Dear all, dear organizers,
\nI looked at one flying ball and the tracking data seems to be against what the kinematics of ballistic motion predicts.
\nThe football acceleration should be constant during the flight, at 10.72 yards/s^2 ( 9.81 m/s^2 in SI), the gravitational acceleration , and according to the tracking data it is not.
\nThe football speed in the horizontal plane (x-y) should be constant during the flight (after the kick), but it is not.
\nIs there any explanation for why the data does not match the data we expect from free fall, or ballistic motion?
\nHere's the notebook where I describe my analysis and my findings.
\nhttps://www.kaggle.com/louisbunuel/one-weird-flying-ball/
Dear all, dear organizers,
\nI looked at one flying ball and the tracking data seems to be against what the kinematics of ballistic motion predicts.
\nThe football acceleration should be constant during the flight, at 10.72 yards/s^2 ( 9.81 m/s^2 in SI), the gravitational acceleration , and according to the tracking data it is not.
\nThe football speed in the horizontal plane (x-y) should be constant during the flight (after the kick), but it is not.
\nIs there any explanation for why the data does not match the data we expect from free fall, or ballistic motion?
\nHere's the notebook where I describe my analysis and my findings.
\nhttps://www.kaggle.com/louisbunuel/one-weird-flying-ball/
Hello,
\nAcceleration (and everything that is measured) is always in the horizontal x-y direction and not in the vertical z direction. Thus, acceleration due to gravity is not measured so the value for acceleration will not be a close to constant value around 9.8 m/s^2 with gravity and air resistance.
\nMoreover, in terms of speed in the air, part of it is the game is not played in a vacuum and factors such as air resistance / wind are active. Also, the spin of the ball about either of its axes could have an effect.
\nHowever, all in all part of it is the data is imperfect and measurements could be off. The tagging of events could also be off. My guess is that the ball actually landed closer to the 3 second mark on your x axis as opposed to when the event was tagged \"touchback\".
", "rawMarkdown": "Hello,\n\nAcceleration (and everything that is measured) is always in the horizontal x-y direction and not in the vertical z direction. Thus, acceleration due to gravity is not measured so the value for acceleration will not be a close to constant value around 9.8 m/s^2 with gravity and air resistance.\n\nMoreover, in terms of speed in the air, part of it is the game is not played in a vacuum and factors such as air resistance / wind are active. Also, the spin of the ball about either of its axes could have an effect.\n\nHowever, all in all part of it is the data is imperfect and measurements could be off. The tagging of events could also be off. My guess is that the ball actually landed closer to the 3 second mark on your x axis as opposed to when the event was tagged \"touchback\".", "votes": null }, { "id": "1568171", "postDate": "11/02/2021 14:53:53", "content": "Hello,
\nMay I ask how is the ball acceleration measured? If it is with an accelerometer stuck to it, then the acceleration will be a = sqrt(ax^2+ay^2+az^2). I am not aware of \"directional accelerometers\", but I may be wrong. Now the players stay pretty much in the x-y plane, but the football does not. If the measured ball acceleration were only the one in the x-y plane, I would have expected an acceleration proportional to the viscous drag of the air, that is, proportional to the instant velocity of the ball, which we don't see either. You are very right about the spin of the ball, it may be a large effect, though I still cannot tell how can it account for the spikes such as the one at t=2s.
\nThe thing is I very much want to correctly identify the flight parameters and I wanted to doublecheck before I go with further with my idea.
All of the data is measured via 20-30 RF antenna's installed inside the stadium and RFID tags in each player's shoulder pads and the ball.
\nThus, acceleration is not measured via an accelerometer. In our data, the acceleration is the magnitude of the ball's change in velocity in the direction it is moving. Thus, it is calculated by a = sqrt(ax^2+ay^2). If the data appears to be unclean (which it may be in a few examples), I would recommend recalculating it using a smoothed change in x/y or change in speed.
\nOf course the ball and players change in the z direction, but that is not captured in our data.
", "rawMarkdown": "All of the data is measured via 20-30 RF antenna's installed inside the stadium and RFID tags in each player's shoulder pads and the ball.\n\nThus, acceleration is not measured via an accelerometer. In our data, the acceleration is the magnitude of the ball's change in velocity in the direction it is moving. Thus, it is calculated by a = sqrt(ax^2+ay^2). If the data appears to be unclean (which it may be in a few examples), I would recommend recalculating it using a smoothed change in x/y or change in speed.\n\nOf course the ball and players change in the z direction, but that is not captured in our data.", "votes": null } ], "comments": [ { "id": 1568143, "author_name": "tombliss", "author_url": "", "post_date": "11/02/2021 14:25:41", "content": "Hello,
\nAcceleration (and everything that is measured) is always in the horizontal x-y direction and not in the vertical z direction. Thus, acceleration due to gravity is not measured so the value for acceleration will not be a close to constant value around 9.8 m/s^2 with gravity and air resistance.
