Market anomaly education boundary

Overview

Barfinex market anomaly content on the public website is educational. It can explain anomaly categories, delayed aggregate counts, source-freshness concepts, evidence classes, and review questions. It does not publish a live anomaly feed and does not expose private Provider candidates, focus plans, labels, runtime payloads, account data, order state, position state, or operator controls.

The private anomaly scanner, candidate triage, scorecards, focus-plan review, and any runtime promotion workflow remain in Provider and Studio.

Public-safe delayed aggregates

If public anomaly summaries are published, they must be delayed and aggregate. The first approved shape is:

• anomaly category;
• market type or aggregate market family;
• delayed count;
• severity buckets;
• evidence class names such as delayed_aggregate or historical_markout;
• redacted blocker and warning names from an allowlist;
• generation time for the aggregate, not a real-time alert timestamp.

The public projection must use at least a 60-minute delay. It must not include symbol-level candidates, Provider candidate IDs, source references, private URLs, raw payloads, execution instructions, or performance claims.

Educational categories

Public pages may describe these anomaly families at a conceptual level:

• relative-volume momentum breakouts;
• liquidity vacuum reversals;
• volatility regime shifts;
• cross-venue dislocations;
• funding and open-interest crowding;
• listing metadata events;
• spread and liquidity stress;
• market-wide residual shocks;
• toxic-flow standdown conditions;
• macro-event volatility.

These categories are research labels. They are not signals, instructions, or proof that a strategy is profitable.

Mentions of cross-venue dislocation, funding/open-interest crowding, or future model-based anomaly review are category education only. They do not claim that Bybit, OKX, Kraken, Coinbase, MOEX, equities, cross-venue basis, runtime-proof, symbol-map, model-score, or ML/deep anomaly support is publicly active.

Private fields

Public anomaly pages must exclude:

• Provider tokens, localhost URLs, credentials, and operator commands;
• accounts, balances, orders, fills, positions, and margin state;
• private candidate IDs, focus-plan IDs, review-label records, idempotency keys, source refs, evidence refs, and raw Provider payloads;
• unredacted score components, replay packets, Inspector decisions, Auditor proof, PnL, win rate, fill quality, connector runtime proof, model scores, ML inventory, and promotion evidence;
• any wording that asks the reader to buy, sell, enter, exit, or change risk.

How to read public anomaly content

Treat public anomaly content as market-structure education. A useful question is "what kind of market condition is being described?" rather than "what should I do now?"

For authenticated operator workflows, use Studio and Provider in a private environment. For public research context, pair this page with the Provider API boundary, Trading research platform, and Risk-aware trading automation.