Straight answers.
Stradion is built for people who would rather check a claim than be asked to believe it. So here are the questions a skeptic should ask — answered plainly, including where the honest answer is "not yet."
What does "P(up)" actually mean?
A calibrated probability that the return is positive over the stated horizon — not a made-up confidence score. Calibration means that across all the times we printed 62%, about 62% should actually have risen. We publish 90% interval coverage so you can check whether that property holds in practice.
Why should I trust the track record?
You don't have to — you can audit it. The record is a walk-forward, out-of-sample replay: at each past date the model used only data known then, and every call is scored once its horizon elapsed. Nothing is dropped for looking bad.
Hit-rate and 90% coverage are published per model and per horizon, and paying members can export the full history — the calls, the inputs behind them, and the realized outcomes — to verify it independently.
Can I trade these signals?
Honestly, not yet at short horizons. Our own trading analysis shows the models' directional confidence does not rank outcomes well enough to trade on over a few days — it only begins to separate around the 20-day window. We show this on the performance page rather than imply an edge we cannot evidence. Treat the outputs as research, not a trading system.
Expected return versus the 90% band — what's the difference?
The expected return is the model's central estimate. The 90% band is its uncertainty: the range it expects to contain the realized return about 90% of the time. A wide band is the model telling you it is not sure. It is an uncertainty range, not a target range.
What is a "signal matrix," and why only eight symbols?
A matrix is a fixed set of symbols paired with the models that forecast them and the live record of how those forecasts scored. The public matrix is eight large-caps — kept small and free so the track record stays legible while it builds. Broader matrices, each with its own scored record, are planned.
Do you redistribute prices or market data?
No. Stradion publishes only our models' percentage outputs — direction, probabilities, expected returns, and return bands — and how they scored. No prices, no OHLCV, no redistributed vendor data.
How are the models chosen?
Two forecast models run side by side, and each symbol is served by whichever scored better for it out of sample (direction hit-rate, with band calibration as a guard). deep-xgb-cqr is an XGBoost quantile model with conformal bands and a calibrated direction; vol-baseline-v1 is a parsimonious volatility-based model with no learned features. The performance page breaks down each.
What's free, and what will cost money?
The public matrix and the daily news brief are free and anonymous, and stay that way while the record builds. Planned paid surfaces: signal-change alerts and watchlist routing, additional matrices, the full Oikos research terminal, and programmatic API access — plus full export of the published history for members.
Pricing is not set. The record has to earn it first.
Is this investment advice?
No. Stradion is research. It does not know your goals, capital, or constraints, and nothing here is a recommendation to buy, sell, or hold. Models degrade when regimes shift; backtests flatter; small samples mislead. Use it to form your own view.
Who built this, and why?
Stradion began as a private tool — the terminal its author wanted for his own market work, held to the standard he would have paid for. There was never an incentive to dress up the numbers for a buyer, so the honest version is simply the one that exists.