When the same event trades at 52% on one screen and 67% on another, something has to give. Arbitrage is the force that sells expensive YES and buys cheap YES until gaps shrink—turning individual greed into collective calibration.
This chapter builds the core logic used in later lessons on cross-market, cross-contract, and structural arbitrage.
What arbitrage is (and is not)
Arbitrage — locking profit (or zero-risk improvement) by combining positions whose payoffs offset, exploiting a pricing inconsistency.
Not arbitrage:
- Buying because you “like” the candidate (directional bet)
- Hedging macro exposure (risk reduction, may lose)
- Copying a guru on social media
Quasi-arb — same narrative, different resolution rules; looks like gap, eats you at settlement.
The law of one probability
For the same event under identical resolution, YES price plus NO price should approximate one minus frictions, and the same YES should trade at the same implied percentage everywhere. Violations create risk-free trees when capital moves freely.
Binary arb tree (same venue)
If YES ask = 0.40 and NO ask = 0.55, sum = 0.95 < 1.
Action: Buy both for $0.95; one pays $1 at settlement → $0.05 gross per pair (before fees).
Competition consumes gap; makers adjust quotes.
If YES bid = 0.60 and NO bid = 0.50, sum bids > 1—impossible for simultaneous sale of both sides to the market; indicates crossed book or stale data.
Two-venue arb (sketch)
Venue A: YES 0.55
Venue B: YES 0.63
If resolution equivalent:
- Buy YES on A, sell YES on B (or buy NO on B)
- Lock 8¢ per share if holds to settlement
Frictions that kill arb:
The cross-market chapter details platforms and capital rails.
Who plays the arb game
Adverse selection for slow trader: when you see obvious arb on UI, bots already cleared it.
Speed and the arb clock
Timeline after news:
- T+0s — Informed trade on fastest venue
- T+1–30s — Arb bots scan APIs
- T+minutes — Retail apps refresh
- T+hours — Polls and articles cite “markets”
Your edge is in T+1–30s if you have automation—not in reading yesterday’s gap.
Inventory and convergence trade
Arb need not finish before settlement. Convergence trade:
- Buy cheap YES, wait for prices to align, sell into strength
- Risk: gap widens if fundamental news hits one venue only
Game: you bet liquidity returns vs information split.
Structural role in mechanism design
Platforms want arbs:
- Prices align with “consensus” faster
- Manipulation costly
- Users trust displayed odds
Platforms also fear arbs:
- Regulatory view as unlicensed cross-border wagering
- Oracle risk if arbs stress settlement
Failure modes
False arb — Trump wins “national popular vote” vs “presidency” — different events.
Stale quote arb — One venue halted trading; screen shows old 40%.
Capital arb — Need USD on A, USDC on B; FX and bridge risk.
Fat finger — 8% gap real for 200ms; you cannot fill.
Expected value of arb hunting
Professional arb is negative-sum after tech spend but positive EV per trade if:
E[gaps captured] > infrastructure + compliance
Retail picking obvious gaps on phone app competes against firms with co-location and lawyers.
Relationship to market efficiency
Weak-form — prices reflect past trades.
Semi-strong — prices reflect public news.
Arb pushes semi-strong efficiency across venues when frictions low. Does not imply correct forecast—only consistent given public info and rules.
Ethical and legal note
Arb is not laundering. Insider trading then arb on other venue is still illegal. Regulatory arb (using venue A because B banned) is compliance strategy, not price gap.
When you see a ten-point gap, write the resolution sentence for both legs, list fees limits and currency, estimate size at touch for your capital, and ask why the gap exists—often capital cannot move. Only then call it arb.
What comes next
Next: Cross-market arbitrage across venues.