Free lesson · Point-in-time data

An as-of join that never selects the future

Open interactive lessonPractice calculationsExplore labs

At a decision time, choose the most recent eligible information rather than the nearest observation.

Nearest-time matching can reach forward. A causal as-of rule first filters by usable time, then picks the latest eligible record. Entity identity, observation period and version key must also match; one global latest record is insufficient for a multi-company dataset.

Reject or explicitly aggregate duplicate eligible timestamps. If no record is available, preserve missingness instead of filling it from the future. A maximum staleness rule can be added after the causal join.

Apply it in a strategy

• Freeze inputs at the stated decision time and record their units.
• Join fundamentals, macro releases and alternative data into a causal feature table.
• Recompute the example, then change the material assumption and explain the difference.

Research deliverable

An as-of join that never selects the future: produce the worked calculation, a timestamped input record and a written decision addressing this limitation: Filtering by availability does not fix incorrect entity mapping or revisions that overwrote their earlier version.

These are synthetic mechanics examples, not historical performance or paper replications. Module evidence and research boundaries record the 12 September 2026 review.

# Python 3.10+; standard library and NumPy only.
# Synthetic teaching inputs; conventions and units are defined in the notation above.
def asof(records,decision):
    # records are (usable_timestamp,value) for one entity and series.
    eligible=[row for row in records if row[0]<=decision]
    if not eligible: return None
    latest=max(row[0] for row in eligible)
    matches=[row for row in eligible if row[0]==latest]
    if len(matches)!=1: raise ValueError('Resolve duplicate version timestamps first')
    return matches[0][1]

assert asof([(5,10),(9,12)],8)==10
assert asof([(5,10),(9,12)],4) is None
print(asof([(5,10),(9,12)],9))