Predict & Profit
0 out of 5 starsBe the first to reviewSelf-hosted Python suite for Kalshi weather and inflation markets — two bots, a monitoring dashboard and one shared risk engine, installed and managed as a single system. Version 3.0 prices temperature contracts from NOAA's National Blend of Models, the calibrated station-level guidance published for exactly the stations Kalshi settles on, at 0.75 weight; GFS, AIGEFS, ECMWF IFS, AIFS and HRRR hold the remaining 0.25 as a disagreement check rather than as the model. Settlement stations come from Kalshi's own series metadata across 20 cities — Chicago settles on Midway, Houston on Hobby — and any market NBM does not cover is skipped, which rules out every same-day contract, because NBM coverage starts at forecast hour 24. The inflation half scores CPI, Core CPI and PCE against the Cleveland Fed's nowcast and an internal weighted one. Candidates then pass a four-signal gate — spread 30%, volume 20%, order-book imbalance 25%, model mispricing 25% — and the author says over 95% of scanned contracts are rejected. There is no public performance record: the trade log the earlier version published was retired with 3.0, and the site now publishes release and validation status instead of a P&L scoreboard.
Overview
Self-hosted Python suite for Kalshi weather and inflation markets — two bots, a monitoring dashboard and one shared risk engine, installed and managed as a single system. Version 3.0 prices temperature contracts from NOAA's National Blend of Models, the calibrated station-level guidance published for exactly the stations Kalshi settles on, at 0.75 weight; GFS, AIGEFS, ECMWF IFS, AIFS and HRRR hold the remaining 0.25 as a disagreement check rather than as the model. Settlement stations come from Kalshi's own series metadata across 20 cities — Chicago settles on Midway, Houston on Hobby — and any market NBM does not cover is skipped, which rules out every same-day contract, because NBM coverage starts at forecast hour 24. The inflation half scores CPI, Core CPI and PCE against the Cleveland Fed's nowcast and an internal weighted one. Candidates then pass a four-signal gate — spread 30%, volume 20%, order-book imbalance 25%, model mispricing 25% — and the author says over 95% of scanned contracts are rejected. There is no public performance record: the trade log the earlier version published was retired with 3.0, and the site now publishes release and validation status instead of a P&L scoreboard.
How it works
- Prices every temperature contract from NOAA's published percentiles for the station that contract settles on.
- Declines any market NBM cannot price, including every same-day contract.
- Scores spread, volume, order-book imbalance and model mispricing into one composite before an order goes out.
- Trades inflation when the nowcast stack sits far enough from where the strike is priced.
- Halts on ambiguity — unresolved exposure or a state mismatch stops trading instead of compounding.
Supported markets
- Kalshi weather · 20 US cities
- CPI / Core CPI / PCE
$129 one-time on Gumroad — name-a-fair-price with $129 as the floor — covering both bots, the dashboard and the full Python source, with six months of updates. No subscription, no licence server, no paid weather or economic-data APIs: you run it on your own machine or VPS with your own Kalshi account and API key. Ubuntu 24.04 with Python 3.12 or 3.13 and PostgreSQL 16 or 17; Windows works only under WSL2. The $9.99 ebook sold alongside it is the manual strategy, not the software.
Get Predict & Profit- NOAA National Blend of Models at 0.75 weight — mean, sigma and percentiles for the settlement station, interpolated with normal tails beyond P10/P90 and sigma × 1.15 for fat tails
- GFS, AIGEFS, ECMWF IFS, AIFS and HRRR at 0.25 combined, kept as a disagreement check; AIGEFS and HRRR are pulled as byte ranges out of GRIB2 files on S3
- 20 Kalshi cities, with the settlement station read from Kalshi's series metadata instead of guessed from the city name
- Inflation bot on CPI, Core CPI and PCE against the Cleveland Fed's nowcast and an internal weighted one, with FRED and BLS inputs when their free keys are configured
- One account-risk engine across both bots: cost cap per trade, per-city and per-series exposure limits, and a daily-loss kill switch
- Fills, fill cost, fees and settlements taken from the exchange's own order record — missing evidence fails closed instead of being assumed
- Visual dashboard for account, trades, positions, model output, risk and system health; it reports what the bots did and places nothing itself
- Guided installer, management CLI (configure, start, stop, status, doctor, upgrade, rollback), PostgreSQL state and migration and recovery tooling
- Author
- Steve Farmer
- Category
- Weather Markets
- Website
- predictandprofit.io
- Risk level
- Medium risk
- Added
- Jul 16, 2026
Link to this page from your site and the listing gets the Verified Partner mark and a followed link back to you.
Get the badgePredict & Profit reviews
No one has reviewed Predict & Profit here yet. If you have run it, your write-up is the one the next reader will see.
Be the first to review Predict & Profit. Tell the next reader what it traded for you, what it cost and whether you would run it again.
Reviews are written by readers, not by us or by the bot's author. Each one is read by a human before it is published, and we never accept payment for a review or for a rating.
Predict & Profit alternatives
PolyTraderBot
Six self-hosted Polymarket bots from one developer, sold separately at $10 a month each and delivered as full source with a browser dashboard, tests and a Docker setup for your own PC or VPS. CopyBot follows wallets you choose, or candidates it ranks from Polymarket's public leaderboard, with fixed, proportional or Kelly-fraction sizing, ten-plus individually switchable guards and exits that follow the leader out. Crypto5min and Crypto15min trade the 5- and 15-minute up/down windows on BTC, ETH, SOL, XRP, DOGE, BNB and HYPE from tree-based models retrained every 60 or 90 minutes. ARB scans the order books for binary YES and NO asks that sum to less than $1 after fees. WeatherBot prices temperature buckets from a weighted blend of Open-Meteo models, NWS, NOAA and Met.no forecasts and compares the result with the live asks. PolyLiq rests two-sided quotes in the markets that pay Polymarket's liquidity rewards, or takes over the resting orders you placed yourself. Every bot starts with real orders switched off, keeps the wallet key in a local env file and, from the current releases, updates itself from its own dashboard. The vendor advertises no returns; a free Polymarket terminal from the same developer is in beta on the same domain.
PolymarketAlpha
Autonomous AI agents that scan, analyze and execute on Polymarket across sports, crypto (5m, 15m, 1h) and weather markets — no manual intervention. Backed by a custom-built desk that also ships professional on-chain Solana bots (MEV, arbitrage, sniping and market making).
PolymTradeBot
Editor's ChoiceAutomated Python bot for Polymarket's 5- and 15-minute Bitcoin & Ethereum up/down markets. Trades live or in paper mode with configurable risk controls, self-hosted from full source.