Probability
Conditional Probability Engine
Estimate P(Y | X) across historical market states.
Status research · Version 0.2 · Sep 2026
P(Xₜ₊₁ | Xₜ)
Illustrative distribution, not a live forecast.
Bull 34%
Neutral 41%
Bear 25%
Overview
Given a conditioning state X, the engine counts historical successors Y and reports the empirical conditional distribution with uncertainty.
Figures on this page are model output examples. They are estimated distributions, not a statement of where the market will trade.
How it works
- 01 Define X and Y
- 02 Match historical states
- 03 Estimate P(Y|X)
- 04 Correct for search
Inputs
| Condition | funding z > 2 |
| Outcome | 24h return < 0 |
Outputs
- P(Y|X)
- Support
- Interval
Model statistics
Published as quality of the estimate. Directional accuracy is not the headline for a probabilistic model.
- Output
- probability, not a point forecast
Validation
- Permutation test
- Multiple-hypothesis correction
Data sources
ArcDelta: trades, order book, open interest, funding, OHLCV.
Access
From Quant plan. Included runs, path limits, and API access follow the Labs tier on the TreVmS account. Individual metrics are not sold separately.
Experimental. This is an active research project. Interface, methodology, and outputs may change.