A research snapshot checklist for market observations

A number needs an observation context

An output without its instrument, timestamp and horizon is difficult to compare. A price observed at one moment and an estimate calculated from a different input snapshot may describe different conditions.

Record the observation before interpreting the disagreement. A missing source or timestamp is a research gap, not a reason to fill in a convenient value.

Copy this blank snapshot record

Instrument and identifier: ____. Observation time and timezone: ____. Retrieval time and timezone: ____. Source and source timestamp: ____.

Research horizon: ____. Coverage and missing fields: ____. Model or calculation version: ____. Input snapshot reference: ____. Assumptions and units: ____. Reviewer and unresolved questions: ____.

Keep observation time distinct from retrieval time. Retrieving an older observation now does not make the underlying observation current. Keep raw inputs and derived outputs distinguishable in the record.

A synthetic timing mismatch

This example is invented and contains no current market information. Researcher A records an observation at 10:00 UTC. Researcher B retrieves a different snapshot at 10:15 UTC and compares its output directly with A's.

Before attributing the difference to model quality, align the inputs and time basis. If alignment is impossible, report the mismatch. Two outputs carrying the same instrument label do not necessarily share a comparable input set.

Separate horizon from update time

A model estimating a short horizon and a model describing a longer horizon can disagree without either being evaluated on its intended task. The update time tells you when an output changed, not the period it is intended to describe.

Write the horizon in the snapshot and define what a later check would compare. Changing the horizon after seeing the outcome makes the comparison harder to interpret.

Keep assumptions and dependencies visible

Different model names do not establish statistical independence. Outputs may depend on the same source, share assumptions or respond to the same missing inputs. Record those dependencies before interpreting agreement as corroboration.

An AI explanation should point back to the recorded snapshot. Clear prose cannot repair a missing timestamp or an unsupported assumption. Keep unresolved questions with the output so another reviewer can inspect them.

Use the checklist as a research gate

If the record is incomplete, narrow the claim or obtain the missing input before comparing outputs. Preserve the original snapshot when adding a correction so the reason for a changed interpretation remains visible.

This checklist supports research review. It contains no trading recommendation, backtest result or prediction about returns.