Price Prediction Workbench
Compare historical bootstrap, block bootstrap, correlated Monte Carlo, and heavy-tail Student-t projections at the same time. Advance all four synthetic paths together, then back up one complete step when needed.
Prediction settings
The first ticker is the displayed prediction. Additional symbols are optional covariance context for the multivariate models.
Recommended defaults
Advance or restore the model paths
Each repeat is a separate undoable step. All four models advance before the next repeat begins.
All four median projections
This chart isolates the center path from each model so disagreement is easy to see.
Consensus projection bands
The pooled bands combine paths from all four models rather than treating one model as authoritative. The inner shaded band is the 25th–75th percentile range; the summary cards show its terminal values and the two inverse-e percentiles.
Stepped paths so far
These are the four independently evolving model anchors. Back one step restores all four together.
Historical bootstrap
Resamples individual observed return rows. Cross-asset relationships from each sampled row are preserved.
Block bootstrap
Resamples consecutive return blocks to preserve short-term volatility clustering and serial structure.
Correlated Monte Carlo
Draws jointly from the historical mean vector and covariance matrix. With one ticker, this reduces to Gaussian Monte Carlo.
Heavy-tail Student-t
Uses a multivariate Student-t process with five degrees of freedom to allow more frequent extreme moves.
What “prediction” means here
Each model estimates a conditional distribution from historical returns. The median is not a guaranteed target, and the displayed percentiles are not confidence intervals for the true future. They are quantiles of the simulated outcomes under the selected models and current settings. The consensus values pool terminal paths from all four models.