A reliable esports dataset begins with precise definitions. Decide what a row represents, preserve the organiser’s match reference and store the time with its date and zone. The small examples below are invented teaching records, not live tournament data or claims about real teams.
Give each level its own identity
Use a tournament edition identifier, a stage identifier, a series identifier and, where needed, an individual game identifier. A team name and date alone can be ambiguous when the same teams meet again or a session crosses midnight. The official International event record is an example of an edition-specific source to consult for actual fixtures. In your own notes, keep the source reference as well as a readable label. Preserve team aliases separately from roster identity. An organiser may rename a team display without changing the underlying match, and a club may change players while retaining the same public name.
Store an instant, not just a clock reading
The IANA Time Zone Database models geographic time rules and their changes. A named zone is useful when a future local schedule must account for seasonal or legal changes. An explicit offset is useful when recording the exact instant shown in a published fixture. The W3C Date and Time Formats Note illustrates dates with UTC and numeric offsets. In practical terms, save the original local date and time, the zone or offset supplied by the source, and a UTC representation when the conversion is unambiguous. Do not guess a zone from a three-letter abbreviation alone.
An explicit conversion example
| Representation | Value | Interpretation |
|---|---|---|
| UTC | 2026-10-23T12:00:00Z | Reference instant |
| Offset +08:00 | 2026-10-23T20:00:00+08:00 | Eight hours ahead of UTC |
| Offset -04:00 | 2026-10-23T08:00:00-04:00 | Four hours behind UTC |
This is arithmetic with explicit offsets, not a schedule for any real match or a claim that a city always uses a particular offset. If the source only says that a match follows the previous series, store that sequence and leave the exact start time pending. Do not manufacture a timestamp to make sorting easier. A reminder can be approximate while an archival record remains honest about what is known.
Preserve status and correction history
Keep scheduled, live, completed, postponed, cancelled and administratively awarded records distinguishable. Store scheduled start and actual start in separate fields when both are known. If a result is corrected, retain the previous value, the replacement, the time of correction and the source that justified it. A simple notes column is better than silent overwriting. Avoid conflating a missing game with an unplayed game: the former needs evidence, while the latter may be a normal consequence of a series ending early. A dataset can contain blanks and still be accurate; invented precision creates a false impression of completeness.
Calculate statistics with a visible denominator
Suppose an invented team wins two series out of three and four games out of seven. Its series rate is 2/3, about 66.7%, and its game rate is 4/7, about 57.1%. Label both values with their sample size and period. If you later add a forfeit, decide explicitly whether it belongs in a results table, a played-game analysis or both under different fields. For average game duration, count only the games for which you have verified duration data and say how many that is. Do not fill missing times with zero. Compare records from compatible formats and explain differences in opponent quality, roster and game version.
Checklist before exporting a table
- Give each row a documented unit and a stable source reference. Keep series outcomes and individual game outcomes in separate tables or clearly separated fields.
- Store dates with the relevant zone or explicit offset, retain the original display and check conversions that cross midnight. Do not use a guessed time to hide an unknown start.
- Mark pending or administrative outcomes clearly and keep a correction log. Preserve evidence links and check dates so another reader can audit a changed result.
- Publish the counting rule, sample period and missing-data policy beside any statistic. Use invented examples only when clearly labelled, and never mix them into a real event archive.
Frequently asked questions
Why is 20:00 without a time zone insufficient?
It does not identify a unique instant or even necessarily the viewer’s calendar date. Keep the source date, location-based zone or explicit offset, and any confirmed UTC conversion.
Should an unused fifth game in a BO5 be recorded as zero?
No. Mark it as not played if the series ended earlier. A zero duration or fabricated score can distort totals and averages.
Do I need a complicated database to start?
No. A small table with clear identifiers, counting units, sources, dates and correction notes is more useful than a larger table filled with unexplained or invented values.
Sources and verification
- The International official event record — Valve — 2026-10-23
- Time Zone Database — IANA — 2026-10-23
- Date and Time Formats (W3C Note) — W3C — 2026-10-23