AoA
Join
← Intel
Edge vs illusion · 5 min read

The real edges in prediction markets — and the illusions

Forecasting is hard and fees eat thin edges. The durable edge is mostly not losing.

Real & repeatable (for a disciplined trader)

  • Fee minimization — the biggest, most certain 'edge', and it's just discipline: be a maker (save 75%), trade extremes not 50/50, hold to settlement instead of round-tripping.
  • Fading genuine overreaction — retail prices one event (an early INT) as a quality shift; a fair-value anchor knows it's one draw from a distribution. Real, but thin, and only when the dislocation clears fees.
  • Disciplined sizing + abstention — fee-adjusted fractional Kelly. Most traders blow up not because they're wrong but because they bet too much when they're right.

Rare / competed-away

Cross-venue arbitrage (Kalshi YES + Polymarket NO < $1) is real but friction-gated — the capital-transfer friction is the spread. It needs ~1.75–2.5¢ gross to clear fees and is hunted by bots scanning 10,000+ markets. You're racing infrastructure.

Illusion / already priced (negative-EV)

  • Linear momentum / 'ride the wave' — the line prices momentum instantly; you buy the local top and pay the fee.
  • Pre-game closing-line arb — the pre-game market is liquid and efficient; fees kill it.
  • Parlays / combos — Kalshi retail has lost $100M+ on parlays this year. The combo is the house's friend.

Before you find an edge, stop bleeding the rake. The single most valuable lesson saves you more than most 'edges' earn you.

Source · Adapted from our Prediction-Market Edges & GTO research

Informational decision-math education, not betting advice. 18+. SIM-safe · read-only odds. We show our calibration and our losers.