← All work
  • Data
  • Automation
  • Engineering

Prediction Market Trading, Automated

Prediction markets move fast, and mispriced contracts disappear before anyone can spot them by hand. We built Janus, a system that compares Polymarket prices against sportsbook odds and trades the gaps, with risk management built in.

Client
Janus
Year
2026
Led by
Sumner
Status
v1 launched
Prediction Market Trading, Automated for Janus

The problem

Prediction markets move quickly. Prices shift with every injury report, line-up change and late swing in a match, and the chances to buy a contract below its fair value last minutes, not hours. Comparing prices across markets and sportsbooks by hand means you’re always a step behind.

What we built

We researched trading strategies, then designed and built an automated system that does the watching and the maths:

  • Fair value from the sportsbooks. Janus pulls odds from sportsbooks across NBA and professional tennis markets and turns them into a fair price for each Polymarket contract.
  • Strategies that act on the gap. When a contract trades meaningfully below fair value, Janus can take the position automatically, following the rules of each strategy.
  • Risk management built in. Position sizes, exposure limits and every open outcome stay visible, so nothing runs unchecked.
The Janus positions screen listing open NBA positions, with a callout comparing a market price of 27 cents to a sportsbook fair value of 33 cents
Every open position, with its entry price, snapshot price and edge against the sportsbooks. Portfolio figures are illustrative.

Testing before trusting

Before putting capital to work, every strategy runs through backtesting: the same window and the same starting capital, compared side by side against a baseline. It shows which ideas earn their place and which should be switched off.

The Janus simulations screen comparing a current strategy against a baseline, with results by strategy and by sport
Simulations compare strategies against a baseline. Figures shown are illustrative, not measured performance.

Where it is now

We launched v1. Early historical simulations show promising results and are helping validate the approach before it scales up.

Why it matters for your business

Lots of businesses have someone checking prices, stock levels or competitor listings every morning. The same approach turns that daily chore into a system that watches continuously and only asks for attention when something is worth acting on.

Want something like this? AI for Business

Less busywork. Smarter business.

→

Next case study

A Web3 Dashboard for Smarter Yield

→