Dataset train-test entity leakage calculator

Calculate entities present in more than one supplied split and record counts by split with your own figures.

Your numbers

Use the units shown. Results update as you change valid inputs.

Use entities present in more than one supplied split and record counts by split to compare changes to your inputs. The method and limits below explain the scope.

Examples

Examples

Try the starting example

Records: unique record ID, entity ID, split label, YYYY-MM-DD date
R1,E1,train,2026-10-01 R2,E1,test,2026-10-20 R3,E2,train,2026-10-15 R4,E3,test,2026-10-05
Supplied chronological cutoff date
2026-10-10
Entities present in more than one supplied split
E1
Record counts by split
train: 2; test: 2
Train records dated strictly after cutoff
R3
Other-split records dated on or before cutoff
R4
Unique entities
3

Try another scenario

Records: unique record ID, entity ID, split label, YYYY-MM-DD date
R1,E1,train,2026-10-01 R2,E1,test,2026-10-20 R3,E2,train,2026-10-15 R4,E3,test,2026-10-05
Supplied chronological cutoff date
2026-10-16
Entities present in more than one supplied split
E1
Record counts by split
train: 2; test: 2
Train records dated strictly after cutoff
None
Other-split records dated on or before cutoff
R4
Unique entities
3
Calculation method

Calculation method

Entities present in more than one supplied split

Exact entity-ID equality across distinct split labels.

Record counts by split

Count supplied records for each literal split label.

Train records dated strictly after cutoff

The literal train split is expected on or before cutoff.

Other-split records dated on or before cutoff

All split labels other than train are compared to strictly-after cutoff.

Unique entities

Distinct literal entity IDs.

Symbols and units
a
Records: unique record ID, entity ID, split label, YYYY-MM-DD date
b
Supplied chronological cutoff date

Use the units shown. A list keeps the order in which you enter its values. The symbol ^ means “raised to the power of”.

Assumptions & limits

Assumptions & limits

  • A supplied identity/date audit, not proof of statistical leakage. Exact identifiers only; train is a literal label, all other splits are treated as holdouts, and dates have no time-of-day. No source data is changed.
Are these inputs private?

These calculations run in your browser. A shared link contains the values you choose to share. Only questions submitted to the separate AI finder are sent to OpenAI.

Further reading

OpenStax: arithmetic and geometry ↗

Background reading; this does not imply endorsement or independent review of SumFizz.

Share this page
Share this calculator

Leave unchecked to share the page without your entered values.