Position Sizing for Binary Contracts: My Kelly-Based System

Position sizing for binary contracts determines whether a sound probability edge on Kalshi or Polymarket actually compounds into account growth, or bleeds out through variance and oversized bets. Most traders spend all their energy on the forecast — is this market mispriced, is the "yes" side underpriced — and almost none on how much to stake once they've found the edge. That's backwards. A mediocre edge sized correctly outperforms a great edge sized carelessly over any meaningful sample. This guide walks through a Kelly-based sizing system built specifically for binary, cash-settled contracts, where payout structure and correlation across markets create sizing traps that don't exist in traditional asset classes.

Kelly Criterion Binary Markets: The Math That Actually Applies

The Kelly criterion was built for exactly this kind of bet: a binary outcome with a known (or estimated) probability and a market-implied price. For a "yes" contract trading at price p_market that you believe resolves "yes" with true probability p_true, the Kelly fraction of bankroll to allocate is:

f* = (p_true × b - (1 - p_true)) / b, where b is the net odds received (payout minus stake, divided by stake).

On a binary contract priced at $0.40 for a "yes" share that pays $1.00, b = (1 - 0.40) / 0.40 = 1.5. If your structured analysis puts true probability at 0.55, Kelly says:

f* = (0.55 × 1.5 - 0.45) / 1.5 = (0.825 - 0.45) / 1.5 ≈ 0.25, or 25% of bankroll.

That number should stop you cold. Full Kelly on a single binary market is aggressive to the point of recklessness for anyone without institutional-grade probability estimates. The formula assumes your probability estimate is exactly right, that outcomes are independent across bets, and that you can size fractionally without execution friction. None of those hold cleanly in prediction markets. This is why every serious binary-contract trader runs a fractional Kelly system, not full Kelly — more on the exact fraction below.

The other subtlety specific to binary contracts: your edge isn't just "will this happen" — it's the gap between your calibrated probability and the market's implied probability, read directly off the order book. If you haven't internalized how that translates price to probability, revisit How to Read Prediction Market Odds before running any of this math, because a sizing model built on a misread price is worse than no model at all.

Position Sizing Binary Contracts: Why Bankroll Percentage Beats Fixed Stakes

A huge share of retail activity on Kalshi and Polymarket uses fixed-dollar staking — $50 on every trade, regardless of edge size or bankroll. This fails for two structural reasons unique to binary markets.

First, binary contracts have a hard payout ceiling: $1.00 per share minus your entry price. There's no scaling upside the way there is in a directional futures trade, so your edge is capped and known at entry. That means the size of your edge (the spread between true and market probability) should directly scale your stake — a 3-point edge and a 15-point edge do not deserve the same dollar allocation, yet fixed staking treats them identically.

Second, bankroll percentage sizing self-corrects for drawdowns. If your bankroll shrinks 20% after a run of losing resolutions, a percentage-based system automatically reduces stake size on the next trade, protecting you from the classic "chase the loss with the same size" spiral. Fixed-dollar sizing has no such governor.

The practical system: convert your Kelly fraction into a percentage of current bankroll, recalculated before every trade — not before every trading session, before every individual trade. Bankroll changes with every settlement, and your sizing should track it in real time.

Prediction Market Bet Sizing: Building In a Fractional Kelly Discount

Structured practitioners essentially never run full Kelly on binary contracts. The standard adjustment is quarter-Kelly to half-Kelly, and here's the reasoning specific to this asset class:

• Probability estimation error compounds fast. Full Kelly assumes p_true is exactly correct. If your real edge is half what you estimated, full Kelly sizing produces bankroll volatility that can wipe out a third of your account on a bad month even with a genuinely positive expected value.
• Binary markets cluster by event. Multiple Kalshi contracts on the same Fed meeting, same election, or same game are correlated — they aren't independent bets even when they look like separate line items in your portfolio. Full Kelly on each individually understates your true aggregate exposure.
• Liquidity and slippage eat into realized edge. Thin order books on smaller Polymarket and Kalshi markets mean your effective entry price is often worse than the quoted mid, silently reducing b in the Kelly formula.

A workable framework: calculate full Kelly, then apply a 0.25x to 0.5x multiplier depending on your confidence in the probability estimate. Higher-confidence markets — ones with hard data inputs like CPI prints or election vote counts — can run closer to half-Kelly. Soft, judgment-heavy markets (geopolitical event contracts, discretionary sports props) should run at quarter-Kelly or less. This single adjustment does more to protect long-run capital than any other rule in this system.

How PillarLab AI Fits Into This

The entire sizing exercise above is only as good as your probability estimate — the "p_true" that feeds the Kelly formula. Guessing at that number, or eyeballing it from vibes and headlines, defeats the purpose of running a quantitative sizing system at all. This is the actual gap PillarLab AI is built to close.

PillarLab AI runs a structured 9-pillar analysis on any Kalshi or Polymarket market, pulling real-time data directly from both platforms' APIs rather than relying on stale screenshots or manual research. Instead of a single opaque probability guess, you get a pillar-by-pillar breakdown covering things like market structure, liquidity depth, sentiment signals, historical base rates, and catalyst timing — each contributing to a final probability assessment you can actually defend and plug into a Kelly calculation.

