Forecasting & calibration
Can a system state a probability that still holds up once the world answers?
ARCANE LABS / RESEARCH OBSERVATORY
Forecasting · Structure · Evidence · Risk
Research that says what would prove it wrong.
Labs is the research arm behind ARCANE Intel. We test whether machines can read markets honestly: registered before the data arrives, scored in public, and published when the answer is no.
- Publications
- 5
- Negative or null
- 2
- In preparation
- 2
Latest · Evaluation · 17 Sep 2026
Models without evidence lose to markets
Given no fresh evidence, a calibrated multi-model ensemble scored a Brier of 0.284 against 0.212 for the market price at the same moment. The only statistically clear improvement came from walk-forward calibration, which cut Brier by 0.030. This is the floor every later ARCANE forecaster has to beat, published before we try to beat it.
Read the evaluation →Programmes
Four questions the lab keeps asking.
Structure & learning
Finding a pattern is cheap. The test is whether a learner can refuse one.
Evidence systems
Every claim should carry where it came from and what would break it.
Risk & mandates
Measure the loss a mandate actually governs, not a convenient proxy for it.
Research record
Public questions. Visible boundaries.
How Labs publishes
Five rules every piece keeps.
- 01
Registered first
The question, test and refusal conditions are written down before the data can answer.
- 02
Every number has an interval
A score comes with its sample size and a confidence interval, or it is not reported.
- 03
Negative results ship
A null or a loss is published in the same format as a win, including our own.
- 04
What would change it
Every piece names the observation that would overturn it, the same discipline as every ARCANE article.
- 05
A stated boundary
Every piece says what it publishes and what it withholds, so the gap is visible rather than hidden.
In preparation
Registered, running, not yet answered.
- Evaluation
Alpha against frontier models: a pre-registered head-to-head
Registered 17 September 2026. Remaining model arms and human judging in progress. Published whatever it shows.
- Evaluation
The ARCANE forecasting benchmark: first prospective window
Planned. Held-out enrollment 25 November to 25 December 2026; review no earlier than 25 January 2027.
- Programme
Order is not a forecast
A reading programme on order, measure and complex systems that generates testable market questions without turning history into a signal.
Every public conclusion remains open to its evidence and its limits.