kelly-js

The sports bettor's math toolkit. Kelly Criterion, CLV, EV, bankroll stats, odds conversion — TypeScript, zero dependencies, tree-shakeable.

import { kelly, clv, bankrollStats } from '@ianalloway/kelly-js';

// Size a bet
const k = kelly(0.58, -110);
console.log(k.fraction);         // 0.0714
console.log(k.halfDollars(1000)) // $35.71

// Measure your edge
const c = clv(-108, -115);       // you bet -108, closed -115
console.log(c.verdict);          // 'positive' — you beat the market

// Bankroll health check
const stats = bankrollStats(myBets, 1000);
console.log(stats.roi);          // +0.054 (+5.4%)
console.log(stats.maxDrawdown);  // 0.12 (12% max drawdown)

Install

npm install @ianalloway/kelly-js
# or
pnpm add @ianalloway/kelly-js

API

Kelly Criterion

kelly(winProbability, americanOdds): KellyResult

Calculate optimal bet sizing using the Kelly Criterion.

const k = kelly(0.60, +120);  // 60% win prob at +120

k.fraction       // 0.1167  — bet 11.67% of bankroll
k.halfKelly      // 0.0583  — recommended for most bettors
k.quarterKelly   // 0.0292  — ultra-conservative
k.dollars(1000)  // $116.67 — full Kelly on $1,000 bankroll
k.halfDollars(1000) // $58.33
k.ev             // 0.12    — expected value per unit
k.edge           // 0.0545  — 5.45% edge over implied prob
k.hasEdge        // true

Rule: Use half-Kelly in practice. Full Kelly maximizes log-growth but variance is brutal.

kellyPortfolio(bets, maxExposure?): PortfolioResult[]

Size multiple simultaneous bets, scaling so total exposure stays within maxExposure (default 25%).

const portfolio = kellyPortfolio([
  { winProbability: 0.58, americanOdds: -110, label: 'Chiefs ML' },
  { winProbability: 0.62, americanOdds: +105, label: 'Warriors ATS' },
], 0.20);

Odds Conversion

impliedProb(american): number

impliedProb(-110) // 0.5238
impliedProb(+150) // 0.4000

toDecimal(american): number

toDecimal(-110)   // 1.909
toDecimal(+150)   // 2.500

toAmerican(decimal): number

toAmerican(1.909) // -110

convertOdds(american): OddsConversion

convertOdds(-110)
// { american: -110, decimal: 1.909, fractional: '10/11', impliedProbability: 0.5238 }

removeVig(side1, side2): { prob1, prob2, vig }

removeVig(-110, -110)
// { prob1: 0.5, prob2: 0.5, vig: 0.0476 }

Expected Value

expectedValue(winProbability, americanOdds, stake?)

expectedValue(0.58, -110, 100)
// { ev: 9.09, evPercent: 9.09, breakEvenProb: 0.5238 }

Closing Line Value (CLV)

CLV is the #1 predictor of long-term betting success. If you consistently get better odds than closing, your process is sound.

clv(openLine, closeLine): CLVResult

clv(-108, -115)
// { openLine: -108, closeLine: -115, clvPercent: 0.45, beatClose: true, verdict: 'positive' }

Verdict	Avg CLV	Meaning
elite	≥ +2.0%	World-class line shopping
positive	≥ +0.5%	Real edge in timing/shopping
neutral	≥ -0.5%	No significant edge either way
negative	< -0.5%	Consistently getting bad lines

clvSummary(bets): SummaryResult

clvSummary([
  { openLine: -108, closeLine: -115 },
  { openLine: +102, closeLine: +100 },
  { openLine: -120, closeLine: -118 },
]);
// { avgCLV: 0.58, beatCloseRate: 0.67, verdict: '✅ Positive — beating the market consistently', totalBets: 3 }

Bankroll Tracking

betPnL(stake, americanOdds, result): BetResult

betPnL(100, -110, 'win')  // { pnl: 90.91, roi: 0.9091 }
betPnL(100, -110, 'loss') // { pnl: -100, roi: -1 }

bankrollStats(bets, startingBankroll?): BankrollStats

bankrollStats(myBets, 1000)
// {
//   totalBets: 150, wins: 84, losses: 63, pushes: 3,
//   winRate: 0.5714, totalStaked: 15000, netPnL: 823.50,
//   roi: 0.0549, peakBankroll: 1901.23, maxDrawdown: 0.089,
//   currentStreak: 4, streakType: 'win'
// }

DFS Helpers

ownershipLeverage(projectedPoints, ownershipPct)

ownershipLeverage(52.4, 8.2)  // 5.98 — high leverage
ownershipLeverage(58.1, 38.5) // 1.48 — chalk

stackBonus(qbProj, receiverProj, correlation?)

stackBonus(33.8, 29.8) // 1.04 — extra projected pts from correlation

The Math

Kelly Formula: f* = (bp - q) / b

• b = net decimal odds (profit per unit)
• p = your estimated win probability
• q = 1 - p (loss probability)

CLV Formula: CLV% = (P_close - P_open) × 100

• Positive = you got better implied prob than close (beat the market)

EV Formula: EV = (b × p) - q

Author

Ian Alloway — Data Scientist, sports analytics & AI.

License

MIT