Probability Math for Profit on Prediction Markets: EV, Kelly & How HolyPoly Helps
Prediction markets like Polymarket price shares from $0 to $1, directly reflecting the crowd’s implied probability of an outcome (e.g. 65¢ means 65% chance). Profiting requires estimating a true probability higher (for YES) or lower (for NO) than the market price, then using math like expected value (EV) to identify edges and Kelly Criterion for sizing bets. This guide covers the core formulas, profitable strategies, and how tools like HolyPoly automate edge detection and execution on Polymarket.
Core Probability Math
Two ideas underpin systematic profit: (1) only bet when your edge is positive in expectation, and (2) size bets so you grow bankroll over time without blowing up. That’s EV and Kelly.
Expected value (EV) for a YES share
For a YES share that pays $1 if the event happens and $0 if not, the expected value of buying at cost c (market price + fees) when your estimated probability is p is:
Only buy if EV > 0—i.e. when your p is greater than the price you pay. Example: if you assess the true probability at 58% (p = 0.58) but YES shares cost 52¢ (plus e.g. $0.01 fees), then c = 0.53 and EV = 0.58 − 0.53 = $0.05 per share—a positive edge.
Kelly Criterion for bet size
The Kelly Criterion tells you what fraction of your bankroll to bet to maximize long-term growth. For binary outcomes with even-money style payouts (win $1 per $1 risked), the formula simplifies to:
Here p is your estimated probability of winning, q = 1 − p, and b is the odds (e.g. b = 1 for prediction markets where you risk c to win 1 − c). So if p = 0.55 and the market prices YES at 48¢, full Kelly says bet 2(0.55) − 1 = 10% of bankroll. In practice, many use fractional Kelly (e.g. 0.25× or 0.5×) to reduce volatility—so 2.5% or 5% of bankroll in that example.
Profitable Strategies
Once you have EV and sizing, you still need a source of edge. These are the main ways to get it.
Value betting
Build models (polls, Bayesian updates, fundamentals) that outperform market consensus by a few percentage points (e.g. 3–5%). When your implied p is higher than the market price for YES (or lower for NO), you have positive EV. Trade those discrepancies consistently and compound. Tools that surface which wallets actually have edge (win rate, PnL over many trades) help you copy or benchmark against proven value bettors instead of guessing—see how to evaluate Polymarket traders.
Arbitrage
Intra-market: In binary markets, YES + NO should sum to $1. If they sum to less (e.g. YES 50¢, NO 48¢ → 98¢), you can buy both sides and lock in roughly 2¢ profit per share at resolution. Cross-platform: Same event listed on Polymarket vs Kalshi (or others) can be mispriced; buy the cheap side on each platform for a risk-free or near-risk-free edge. HolyPoly’s arbitrage bot automates the intra-market case: it scans for YES+NO < $1 and executes both legs so you don’t have to monitor manually.
Exploit biases
Longshot bias: Crowds often overprice underdogs (high payoff, low probability). If your model says the underdog’s true probability is lower than the market price, you have positive EV on NO (or short the YES). Order flow: Short-term price moves can follow order flow imbalance (more buying than selling into one side). Tools that show Flow (net $ into YES vs NO over 24–48h) help you see where money is moving and combine that with leaderboard wallets for conviction.
Strategy summary
| Strategy | Math edge | Example profit |
|---|---|---|
| EV value bet | p > c | ~5¢/share on 52¢ buy when true p = 58% |
| Intra-market arb | Sum of prices < $1 | ~$0.02 per share on $0.98 total outlay |
| Kelly sizing | f* = 2p − 1 (full Kelly) | e.g. 10% bankroll at 55% edge; use ¼ Kelly for lower vol |
Using HolyPoly to Find and Execute Edges
HolyPoly scans Polymarket wallets for proven edges: 60%+ win rate, 50+ closed trades, positive PnL. It ranks them by PnL and volume in categories like crypto, politics, and sports, so you’re not guessing who to copy—you’re starting from wallets that have already compounded edge over time.
Pro ($199/mo) unlocks:
- Top Wallets leaderboard — filters by time window, category, PnL, edge score
- Recent trades — what top wallets did in the last 24h, with current price vs their entry
- Winning trades — only closed profitable trades with patterns (Longshot, Scalp, etc.) for strategy research
- Flow — 48h net $ into YES vs NO per market; combine with leaderboard to see if flow agrees with top wallets
- Picks of the day — flow-backed 15m Bitcoin pick and markets where top wallets agree
- Copy backtest — simulate copying a wallet’s closed trades over 1–3 months (fixed $ or % bankroll) for equity curve and max drawdown
- Playbooks — market, side, entry, size per wallet so you can place the same trades on Polymarket
- 24/7 arbitrage bot — connect your wallet; the bot auto-scans binary markets for YES+NO < $1 and executes both legs for locked-in profit at resolution, ideal for intra-market arb without manual monitoring
You can use HolyPoly as your edge layer and still apply your own probability math (EV, Kelly) for sizing and selection. For custom bots that integrate your models with Polymarket’s API, you can combine HolyPoly’s leaderboard and flow data with the Polymarket API and CLOB clients to place orders programmatically.
Next steps
To put this into practice: (1) read how prediction market prices relate to probability; (2) use the HolyPoly leaderboard to find wallets with a proven edge; (3) run copy backtests and size with fractional Kelly; (4) for arb, enable the HolyPoly bot or hunt YES+NO < $1 yourself. For a full workflow, see how to copy trade on Polymarket and how to use HolyPoly.