How Prediction Market Arbitrage Works (Polymarket, Kalshi)

Polymarket and Kalshi are prediction markets. You trade on the outcome of real events — elections, economic data, whether it’ll snow in April. Every question has two contracts: YES and NO.

Each contract pays out $1 if it’s right, $0 if it’s wrong. You buy at whatever price the market sets.

Here’s where it gets interesting. In a perfectly efficient market, YES + NO always equals $1.00. But markets aren’t always efficient.

Take a question like “Will Bitcoin hit $100k by March?”

YES + NO = $0.97. That’s less than a dollar.

You buy both. One YES contract for $0.42 and one NO contract for $0.55. Total cost: $0.97.

Now think about what happens:

• Bitcoin hits $100k → YES pays $1.00, NO pays $0. You get $1.00.
• Bitcoin doesn’t hit $100k → YES pays $0, NO pays $1.00. You get $1.00.

Either way, you get $1.00 back. You paid $0.97. That’s $0.03 profit per pair, no matter what happens.

1
        YOU
2
         │
3
    pay $0.97
4
    ┌────┴────┐
5
    ▼         ▼
6
  YES       NO
7
 $0.42     $0.55
8
    │         │
9
    ▼         ▼
10
 ┌─────────────────┐
11
 │  ONE of these   │
12
 │  pays out $1.00 │
13
 └────────┬────────┘
14
          ▼
15
   You get $1.00
16
   Profit: $0.03

This is arbitrage. You’re not predicting anything. You’re exploiting a pricing gap.

Why the Gap Exists

In theory, YES + NO should always equal $1.00. In practice, they drift apart:

1
  Efficient market          After news breaks
2

3
  YES  $0.50                YES  $0.58  ← buyers rush in
4
  NO   $0.50                NO   $0.39  ← hasn't adjusted
5
  ──────────                ──────────
6
  Sum  $1.00                Sum  $0.97  ← gap opens

New information hits. A news story breaks and traders rush to buy YES. The YES price jumps, but NO hasn’t adjusted yet. For a brief moment, the sum drops below $1.00.

Low liquidity. If barely anyone is trading a market, prices get stale. One side updates while the other lags behind.

Different platforms. Polymarket might price YES at $0.45 while Kalshi prices NO on the same event at $0.52. Same event, different markets, different prices.

1
  Same event: "Will X happen?"
2

3
  Polymarket          Kalshi
4
  ┌──────────┐        ┌──────────┐
5
  │ YES $0.45│        │ YES $0.51│
6
  │ NO  $0.58│        │ NO  $0.52│
7
  └──────────┘        └──────────┘
8

9
  Buy YES here ───────────────────► Buy NO here
10
      $0.45          +                 $0.52
11
                                  ──────────
12
                          Total:     $0.97
13
                          Payout:    $1.00
14
                          Profit:    $0.03

Three cents doesn’t sound like much. But it scales linearly.

The return is small per unit, but it’s risk-free. The only question is how many pairs you can buy before the price moves.

The Catches

Fees. Polymarket only charges taker fees — if you place a limit order that sits on the book (maker), you pay zero fees. You only pay when you take someone else’s order. Kalshi charges a transaction fee on your expected earnings, and some markets also have maker fees on resting orders (common during elections, big sporting events, awards). A $0.03 gap can disappear after fees eat into it, so how you place orders matters. Maker orders on Polymarket keep the full spread.

Slippage. The price you see isn’t always the price you get. If you try to buy 1,000 YES contracts at $0.42, there might only be 200 available at that price. The rest fill at $0.43, $0.44, and now your gap is gone.

1
  Order book for YES:
2

3
  Price    Available
4
  $0.42    200 contracts  ← you want 1,000
5
  $0.43    300 contracts
6
  $0.44    500 contracts
7
           ───
8
           1,000 total
9

10
  Average price: $0.434, not $0.42
11
  Your gap just shrank from $0.03 to $0.016

Speed. These windows last seconds, not minutes. By the time you spot the gap and click buy, someone else already took it. This is why most prediction market arbitrage is done by bots — programs that watch prices across markets and execute trades in milliseconds.

Liquidity. Thin markets have bigger gaps but you can’t buy enough pairs to make meaningful money. Liquid markets have smaller gaps but can absorb larger orders.

The gap has to be large enough to survive fees, slippage, and execution time. In practice, that means most opportunities aren’t worth it for humans clicking buttons. But for automated systems watching hundreds of markets simultaneously, the small edges add up.