That matters directly for sizing. A probability estimate backed by nine distinct analytical angles carries meaningfully less estimation error than a single gut call, which means you can run a slightly higher fractional Kelly multiplier with more confidence — and conversely, when the pillars disagree with each other, that internal disagreement is itself a signal to size down regardless of what the headline probability says.

The actionable output format is built for exactly this workflow: a probability estimate, a confidence read, and the underlying reasoning, all in one pass, so you're not stitching together data from five browser tabs before you can even open your position-sizing spreadsheet. For traders running any disciplined Kelly-based system across dozens of markets a week, that speed compounds the same way correct sizing does — more accurate inputs, more consistently, across a larger sample of trades.

Kalshi Trading Strategy: Sizing Rules for Correlated Markets

Correlation is the sizing trap most binary-contract traders miss entirely. If you hold "yes" positions on three different Fed-rate-decision-adjacent Kalshi contracts, they are not three independent bets — they're one macro bet expressed three ways. Sizing each at quarter-Kelly independently can leave you with an aggregate exposure equivalent to three-quarters of bankroll on a single underlying event, which defeats the entire purpose of fractional sizing.

The fix: group markets by underlying driver before sizing, not after. Build a simple tagging system — election-outcome cluster, Fed-decision cluster, single-game sports cluster — and cap total bankroll allocation per cluster, independent of how many individual contracts sit inside it. A reasonable starting cap is 15-20% of bankroll per correlated cluster, regardless of how attractive the individual edges look.

This is also where a documented, repeatable strategy earns its keep over ad hoc trading. If you haven't formalized your overall approach yet, pair this sizing framework with a broader read on Kalshi Trading Strategy 2026 so your entry criteria and your sizing rules are built on the same underlying logic instead of pulling in different directions.

Prediction Markets vs Sportsbooks: Why Binary Sizing Differs From Point-Spread Betting

Traders migrating from sportsbook betting to Kalshi or Polymarket often import point-spread sizing habits that don't map cleanly onto binary contracts. Sportsbook units are typically sized as a flat percentage of bankroll regardless of the odds, because vig is baked uniformly into the line. Binary contract pricing is different — the "vig" (the gap between the sum of yes/no prices and $1.00) varies market to market and directly changes your effective b in the Kelly formula. A market priced at 52/51 (yes/no) has a tight 3-cent total spread and a much smaller sizing penalty than one priced at 48/58, where the wider gap eats into edge before you've even accounted for your probability estimate. Check the combined yes/no price before sizing, not just the yes price in isolation — a wide combined spread should mechanically shrink your position size even if your directional read is strong.

If you're still getting oriented on how these mechanics differ structurally from sportsbook markets, Prediction Markets vs Sportsbooks covers the settlement and pricing differences in more depth, and it's worth internalizing before you assume your old sportsbook bankroll rules transfer over cleanly — because they don't, particularly around effective vig and liquidity.

Position Sizing Across Kalshi and Polymarket: Platform-Specific Adjustments

Liquidity depth and fee structure differ enough between Kalshi and Polymarket that a single sizing model applied identically to both platforms will misfire on one side. Kalshi's regulated, CFTC-overseen order books tend to have tighter spreads on high-volume markets but thinner depth on niche or newly listed contracts. Polymarket's on-chain execution introduces gas-cost and slippage considerations that don't exist on Kalshi, particularly on smaller-cap markets.

Practical adjustment: treat quoted liquidity depth as a direct input into your Kelly fraction, not just a pre-trade checklist item. If the order book can't absorb your calculated Kelly stake within a reasonable slippage tolerance — say, half a cent of price impact — cut the position down to what the book can actually support rather than forcing the full calculated size across multiple price levels.

This is also where platform selection itself becomes a sizing decision. If you're deciding where to route a given trade in the first place, Kalshi vs Polymarket 2026 breaks down the structural differences in depth, fees, and settlement that should inform not just where you trade but how large a position each platform can realistically support.

Frequently Asked Questions

What percentage of bankroll should I risk per binary contract trade?

Most disciplined traders cap individual positions at quarter- to half-Kelly, which typically lands between 2-10% of bankroll depending on edge size and confidence in the probability estimate.

Is full Kelly ever appropriate for prediction market trading?

Rarely. Full Kelly assumes a perfectly accurate probability estimate and independent bets, neither of which holds reliably across correlated binary markets like Kalshi or Polymarket contracts.

How do I size positions across correlated markets on the same event?

Group markets by underlying driver and cap total exposure per cluster — around 15-20% of bankroll — rather than sizing each contract independently at full Kelly.

Does PillarLab AI calculate position size directly?

PillarLab AI's 9-pillar analysis produces the probability and confidence inputs your Kelly calculation needs; you apply your own fractional multiplier and bankroll rules on top of that output.

How often should I recalculate my bankroll for sizing purposes?

Before every trade, not every session. Bankroll shifts with each settlement, and stale bankroll figures produce systematically oversized or undersized positions.

A Kelly-based sizing system only works if the probability feeding it is defensible, and platform noise, thin order books, and correlated event clusters all conspire to make that estimate harder to trust than it looks. Building the habit of checking pillar-level data before every position — not just before every session — is what separates traders who compound a real edge from those who get run over by variance on a mathematically "correct" bet. Start free with 10 credits