Tradewink Predictions: Autonomous AI Prediction Market Trading

What Is Tradewink Predictions?

Tradewink Predictions is an autonomous AI system that trades binary event prediction markets on Kalshi, a CFTC-regulated prediction market exchange in the US. The AI estimates event probabilities, detects mispricings, sizes bets mathematically, and executes trades — all without manual intervention.

You set your risk limits. The AI handles everything else, running 12 concurrent agent loops 24/7.

What Are Prediction Markets?

Prediction markets let you trade on the outcome of real-world events. Each contract is binary: it pays out $1 if the event happens, $0 if it doesn't. If you buy a "Yes" contract at $0.60, you're betting 60 cents to win $1 (a 40-cent profit) if the event occurs.

The market price reflects the crowd's estimated probability. If a contract trades at $0.70, the market thinks there's roughly a 70% chance the event happens. When the AI estimates a different probability, that gap is the trading edge.

Market Categories

Tradewink Predictions trades across 10 categories on Kalshi:

Politics

Elections, legislation, Supreme Court rulings. The AI synthesizes polling data, prediction market history, expert forecasts, and news sentiment.

Economics

CPI prints, jobs reports, GDP growth. These markets have clear resolution dates and the AI can analyze leading indicators and economic models to estimate outcomes.

Fed Policy

Federal Reserve rate decisions and forward guidance. A dedicated category because Fed markets are among the highest-volume and most predictable using economic data.

Earnings

Corporate earnings surprises and revenue beats/misses. Ties directly into Tradewink's existing earnings analysis infrastructure.

Financials

S&P 500 closing levels, index milestones, and market-wide outcomes.

Companies

Individual company events — IPO outcomes, M&A completions, product launches.

Crypto

Bitcoin price thresholds, ETF approvals, regulatory actions. The AI monitors on-chain metrics, exchange flows, and regulatory filings.

Sports

Game outcomes, season milestones, and championship results.

Climate

Temperature records, hurricane landfalls, precipitation thresholds. Tradewink ingests weather forecast data from Open-Meteo and compares it to market pricing.

Culture

Awards shows, viral moments, and pop culture events. Lower-frequency but can offer large mispricings when the crowd overreacts to narratives.

Additional internal categories include Geopolitics, Mentions, and Tech & Science.

Trading Strategies

The AI uses a strategy router to select the best approach for each market. Here are the active strategies:

Ensemble Forecasting

Up to three AI models — Claude (primary), GPT-5 Nano, and Gemini 3.1 Flash-Lite — independently estimate the probability of each event. The system takes the geometric mean of their odds to produce a calibrated forecast. GPT and Gemini are optional and activate when their respective API keys are configured; Claude always runs.

Kelly Criterion Sizing

Once the AI estimates a probability and the market offers a different price, the Kelly criterion calculates the mathematically optimal bet size. Tradewink uses fractional Kelly (typically half-Kelly) for bankroll protection — it sacrifices some expected growth for significantly lower risk of ruin.

For example: if the AI estimates 70% probability and the market prices at 55%, half-Kelly might recommend risking 8% of bankroll. If the edge is smaller, the bet is proportionally smaller.

Value

Identifies markets where the AI's probability estimate diverges significantly from the current price. This is the core strategy — buy underpriced contracts, sell overpriced ones.

Momentum

Detects price trends in prediction markets. When a contract is moving in one direction with increasing volume, the momentum strategy rides the trend.

Contrarian

Takes the opposite side of extreme crowd sentiment. When a market is heavily skewed in one direction without strong fundamental justification, the contrarian strategy fades the crowd.

Event-Driven

Targets markets with upcoming catalysts (economic data releases, earnings, Fed meetings). The AI pre-positions before the event based on its probability estimate and adjusts immediately after.

Cross-Platform Arbitrage

The same event can be priced differently on Kalshi vs other platforms like Polymarket. When the AI detects a meaningful price divergence, it captures the spread. This is a lower-risk strategy since you're exploiting pricing inefficiency rather than predicting outcomes.

Longshot Sell

Sells contracts with very low probability events that the market is overpricing. These are high-win-rate, small-payout trades that profit from the crowd's tendency to overpay for unlikely outcomes.

Adaptive Calibration

The AI continuously tracks its own accuracy using Brier score decomposition (reliability + resolution + uncertainty). When it notices systematic bias — say, consistently overestimating political event probabilities — it applies adaptive Platt scaling to correct future estimates. The system literally learns to be less wrong over time.

Planned Strategies

Several additional strategies are implemented but not yet active in the agent pipeline: VPIN-based smart money detection, settlement arbitrage (trading the 5-10 minute repricing window after major events), Avellaneda-Stoikov market making, and risk parity allocation. These will be activated as the system matures.

The Predictions Dashboard

Your predictions dashboard shows:

• Performance stats — Total P&L, win rate, total bets placed, and open bet count at a glance
• Active positions — Current bets with entry price, size, strategy, side (yes/no), and live P&L. Filter by open or settled bets.
• Agent & risk status — Whether the prediction agent is running, current risk utilization, and any active risk limits being hit

Setting Your Risk Limits

Before the AI starts trading, configure your risk parameters:

• Bankroll — Total capital allocated to predictions
• Max single bet — Maximum amount on any single contract (default: $50)
• Max daily loss — Stop trading for the day if losses exceed this amount (default: $100)
• Max open positions — Limit how many bets can be active at once
• Max exposure — Maximum total exposure as a percentage of bankroll
• Kelly fraction — How aggressive the sizing should be (half-Kelly is the default; quarter-Kelly for conservative)
• Min edge threshold — Minimum probability gap required before the AI trades (default: 5%)
• Min confidence — Minimum AI confidence score to place a bet
• Preferred / excluded categories — Choose which market categories the AI trades or avoids
• Strategy toggles — Enable or disable specific strategies (contrarian, arbitrage, longshot sell, etc.)
• Paper trade mode — Enabled by default. Simulates trades without real money until you're ready.

How It Works End-to-End

1. The AI scans all active Kalshi markets every few minutes
2. For each market, the AI estimates the probability (using up to three models if configured)
3. The ensemble estimate is compared to the current market price
4. If the edge exceeds your minimum threshold, the strategy router selects the best approach and Kelly criterion calculates the optimal bet size
5. Risk checks confirm the trade fits within your limits (single bet max, daily loss, max exposure, max open positions)
6. The order is submitted to Kalshi
7. The AI monitors the position, tracking P&L and updating calibration scores
8. When the event resolves, the outcome feeds back into the calibration system

Why Kalshi?

Kalshi is a CFTC-regulated prediction market exchange in the US (one of the first to receive designation). This means:

• Legal — Fully regulated, available to US residents
• Settlement guarantee — The exchange guarantees payouts; no counterparty risk
• Transparent pricing — Real order book with visible bids and asks
• API access — Programmatic trading via REST API (which Tradewink uses)

Getting Started with Predictions

1. Create a Tradewink account at tradewink.com
2. Create a Kalshi account at kalshi.com and fund it
3. Connect Kalshi in your Tradewink dashboard settings by entering your Kalshi API credentials
4. Set your risk limits — Start conservative (quarter-Kelly, low max bet, strict daily loss limit)
5. Enable autopilot — The AI starts scanning and trading immediately
6. Monitor the dashboard — Review positions, P&L, and calibration daily for the first few weeks

Start with a small bankroll and conservative settings. As you build confidence in the system's calibration and edge, gradually increase your limits.