Skip to content
QuantCalc Tools
Directory

Kelly Margin Leverage Sizing

Risk of Ruin Math: Fixed-Dollar Markov Absorption vs Fixed-Fractional Sizing

Betting a constant $2,000 per trade carries an exponential Markov absorption Risk of Ruin (q/p)^N, whereas scaling risk dynamically as a fixed percentage of live equity transforms ruin geometry.

By Elena Rostova, FRM, CQFReview not recorded · Updated
Risk of Ruin Math: Fixed-Dollar Markov Absorption vs Fixed-Fractional Sizing — QuantCalc Tools 2026 quantitative & supervisory data card
QuantCalc Tools (Kelly Criterion, Initial/Variation Margin Simulators & Liquidation Math): Feller Markov (q/p)^U, asymmetric polynomial roots r_0^U, and fractional barriers.Source: QuantCalc Tools Institutional Research DeskLicense: CC-BY-4.0

Why can a profitable trader with a verified +55% win rate still blow up an account in six months? Because Risk of Ruin (RoR)—the mathematical probability of hitting an absorbing termination barrier (total $0 wipeout, a prop-firm -10% max loss gate, or an institutional -25% mandate termination) before reaching target wealth—depends hyper-sensitively on whether you bet Fixed Dollars per trade vs Fixed Percentage of Current Equity, and on payoff variance relative to drift! Calibrate survival-safe leverage in our Kelly Criterion Calculator and anchor in our Kelly, Margin & Liquidation Hub.

Exact Derivations: Symmetric Markov Ruin, Asymmetric Characteristic Roots & Brownian Ruin#

Distinguish three classic Risk-of-Ruin regimes:

  1. Classic Feller (1968) Fixed-Unit Gambler's Ruin ($R = 1:1$): Starting with $U$ fixed risk units (e.g., $50,000 capital divided into $2,500 fixed bets $\implies U = 20$ units) with win probability $p > 0.50$ and loss probability $q = 1 - p < 0.50$, second-order difference equation $Ru = p R{u+1} + q R_{u-1}$ has characteristic roots $\lambda = 1$ and $\lambda = q/p < 1$.
  2. Asymmetric Payoff Fixed-Unit Ruin ($+R$ units won vs $-1$ unit lost): Characteristic polynomial $p \lambda^{R+1} - \lambda + q = 0$ has a unique root $\lambda* \in (0, 1)$ whenever expectancy $p R - q > 0$, giving $\text{RoR}(U) = (\lambda*)^U$.
  3. Continuous Brownian / Fixed-Fractional Barrier Drawdown Ruin: Scaling bet size $f$ dynamically with current wealth $Wt$ can never literally touch $0.00 in continuous diffusion, so "Ruin" is defined as hitting termination barrier $B = W0(1 - D_{\text{stop}})$!
Formula or model from the articleNot independently verified
Three Exact Institutional Risk-of-Ruin (RoR) Equations:
  • 1. Classic Even-Payoff Fixed-Unit Gambler's Ruin (U Risk Units to Termination Barrier, R = 1:1):
  • Infinite Target (Survival Forever): RoR = (q / p)^U (if p > q); RoR = 100.0% (if p <= q!)
  • Finite Target M Units (M > U) : RoR = [ (q/p)^U - (q/p)^M ] / [ 1 - (q/p)^M ] (When p = 0.5: RoR = 1 - U/M)
  • 2. Asymmetric Payoff (+R Win, -1 Loss, p*R > q) — Exact Root + Brownian Diffusion Approximation:
  • Exact Polynomial Root : Solve unique root r0 in (0, 1) of p * r^(R + 1) - r + q = 0 => RoR = (r0)^U
  • Diffusion Closed Form : Drift muu = p*R - q, Var sigmau^2 = p*R^2 + q - mu_u^2
  • RoR(U) ~= exp( -2 * muu * U / sigmau^2 )
  • 3. Fixed-Fractional % Sizing Barrier Ruin (Risking fraction r of LIVE equity per trade, Stop Barrier = 1 - D_stop):
  • Effective shrinking units to barrier: Ueff = ln(1 - Dstop) / ln(1 - r) (> U_fixed because losses shrink $ bet!)
  • Continuous c-Kelly Exact Formula : RoR(Dstop) = (1 - Dstop)^( (2 / c) - 1 )
  • Worked Prop-Account Survival Audit ($100,000 Account, Hard Termination at $90,000 => $10,000 Risk Budget, p = 55%, R = 1:1):
  • Over-Aggressive Risk ($2,500/trade -> Only U = 4 Fixed Units to Termination!):
  • RoR = (0.45 / 0.55)^4 = (0.81818)^4 = 44.81% CHANCE OF FAILING DESPITE +55% WINNING EDGE!
  • Disciplined Institutional Risk ($500/trade -> U = 20 Fixed Units to Termination):
  • RoR = (0.45 / 0.55)^20 = (0.81818)^20 = 1.80% Ruin Probability (25x safer simply by cutting ticket size 5x!)

