2025 Abu Dhabi Grand Prix: Object-Centric Race, Tyre-Strategy and Race-Control Event Log (OCEL 2.0)

This dataset contains an object-centric event log of the 2025 Abu Dhabi Formula 1 race at Yas Marina Circuit, held on 7 December 2025. It was extracted from the public OpenF1 API (session 9839, meeting 1276) using a snapshot retrieved on 2 November 2026. The source snapshot contains 20 drivers, 1,156 laps, 27 pit stops, 47 tyre stints, 109 race-control messages, 154 weather observations and 20 session results. The primary file, abudhabi-2025-race.ocel.json.gz, is gzip-compressed OCEL 2.0 JSON. It contains 1,513 events across 21 activity types and 1,281 objects across seven object types: RaceSession, Team, Driver, RaceEntry, Lap, TyreStint and PitStop. The log includes 6,847 qualified event-to-object relationships, 2,557 qualified object-to-object relationships, and timed attributes for weather observations, lap-range stint/compound changes and reported lap-time deletions. Events cover the complete source observation window from 12:06:07.170 to 14:39:07.948 UTC, including pre-race context and post-race notifications. The dataset supports object-centric process discovery, driver and team process comparisons, tyre-strategy variants, pit-duration analysis and race-control notification workflows. RaceEntry is the recommended leading object type. Pit observations explicitly connect outgoing and incoming stints; successive stints remain distinct even when the tyre compound is unchanged. Race-control messages link to explicitly identified drivers, including multi-driver messages, and lap-time deletion notices link to the affected laps. The snapshot exhibits five compound-sequence variants, 13 one-stop and seven two-stop strategies, and 40 deletion notifications concerning 39 distinct laps. Temporal semantics are documented explicitly. Lap-completion times are derived from the API's approximate lap starts plus reported durations. Stint markers identify the start of the first lap in a reported stint range, not the physical tyre-installation time. Pit events use the source timestamp as a recorded observation with an unspecified physical phase; pit-lane and stationary durations remain separate. Twenty additional last-recorded-lap markers are derived milestones, not additional laps. Final classification is stored as untimed retrospective attributes because the API does not provide a result-publication timestamp. Source nulls remain missing; one stationary pit duration and four final race durations are absent. Global weather and race-control context are not duplicated onto all driver traces. The accompanying repository provides archived source responses and checksums, deterministic extraction scripts, pinned dependencies, the mapping specification, quality reports and reproducible process-mining analyses. Validation includes complete source-row reconciliation, a strict local profile of the documented OCEL 2.0 JSON structure, referential and temporal checks, eight regression tests, independent import with PM4Py, object-centric directly-follows discovery and 20 flattened driver traces. The exported gzip file is byte-reproducible in the pinned environment. For time-aware state analysis, use the original OCEL timestamps: PM4Py promotes each object's first attribute value into its static table, including values first observed after the race began. The data are intended for descriptive and methodological research. A single race, approximate timing and delayed control messages do not establish causal strategy advantage or precise physical incident times. Source attribution: OpenF1, https://openf1.org/ and https://openf1.org/docs/. Extraction and documentation: https://github.com/fit-alessandro-berti/abudhabi2025-ocel-extraction. Acknowledgments: We thank Jean-Christophe Schmitt ([e-mail]) for suggesting the extraction of this dataset and performing a preliminary OCEL extraction. The present release provides a further refined extraction.