Question 1

What is the latency of the API feed?

Accepted Answer

The public-facing delivery layer is optimized for low-latency updates. Production enterprise endpoints are designed to distribute updated probability matrices within roughly 200ms of a confirmed state change, subject to upstream feed latency.

Question 2

Do you provide tick data for backtesting?

Accepted Answer

Historical exports containing model probabilities, archived market prices, and settlement states are available for enterprise evaluation on request. Delivery format depends on the retention window and the requested sport coverage.

Question 3

How often are the CatBoost models retrained?

Accepted Answer

Retraining follows a walk-forward temporal validation process rather than a fixed marketing schedule. Model weights are refreshed only after new out-of-sample data is ingested and validated against probability-quality controls.

Question 4

What is included in the public archive?

Accepted Answer

The public archive contains processed events, confirmed archived prices when available, settlement states, and position deltas. It is a transparency layer rather than a full internal research warehouse.

Question 5

Why can an anomaly card appear without an implied price?

Accepted Answer

Some legacy publications do not have a confirmed archived market snapshot. The interface intentionally leaves the field empty rather than synthesizing a historical price that was never captured.

Question 6

Which sports are currently exposed publicly?

Accepted Answer

The public layer currently exposes football, NHL, and basketball. Tennis remains local until its sport-specific pipeline passes holdout and live-audit review.

Infrastructure FAQ

Methodology

Contact

Need a deeper technical discussion?