Look at the brutal exponential sensitivity of $(q/p)^U$ in that worked prop-firm audit: changing NOTHING about your trading signal (p = 55%, R = 1:1), risking $2,500 (2.5% toward a -10% floor, $U = 4$ units) carries a 44.81% ruin rate, whereas risking $500 (0.5%, $U = 20$ units) slashes Risk of Ruin to 1.80%—and $250 ($U = 40$ units) drops Risk of Ruin to 0.032% (1-in-3,080)!

Comparison of linear dollar-unit depletion in Fixed-Dollar betting versus geometric Zeno deceleration in Fixed-Fractional percentage sizing
Comparison of linear dollar-unit depletion in Fixed-Dollar betting versus geometric Zeno deceleration in Fixed-Fractional percentage sizingVerified Exhibit

Why Fixed-Fractional Anti-Martingale Sizing Beats Fixed-Dollar Sizing to Any Barrier $D_{\text{stop}}$#

Suppose your committee terminates your sleeve at a $D_{\text{stop}} = -30\%$ drawdown.

  • Under Fixed-Dollar $2% of initial capital, hitting -30% takes exactly $30\% / 2\% =$ 15.0 net losing units.
  • Under Fixed-Fractional 2% of current live equity, each loss multiplies equity by 0.98, so hitting 0.70 takes $U{\text{eff}} = \ln(0.70) / \ln(0.98) = -0.35667 / -0.02020 =$ 17.65 net losing units—gifting you +2.65 extra survival units during drawdowns while compounding exponentially faster during winning streaks! (And conversely, Martingale doubling after losses collapses `Ueff to $\log_2(N)$, turning a routine 7-trade losing streak into instant 100%` wipeout!)

Exact Infinite-Horizon Risk of Ruin Matrix Across Edge & Unit Cushion ($U$)#

Inspect exact Risk of Ruin across asymmetric payoff ratios $R$, win rates $p$, and risk-unit buffers $U$ to your account's stop-out floor:

Quantitative Distribution · Expectancy / Risk
Expectancy / RiskCharacteristic Root ($r_0$)
01Zero-Edge Random (p=50%, R=…$0.0002Thin Scalper (p=53%, R=1:1)+$0.0603Solid Swing (p=55%, R=1.5:1)+$0.37504Breakout Trend (p=35%, R=3.…+$0.4005Tail Macro (p=22%, R=6.0:1)+$0.54
Strategy Profile (Win Rate p, Payoff R)Expectancy / RiskCharacteristic Root ($r_0$)U = 5 Units (reckless)U = 15 Units (moderate)U = 30 Units (institutional)
Zero-Edge Random (p=50%, R=1:1)$0.00$1.0000$$100.0\%$$100.0\%$$100.0\%$ (Guaranteed ruin!)
Thin Scalper (p=53%, R=1:1)+$0.06$0.8868$$54.8\%$$16.5\%$$2.72\%$
Solid Swing (p=55%, R=1.5:1)+$0.375$0.6186$$9.05\%$$0.074\%$$< 0.0001\%$
Breakout Trend (p=35%, R=3.0:1)+$0.40$0.7686$$26.8\%$$1.93\%$$0.037\%$
Tail Macro (p=22%, R=6.0:1)+$0.54$0.8542$$45.5\%$$9.41\%$$0.89\%$ (Needs U >= 30!)
Risk of Ruin percentage for a 22 percent win-rate 6-to-1 trend follower versus a 55 percent win-rate 1.5-to-1 swing trader at 10 risk units
Risk of Ruin percentage for a 22 percent win-rate 6-to-1 trend follower versus a 55 percent win-rate 1.5-to-1 swing trader at 10 risk unitsVerified Exhibit

