The Bot Ledger
EN

How to build a
prediction-market bot

Where to get data, how the exchange APIs work, where to host, what the fees cost. And which strategies held up in two months of live testing on Polymarket, Kalshi and two smaller exchanges.

How to connect your bot to an exchange

All guides

The three numbers that decide a bot

The fee peaks at 1.75 cents a share

Taker fee by price on Polymarket and Kalshi. Nothing at the extremes, most at 50 cents. Makers pay nothing.

What the fee costs

The book is wrong for 20 seconds

Buying the favourite loses at every second of a 5-minute window except the last 30, and needs a sub-second order path to trade.

The favourite in the last 15 seconds

Region cut the order path by 3.7x

The same bot on three setups. Only the dedicated instance in the exchange's region kept the edge.

Where to put the server

Tools

Where the exchanges are

Matching-engine locations as reported or inferred, not CDN edges. Drag to rotate, scroll to zoom, hover a place for the hosting suggestion.

What a trade must win

85c
Fee per share and breakeven win rate by exchange
ExchangeFee / shareWin rate to break even

Taker fees, hold to settlement. Limitless uses the published curve as an approximation.

Can you trade from where you are?

Each exchange blocks a different set of countries. This checks the country your connection reports and lists which exchanges will refuse an account from there. Nothing is stored.

How far are you from each exchange?

Times a few small requests from this browser to each exchange's public API. It measures your path to the API edge, so it tells you how a bot on your own machine would fare, not a server in the exchange's region.

Your round trip to each exchange

Round trip from this browser to each exchange API
ExchangeWarm (median of 4)First request

Checking where you are

  1. Reading your connection
  2. Country:
  3. Checking each exchange's terms

Articles

All articles
Polymarket Bot Rules: 5 Things to Do and 5 to AvoidTen rules for a Polymarket or Kalshi trading bot from two months of live testing: five that saved money and five that cost it.Most of the losses came from the same handful of mistakes: trusting derived labels, stopping out, chasing a rebate, sizing up on a paper edge. The five rules that held were cheaper than any of them, and each one is written here with the number that proved it.