Position Sizing Research

The Kelly-Ahrens Algorithm

Half-Kelly is overcautious for insured defined-risk options. When your edge is statistically proven and your downside is structurally bounded by insurance, you deserve a smarter sizing formula.

1. The Problem with Standard Half-Kelly

The Kelly Criterion tells you what fraction of your bankroll to risk on a bet with a known edge. For options traders, the standard adaptation is Half-Kelly — use half the Kelly fraction as a conservative buffer against model error.

Kelly% = (win_rate × avg_win − loss_rate × avg_loss) / avg_win
Half-Kelly% = Kelly% / 2
Contracts = (Account × Half-Kelly%) / max_risk_per_contract

The problem is the denominator: max_risk_per_contract. For a defined-risk spread with the short strike at 7,300, the standard approach computes a theoretical crash scenario assuming the position is naked:

# Standard approach — assumes naked exposure
max_risk = short_strike × 100 × assumed_crash_pct
max_risk = 7300 × 100 × 0.05 = $36,500

# Even when capped (e.g. $10,000), this ignores the insurance leg

The result: Kelly divides by $10,000 and produces 1 contract on accounts under $80,000 — even when the strategy has a 63% win rate and a proven edge at 99% statistical confidence. The formula is punishing you for a disaster that cannot happen with an insured structure.

2. Insurance Changes Everything

A Jade Lizard with a long put insurance wing has a fundamentally different risk profile than a naked put. The insurance leg creates a hard structural floor on losses — it doesn't matter how far the market gaps down.

The Insurance Floor

A defined-risk spread with a long insurance wing has three distinct zones:

The Insurance Math

With a 30-point insurance spread and $7.40 credit received (illustrative example):

True max loss = (insurance_spread_width − credit) × 100
= (30 − 7.40) × 100
= $2,260 per contract

# This is the same whether SPX drops 2% or 20%
# The insurance leg captures all downside beyond the spread
Gap ScenarioMove vs EntryGross LossInsurance GainNet Loss
−2% (earnings / CPI surprise)−2%variesinsurance activates$2,260
−3% (macro shock)−3%variesinsurance activates$2,260
−5% (severe gap)−5%variesinsurance activates$2,260
−10% (black swan)−10%variesinsurance activates$2,260

Key insight: The insurance leg converts a theoretically unlimited loss into a precisely bounded one. Every gap scenario produces the same net loss. A position sizing algorithm that ignores this is leaving contracts — and profit — on the table.

3. The Kelly-Ahrens Algorithm

The Kelly-Ahrens Algorithm makes three targeted modifications to standard Half-Kelly:

# Step 1: Standard Kelly fraction
kelly_pct = (win_rate × avg_win − loss_rate × avg_loss) / avg_win

# Step 2: Scale by statistical confidence in edge (t-test)
if p_value < 0.001: confidence_scalar = 2.0 # 99.9% — approach full Kelly
elif p_value < 0.01: confidence_scalar = 1.5 # 99% — reduce uncertainty discount
elif p_value < 0.05: confidence_scalar = 1.0 # 95% — standard half-Kelly
else: confidence_scalar = 0.75 # unproven — extra conservative

scaled_kelly = min(kelly_pct, (kelly_pct / 2) × confidence_scalar)

# Step 3: EV anchor (when confidence ≥ 1.5×)
ev_per_contract = (win_rate × avg_win) − (loss_rate × avg_loss)
ev_fraction = ev_per_contract / avg_win
effective_pct = max(scaled_kelly, ev_fraction / 2) # if confidence ≥ 1.5×

# Step 4: Insurance-bounded max risk (not theoretical crash)
true_max_risk = (insurance_spread_width − credit) × 100

# Step 5: Contracts
contracts = floor(account × effective_pct / true_max_risk)

4. Three Components Explained

Component 1: Confidence Scalar

Standard Half-Kelly uses a fixed 50% discount on the Kelly fraction. This discount exists to protect against model error — what if your estimated edge is wrong? As you accumulate trades and statistical evidence, that uncertainty shrinks. The confidence scalar reduces the discount proportionally.

At 99% confidence (p < 0.01), you've proven your edge with enough data that the uncertainty discount drops from 50% to 33% — you use 1.5× half-Kelly, which equals 75% of full Kelly. Still conservative, but appropriately so.

Component 2: EV-Based Anchor

Expected Value per contract is a second sizing anchor. When confidence is high, both the Kelly fraction and the EV-normalized fraction are valid. Taking the higher of the two prevents the algorithm from being artificially conservative when your win/loss ratio is favorable but the Kelly fraction calculation produces a lower number.

Component 3: Structural Max Risk

This is the core insight. For any insured defined-risk position, the theoretical worst case has been replaced by a structural one. The formula uses the actual bounded loss — not a scenario that cannot happen given the position's architecture.

For uninsured strategies (naked puts, credit spreads without a long leg), use the traditional theoretical max risk. The Kelly-Ahrens structural adjustment only applies when a long insurance leg creates a hard floor.

Kelly-Ahrens Calculator

Enter your actual trade statistics to compute your optimal contract size.

Pure trading capital (exclude tax savings, emergency funds)
Minimum 10 for meaningful statistics
Absolute value (positive number)
Distance between put short and put long
Per contract, in option dollars (e.g. 7.40)
0.025 = 97.5% confidence. Use 0.05 if unknown.
Contracts
Effective %
EV / Trade
expected per contract

6. Illustrative Example

A defined-risk options strategy with the following characteristics after 33 trades and a 30-point insurance spread:

MethodMax Risk UsedEffective %Contracts at $16KContracts at $20K
Standard Half-Kelly (theoretical crash)$36,50024%11
Standard Half-Kelly (capped $10K)$10,00024%11
Kelly-Ahrens (structural max risk)$2,26024%12

The algorithm correctly identifies that 2 contracts becomes appropriate at $20,000 of pure trading capital — a threshold the standard approach would never reach without $80,000+ in the account.

As sample size grows and the p-value crosses 0.01, the confidence scalar rises to 1.5×, pulling the 2-contract threshold even lower. The algorithm self-adjusts as evidence accumulates.

7. Kelly vs Kelly-Ahrens

FeatureStandard Half-KellyKelly-Ahrens
Max risk sourceTheoretical crash scenarioInsurance-bounded structural floor
Confidence adjustmentFixed 50% discount alwaysScales from 37.5% to 50% by p-value
EV anchorNoneSecondary anchor when confidence ≥ 1.5×
Behavior when edge unprovenStandard half-KellyExtra conservative (0.75×)
Behavior when edge provenStandard half-Kelly (unchanged)Appropriately less conservative
Accounts for strategy structureNoYes — insured vs uninsured

When to use Kelly-Ahrens: Defined-risk options strategies with a long insurance leg (Jade Lizard, Iron Condor, Bull/Bear spreads with long wing), at least 30 historical trades, and a computable t-test p-value on your trade history.

When NOT to use it: Naked options (no insurance leg), fewer than 30 trades (bootstrap — use 1 contract), or when you cannot compute an insurance-bounded structural max loss.


From the same desk

MyIntelBrief — AI Competitive Intelligence

Monitor your competitors daily. Get AI-powered intelligence briefs delivered to your inbox automatically. Know what your competition is doing before your customers do.

Try Free →
No credit card required
💬