Paper Bot Data Feeds: Live Bars, REST Backfills, and Fail-Closed Logic

Abstract

Paper trading depends on data freshness, but live streams are not always complete. A robust paper loop should define how it combines provisional live bars, completed-bar backfills, option chain checks, and quote validation.

The goal is not to trade every possible signal. The goal is to trade only when the data route is understood.

Provisional vs Completed Data

A live feed can provide fast current-minute information. A REST backfill can provide more reliable completed bars after the minute closes. These are different objects and should be labeled differently in the system.

Developers should avoid silently mixing them. If a completed backfill overwrites a provisional bar, that should be part of the data policy, not an accident.

Signal Age Controls

When completed bars arrive slightly after close, the paper loop needs a max signal age. Too strict, and the system misses legitimate completed-bar signals. Too loose, and it starts acting on stale context.

A max entry signal age makes this tradeoff explicit. The paper bot can accept delayed completed data inside a defined window and fail closed outside it.

Option Selection Still Rules

Even with clean underlying bars, the option route can reject the trade. No listed expiry, empty contract pool, stale quote, wide spread, or missing bid/ask should block the order. These blocks are not bugs; they are the live version of realistic replay.

Takeaway

Paper bot data feeds should be boring, explicit, and conservative. Label provisional data, backfill completed bars, enforce signal age, and fail closed when the option surface does not support the trade.