CPX AGENCY / QUANTITATIVE RESEARCH / RELEASE V5

THE PREDICTION NEVER STOPS.

Automated Kalshi public-market observations every five minutes. Models predict next-interval YES quote midpoint, record their predictions before the future observation, and compare against an unchanging no-change baseline. No claims of winning odds or future settlement accuracy.

5-minute scheduled collection4 competing frozen modelsNo look-ahead labelsPaper execution paused

PROSPECTIVE FORECASTS—New V5 stream

VERIFIED OUTCOMES—Actual later observations

MISSING OUTCOMES—Never silently filled in

MODEL EVIDENCENO-GOMultiple days required

LAST PREDICTION—New data every ~5 minutes

Connecting to the live Neon forecasting ledger...

Head-to-head probability-price prediction

No-change benchmark MSE—

Momentum · 10% extrapolation—

Momentum · 25% extrapolation—

Reversion · 10% pullback—

Reversion · 25% pullback—

Eligible evaluated observations—

Distinct observation days—

Lower mean squared error is better. All candidates were defined before scoring. These are experimental market-price forecasts—not calibrated YES settlement probabilities, not profitability forecasts, and not proof of alpha.

Perpetual data quality & risk governor

Five-minute price-matching coverage—

Adjacent prior quote available—

Research-quality paired quotes—

Liquidity-screened quotes—

Minimum forward-tested outcomes500

Minimum independent time windows100

Minimum separate days3

Unpriced old paper entries—

Real-account authenticationNOT CONNECTED

Real trades / automatic fundingDISABLED

Legacy paper entries with no verified exit prices remain unpriced. The paper bankroll cap was $25, $5 per order, and $10 concurrent exposure. No fresh simulated entries will be released without independent validation of fills and settlement.

Coverage by Kalshi series

Asset series	Predictions	Scored	Research quality	Trade-screened	Mean spread

Latest prospective forecasts

Forecast time	Contract	No change	Momentum 10%	Observed 5m later

Verified YES/NO settlement outcomes

Verified finalized market results—

Independent markets in Brier analysis—

Distinct settlement days—

Observed public market-price Brier—

Naive 50% probability Brier—

Observed market log loss—

Verified OMEGA model edgeNOT ESTABLISHED

Brier = average squared error between a market-implied YES probability and its final YES (1) or NO (0) outcome. Lower is better. Each resolved contract counts once. The probability is the Kalshi market midpoint, not a calibrated OMEGA model prediction. Inputs were frozen at least 60 seconds before the market closed.

Settlement calibration by implied probability

YES probability band	Markets	Mean forecast	Actual YES rate
Waiting for verified outcomes...

These are empirical frequencies for a small, correlated sample. No market-price superiority or tradable statistical edge should be inferred without independently held-out observations across multiple days.

Completed Kalshi settlement ledger

Contract	Official result	Settlement time
Awaiting settlement data...

Paper-only diagnostics include conservative estimated fees and 1¢ slippage but do not verify order-book depth, actual fills or executable returns. Live orders remain blocked.