Look at the Diffusion Bridge: $\exp(-2 \muu U / \sigmau^2)$ IS the Continuous Kelly Formula!#

Inspect Equation 2's Brownian diffusion exponent $-2 \muu U / \sigmau^2$: notice $\muu / \sigmau^2$ is literally *Merton-Kelly optimal leverage $f^$! Thus drawdown-constrained fractional Kelly, discrete binomial Kelly, and Markov Gambler's Ruin are three projections of the exact same stochastic barrier differential equation**!

Stress-Testing Correlated Losing Streaks via Monte Carlo#

IID Markov formulas assume trade $k+1$ has the same win probability $p$ after five straight losses; in real markets, chop regimes cluster losses (positive streak autocorrelation). Always stress-test non-IID cluster ruin using our Cholesky Monte Carlo Simulator, Ralph Vince Optimal f, and regulated client protection disclosures under the UK FCA.

Frequently asked questions

Why is infinite-horizon Risk of Ruin equal to 100% on a perfectly fair 50/50 zero-commission game?

Because 1D symmetric Brownian motion / simple random walk is neighborhood-recurrent! With zero upward drift (mu = 0) and an absorbing floor at $0 against an unbounded upside (+infinity), the random walk fluctuates ever wider as sqrt(t) and crosses ANY finite lower barrier with probability 1.00 as t -> infinity!

Can a Fixed-Fractional trader ever literally hit $0.00, and why can they still suffer "practical ruin"?

In continuous math, multiplying by (1 - 0.02)^k never reaches literal zero; in real markets, "practical ruin" hits via three floors: (1) Account equity drops below minimum exchange lot/margin size; (2) An allocator or prop firm pulls the plug at -15% or -25% drawdown; or (3) An overnight gap exceeds 1/Leverage!

Why does Martingale betting (doubling bet size after every loss) guarantee eventual blowup?

Winning 1 unit most days feels safe until a routine 1-in-512 nine-loss streak requires staking 2^9 - 1 = 511 units just to win +1 unit—instantly breaching either your total bankroll or the broker’s statutory Initial Margin cap and crystallizing catastrophic ruin! Martingale swaps frequent tiny gains for certain tail wipeout.

How do I calculate the maximum expected consecutive losing streak over N trades for win rate p?

With loss probability q = 1 - p across N independent trades, the expected longest losing run is approximately E[L_max] ~= ln(N * p) / ln(1 / q). For a trend follower with p = 30% (q = 0.70) over N = 500 trades, E[L_max] ~= ln(150) / 0.3567 = 14.0 consecutive losses—size U to survive 2.5x E[L_max]!

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). Risk of Ruin Math: Fixed-Dollar Markov Absorption vs Fixed-Fractional Sizing. QuantCalc Tools. https://quantcalctools.com/guides/risk-of-ruin-gamblers-ruin-markov-absorption-fixed-fractional-math
Explore Kelly Margin Leverage Sizing Hub (20) →

QuantCalc Tools Data, Benchmarks & Contributor Records

Risk of Ruin Formula: Markov Gambler Ruin vs Fractional Bet | QuantCalc Tools