Mithrandir Metrics
Fangorn Selective interaction lane Refreshed Jun 1 06:00 AM ET
Nonlinear / Interaction Discovery / Selective · Selective interaction lane

Fangorn Model

Nonlinear interaction discovery. Allowed to abstain - it will not have an opinion on every row.

Markets
6
Active
4
Date
latest
Method Random Forest, ExtraTrees, GBM, or explicit segmented role-stack methods used for interaction discovery. Status Selective interaction lane Coverage 4 of 6 markets currently have a Fangorn lane.
Definition

What this lane is

01
Method family
Fangorn looks for cases where the mainline prior is too smooth: pitch mix by batter weakness, park shape by contact profile, lineup exposure by pitcher leash, and other interaction-heavy pockets.
fangorn
02
Not this
Not a stricter threshold on Palantir, not a trust label over the same probability, and not a pure calibration repair layer.
guardrail
Distinctness

Falsifiable test

What nonlinear or segmented interaction does Palantir miss? Fangorn should create different rankings because of tree splits, role pockets, or feature interactions.

Comparison first

The value of the three-lane system is comparison. No single lane is the right one in isolation.

Coverage matrix

Fangorn across six betting families

MarketStatusMethodDate
Home Runs Production Random Forest / RF-safe stack for nonlinear HR interaction pockets. 2026-06-01 Open
Strikeouts Missing Role-stack prototype exists; planned RF/segmented model over pitcher-start contexts. 2026-06-01 Open
Total Bases Experimental Selective TB strategy / under-specialty artifacts exist; planned contact-shape and park interaction model. 2026-06-01 Open
Hits Experimental Missing because current confidence artifacts are filters, not a distinct nonlinear engine. 2026-06-01 Open
Pitcher Outs Experimental Missing; planned leash-volatility RF/segmentation model. 2026-06-01 Open
Futures Missing Missing; planned scenario-driven season simulator / stress lane. -- Open