I didn't see another post about this data quality issue, so I thought other teams might benefit from this, especially teams using the PFF special teams annotations.
\nWe found that the PFF data annotations (e.g. gunners, vises, tacklers, etc.) sometimes reference jersey numbers that do not match any players in the tracking data for the corresponding play. We looked at all the cases and found the correct jersey number from the tracking data.
\nFor example, in game 2020100400, play 169, vises are listed as CAR 25 and CAR 20, but there is no CAR 20 in the tracking data. It should be CAR 41, Eli Apple.
There was also one game where two Bills players had the jersey number 44, where one of them should be number 40.
\nThe patches can be found in the last section of this notebook: https://www.kaggle.com/vingkan/pff-tracking-data-player-mismatches
\nI included both the raw CSV data for the patches for everyone to use, as well as sample code to apply the patches for Python users.
\nThank you to the organizers, as the dataset is very high quality! Hopefully this small patch helps other teams.
", "messages": [ { "id": 1634827, "postDate": "2022-01-01T08:28:30.863Z", "content": "I didn't see another post about this data quality issue, so I thought other teams might benefit from this, especially teams using the PFF special teams annotations.
\nWe found that the PFF data annotations (e.g. gunners, vises, tacklers, etc.) sometimes reference jersey numbers that do not match any players in the tracking data for the corresponding play. We looked at all the cases and found the correct jersey number from the tracking data.
\nFor example, in game 2020100400, play 169, vises are listed as CAR 25 and CAR 20, but there is no CAR 20 in the tracking data. It should be CAR 41, Eli Apple.
There was also one game where two Bills players had the jersey number 44, where one of them should be number 40.
\nThe patches can be found in the last section of this notebook: https://www.kaggle.com/vingkan/pff-tracking-data-player-mismatches
\nI included both the raw CSV data for the patches for everyone to use, as well as sample code to apply the patches for Python users.
\nThank you to the organizers, as the dataset is very high quality! Hopefully this small patch helps other teams.
", "rawMarkdown": "I didn't see another post about this data quality issue, so I thought other teams might benefit from this, especially teams using the PFF special teams annotations.\n\nWe found that the PFF data annotations (e.g. gunners, vises, tacklers, etc.) sometimes reference jersey numbers that do not match any players in the tracking data for the corresponding play. We looked at all the cases and found the correct jersey number from the tracking data.\n\nFor example, in game `2020100400`, play `169`, vises are listed as CAR 25 and CAR 20, but there is no CAR 20 in the tracking data. It should be CAR 41, Eli Apple.\n\nThere was also one game where two Bills players had the jersey number 44, where one of them should be number 40.\n\nThe patches can be found in the last section of this notebook: https://www.kaggle.com/vingkan/pff-tracking-data-player-mismatches\n\nI included both the raw CSV data for the patches for everyone to use, as well as sample code to apply the patches for Python users.\n\nThank you to the organizers, as the dataset is very high quality! Hopefully this small patch helps other teams.", "votes": 2 } ], "comments": [], "raw_markdown_by_id": { "1634827": "I didn't see another post about this data quality issue, so I thought other teams might benefit from this, especially teams using the PFF special teams annotations.\n\nWe found that the PFF data annotations (e.g. gunners, vises, tacklers, etc.) sometimes reference jersey numbers that do not match any players in the tracking data for the corresponding play. We looked at all the cases and found the correct jersey number from the tracking data.\n\nFor example, in game `2020100400`, play `169`, vises are listed as CAR 25 and CAR 20, but there is no CAR 20 in the tracking data. It should be CAR 41, Eli Apple.\n\nThere was also one game where two Bills players had the jersey number 44, where one of them should be number 40.\n\nThe patches can be found in the last section of this notebook: https://www.kaggle.com/vingkan/pff-tracking-data-player-mismatches\n\nI included both the raw CSV data for the patches for everyone to use, as well as sample code to apply the patches for Python users.\n\nThank you to the organizers, as the dataset is very high quality! Hopefully this small patch helps other teams." } }