audit your Omega & Calmar ratios, and check firm leverage disclosures in our rankings and methodology.

Frequently asked questions

Why is the formula (1 - D)^(2/c - 1) independent of my strategy’s Sharpe Ratio and volatility?

Because volatility sigma and Sharpe SR are ALREADY normalized inside Full-Kelly f* = SR / sigma! Regardless of whether SR is 0.5 or 2.0, scaling to Full Kelly sets drift equal to variance (mu_p = sigma_p^2), making the dimensionless drift-to-variance ratio 2*mu_p / sigma_p^2 depend purely on the fraction c = f / f*!

Does (1 - D)^(2/c - 1) apply over 1 year or over an infinite lifetime?

It is the exact INFINITE-HORIZON probability of EVER hitting drawdown D across an eternal track record—making it a rock-solid conservative upper bound for finite 3-year or 5-year institutional horizons! Over a short 1-year window with high Sharpe (SR >= 1.5), finite-horizon drawdown probability is substantially lower.

How should I size a proprietary trading firm challenge that terminates the account on a -10% trailing drawdown?

Plug D_max = 0.10 and survival failure tolerance alpha = 0.05 into c* = 2 / [1 + ln(0.05) / ln(0.90)] = 2 / [1 + 28.43] = 0.068x Kelly (~1/15th Kelly)! Attempting to trade Half-Kelly (0.50x) against a tight -10% termination barrier carries a (0.90)^3 = 72.9% failure probability!

What is "cash lock" in Grossman-Zhou dynamic drawdown control and how do institutions prevent it?

In pure Grossman-Zhou / CPPI, position size equals Multiplier * (Current Equity - Hard Floor). If an overnight price gap drops Current Equity down to or below Hard Floor, position size drops to ZERO (100% cash forever), making recovery mathematically impossible! Desks prevent cash lock by resetting HWM floors annually or clamping minimum exposure at 0.10x Kelly.

Publication record

Published
13 November 2026
Updated
14 November 2026
Author
Elena Rostova, FRM, CQF
Review
Review not recorded
Automated fact check
Passed in the stored pipeline record
Source URLs
4 recorded

AI assistance recorded: anthropic/claude-opus-4.6. This field does not establish human review or source verification.

Academic & Institutional Attribution

Cite This Research & Embed Live Data Card

Open Attribution License (CC BY 4.0)

Standard academic & institutional research paper citation format.

Elena Rostova, FRM, CQF. (2026, November 14). Drawdown-Constrained Fractional Kelly: Exact Probability of X% Loss Math. QuantCalc Tools. https://quantcalctools.com/guides/drawdown-constrained-fractional-kelly-probability-of-50pct-loss
Explore Kelly Margin Leverage Sizing Hub (20) →

QuantCalc Tools Data, Benchmarks & Contributor Records

Drawdown-Constrained Fractional Kelly Formula (Grossman-Zhou) | QuantCalc Tools