Is it a fit?Book a call

Cases

Every case in one format.

The task, the test design, the result and the decision. Losses next to wins, so cases can be compared and trusted.

The Figure cast as light on a mineral green sphere
Eigna TLM, Case 001
Release record

  • Candidate 3

    23 October, shrinkage head

    Held

  • Candidate 2

    14 October, calibrated

    Held

  • Candidate 1

    14 October, LoRA, 4 runs

    Held

  • Release 1

    Not issued

    No release
Held at the gateCase8 min

Case 001: held at the gate

A forecasting model of its own. Measured in the open, held at the gate. Our first client, a forecasting model for trading; it shares founders with Eigna, and we say so up front.

  • 17.8%Lower forecast error than its base
  • 0 of 3Simple baselines beaten
  • 57.3%Band coverage
  • +4.4%Paper trading
Read Case 001

The format

The rules for an honest case

  1. 01Losses next to wins, on the same page, at the same size.
  2. 02Every number with a date and a source: the run date, the read date and where to check.
  3. 03Compare with the simplest baseline first, named before training.
  4. 04The decision is made at the gate, by rules written before the test. A tie keeps the current version.
  5. 05Plain words. Say what was measured, on what, against what.
  6. 06No invented curves. Draw only measured values.
  7. 07Disclose relationships: shared founders, fees, data rights.
  8. 08All money in US dollars.

Test Lab

Case 002 comes from a Test Lab. It could be yours.

Four weeks, one task, your own data in a sealed setup. If the expert does not beat your baseline, you keep the report.

Request a Test LabBook a call