Validation and release status
Predict & Profit is active software. This page explains what the current release does, what has to pass before a release ships, what remains unproven, and why a daily public trade feed does not belong on the site.
Reviewed against the shipped 3.0.0 release material
Predict & Profit 3.0
- Current release
- 3.0.0
- Strategies
- Weather + Inflation
- Weather cities
- 20
- Platform
- Ubuntu 24.04
- Runtime
- Python 3.12 or 3.13
- State
- PostgreSQL 16 or 17
- Paid data APIs
- 0
- Subscription
- None
Experimental activity is not product performance
I use these bots as active engineering projects. I test forecast models, market filters, position sizing, execution rules, settlement handling and accounting changes. Those experiments can create losses.
The old page mixed experimental runs, retired models, paper trades, live trades, old accounting rules and current releases into one scoreboard. The numbers looked precise, but they did not answer the question a buyer needs answered: what does the current release do, and what has been validated?
The detailed records remain in private monitoring and reconciliation systems. I use them for debugging, model evaluation and risk review. This page reports release status, qualification evidence and validation limits instead.
What has to pass before 3.0 ships
3.0.0 is functionally identical to the release candidate that was qualified and then deployed to production; only release metadata and archive examples changed between them. The gates below describe what that qualification covers.
Fresh installation
A clean install is driven end to end against a disposable PostgreSQL cluster, synthetic signed provider transport and a real Chromium browser: setup, all four components in dry mode, and dashboard login, navigation and logout.
Migration from a supported earlier layout
Upgrades are rehearsed against realistic historical data. The tooling inspects the existing installation, validates credentials and order ownership without modifying the source, converts the schema exactly, and reconciles against the exchange before activation.
Exact archive acceptance
The gate runs against the built release ZIP itself rather than the working tree, so what is qualified is the artifact a customer actually receives.
Shared Weather and Inflation account risk
Both strategies are exercised against the same account under one risk engine, because that is where two-strategy accounting bugs appear.
Order identity handling
Accepted orders are bound to authenticated order evidence fetched by signed request. Unavailable evidence leaves a submission unresolved; an invalid or foreign user latches a conflict rather than being guessed at.
Fill and settlement reconciliation
Fills, actual fill cost and fees are taken from the exchange's own order record. Without that evidence the row fails closed for reconciliation and no fee is assumed.
Immutable release packaging
Each release declares its version, package format, supported Python versions, supported upgrade sources and rollback compatibility. A changed build gets a new version and the previous archive and its evidence are preserved.
Fail-closed safety behaviour
Unresolved exposure, incomplete account snapshots, identity conflicts and state mismatches halt trading. Installation does not start trading, and activation is a separate explicit step.
Passing a test suite is not the same as proving a trading edge. Do not read production live authorization into any of this.
What I publish
- Current release status and version contracts
- Forecast and nowcast methodology and data sources
- System architecture and qualification coverage
- Known limitations and unresolved questions
- Material bugs and accounting fixes
- API migrations and retired approaches
What stays private
- Daily account balance
- Individual experimental trades
- Open positions
- Paper-trading and dry-run activity
- Temporary research branches
- Intraday profit and loss
- Unreconciled settlement records
- Tests of new filters, models, or sizing rules
Current limits are part of the release status
I am not publishing a performance claim for the weather or the inflation strategy. A responsible conclusion needs a meaningful, version-specific sample with reconciled fees and settlements, and that is not something a launch page can honestly assert.
A cold 20-city weather scan can take longer than the configured scan interval when forecast providers respond slowly. You can reduce the city list or increase the interval.
Execution quality has not been measured well enough to support a conclusion about passive bids versus crossing the spread.
Native Windows lifecycle operation is not supported in 3.0. The Linux procedure is authoritative, and Windows users should run it under WSL2.
Prediction-market APIs, settlement rules and data sources can change. The package includes migration, diagnostic and recovery tooling, but no automated trading system is maintenance-free.
What you are buying
A source-code package for technically comfortable Kalshi users. You operate it with your own account and API key. There is no subscription, licence server, phoning home or paid market-data requirement.
- Weather Bot
- Inflation Bot
- Visual Dashboard
- Full Python source
- Installer and management CLI
- Shared account-level risk controls
- Order, fill, fee and settlement reconciliation
- Diagnostics
- Migration and recovery tooling
- Customer documentation
- No subscription
- No paid data APIs
Price
$129
One-time purchase. No subscription. Includes six months of updates.
Get Predict & Profit 3.0 - $129Predict & Profit is educational software for data science and research. It is not personalized financial, investment or trading advice. Prediction markets are speculative, and you can lose the full amount committed to a position.