Kalshi Edge Engine
A machine that mostly says no.
Every ten minutes this engine scans hundreds of prediction markets, builds its own forecast from independent evidence, and compares its probability to the market's price. Almost every time, it declines to trade. That discipline is the product.
Illustration: the engine only acts when its probability and the market's price disagree by more than fees, slippage, and a margin for being wrong — here, an 84¢ forecast against a 52¢ price.
Method
How it decides
- Scan. Hundreds of open markets are persisted every cycle — order books, spreads, and settlement rules included.
- Forecast independently. Evidence comes from sources outside the exchange — weather models, index feeds — never from the market's own price.
- Charge itself costs first. Exchange fees, slippage, and an uncertainty haircut are subtracted before any edge counts. Small edges die here, on purpose.
- Record before the outcome. Every decision — including every pass — is frozen with a hash before the market resolves. Nothing is graded on hindsight.
Structure, not promises
Safety is built in
The runtime accepts demo credentials only. Production endpoints and live-trading flags are rejected at startup — there is no configuration that makes this software trade real money. Order submission sits behind four separate gates, position sizes are capped, correlated exposure is limited, and a drawdown kill switch halts everything automatically.
The dashboard shows empty states when there is no data. It never invents a number to look busy.
Read this part
What this is not
- Not financial advice. This is research software for studying prediction markets with simulated and demo money.
- Not a promised return. Any performance target you see in the configuration is a measurement benchmark, not an expectation. Nothing meaningful can be claimed until many resolved trades are on the record.
- Not connected to real money. There is no production mode to enable, and the interface exposes no trading controls.
If you take one idea from this project, take the habit: write the forecast down before the outcome, charge yourself the costs, and let “no trade” be a respectable answer.