Model Catalog

Each model is a different explanation of how market behavior might be turned into a repeatable trading rule. Use the catalog to choose hypotheses for closer examination, not as a list of recommended trades.
Model definitions

Start with the market claim, then inspect its evidence

Filter the model universe by research round and operating constraints. Each card opens a standalone dossier; forecasts and runtime metrics remain secondary evidence rather than the model's identity.

Model comparison

Compare selected models across projected paths and a common starting amount

The comparison retains the current filters. Choose one model for the detailed projection and a broader set for the asset and outcome comparisons. Every model uses the same amount selected in Browse.

Asset comparison

outcomes for compared models

The asset lens asks whether the selected model set has evidence for this product. It compares model projected return, current recorded/runtime return, and the selected asset's raw Coinbase buy-and-hold move.

Asset market actual

buy-and-hold since baseline

This line is the selected asset's Coinbase close or recorded runtime price return. It is a benchmark, not a strategy result.

Selected-amount calculation

Projected and recorded outcomes at

Each row below holds the starting amount constant. Projected ending value is the input amount plus the calculated model change; it is not a time series and does not show an account growing from zero. The scale adjustment and minimum-trade drag are forecast assumptions that require order-level validation.

Prediction evidence

Inspect the evidence behind the model rankings

The evidence tables retain the selected comparison set and place the model-only baseline, AI-adjusted paper treatment, and historical entry-window results side by side.

Projection sourceEach model is estimated from rolling historical start points and walk-forward windows, then adjusted for fees, spread, slippage, drawdown, churn, and cash drag.
Actual sourceActuals begin on . Values through Day use recorded paper-run snapshots when the model is tracked by the executor.
Starting-amount assumptionThe calculator can apply the model projection to a selected amount, but that arithmetic is not evidence that a larger account trades more efficiently. The scale and minimum-trade adjustments remain assumptions until order-level results validate them.
Safety boundaryModel cards are public review artifacts only. They cannot execute Coinbase orders.