\nMoreover, in terms of speed in the air, part of it is the game is not played in a vacuum and factors such as air resistance / wind are active. Also, the spin of the ball about either of its axes could have an effect.
\nHowever, all in all part of it is the data is imperfect and measurements could be off. The tagging of events could also be off. My guess is that the ball actually landed closer to the 3 second mark on your x axis as opposed to when the event was tagged \"touchback\".
", "votes": null, "replies": [ { "id": 1568171, "author_name": "louisbunuel", "author_url": "", "post_date": "11/02/2021 14:53:53", "content": "Hello,
\nMay I ask how is the ball acceleration measured? If it is with an accelerometer stuck to it, then the acceleration will be a = sqrt(ax^2+ay^2+az^2). I am not aware of \"directional accelerometers\", but I may be wrong. Now the players stay pretty much in the x-y plane, but the football does not. If the measured ball acceleration were only the one in the x-y plane, I would have expected an acceleration proportional to the viscous drag of the air, that is, proportional to the instant velocity of the ball, which we don't see either. You are very right about the spin of the ball, it may be a large effect, though I still cannot tell how can it account for the spikes such as the one at t=2s.
\nThe thing is I very much want to correctly identify the flight parameters and I wanted to doublecheck before I go with further with my idea.
All of the data is measured via 20-30 RF antenna's installed inside the stadium and RFID tags in each player's shoulder pads and the ball.
\nThus, acceleration is not measured via an accelerometer. In our data, the acceleration is the magnitude of the ball's change in velocity in the direction it is moving. Thus, it is calculated by a = sqrt(ax^2+ay^2). If the data appears to be unclean (which it may be in a few examples), I would recommend recalculating it using a smoothed change in x/y or change in speed.
\nOf course the ball and players change in the z direction, but that is not captured in our data.
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "1567937": "Dear all, dear organizers, \nI looked at one flying ball and the tracking data seems to be against what the kinematics of ballistic motion predicts. \nThe football acceleration should be constant during the flight, at 10.72 yards/s^2 ( 9.81 m/s^2 in SI), the gravitational acceleration , and according to the tracking data it is not. \nThe football speed in the horizontal plane (x-y) should be constant during the flight (after the kick), but it is not. \nIs there any explanation for why the data does not match the data we expect from free fall, or ballistic motion?\nHere's the notebook where I describe my analysis and my findings. \nhttps://www.kaggle.com/louisbunuel/one-weird-flying-ball/", "1568143": "Hello,\n\nAcceleration (and everything that is measured) is always in the horizontal x-y direction and not in the vertical z direction. Thus, acceleration due to gravity is not measured so the value for acceleration will not be a close to constant value around 9.8 m/s^2 with gravity and air resistance.\n\nMoreover, in terms of speed in the air, part of it is the game is not played in a vacuum and factors such as air resistance / wind are active. Also, the spin of the ball about either of its axes could have an effect.\n\nHowever, all in all part of it is the data is imperfect and measurements could be off. The tagging of events could also be off. My guess is that the ball actually landed closer to the 3 second mark on your x axis as opposed to when the event was tagged \"touchback\".", "1568171": "Hello, \nMay I ask how is the ball acceleration measured? If it is with an accelerometer stuck to it, then the acceleration will be a = sqrt(ax^2+ay^2+az^2). I am not aware of \"directional accelerometers\", but I may be wrong. Now the players stay pretty much in the x-y plane, but the football does not. If the measured ball acceleration were only the one in the x-y plane, I would have expected an acceleration proportional to the viscous drag of the air, that is, proportional to the instant velocity of the ball, which we don't see either. You are very right about the spin of the ball, it may be a large effect, though I still cannot tell how can it account for the spikes such as the one at t=2s. \nThe thing is I very much want to correctly identify the flight parameters and I wanted to doublecheck before I go with further with my idea.", "1568342": "All of the data is measured via 20-30 RF antenna's installed inside the stadium and RFID tags in each player's shoulder pads and the ball.\n\nThus, acceleration is not measured via an accelerometer. In our data, the acceleration is the magnitude of the ball's change in velocity in the direction it is moving. Thus, it is calculated by a = sqrt(ax^2+ay^2). If the data appears to be unclean (which it may be in a few examples), I would recommend recalculating it using a smoothed change in x/y or change in speed.\n\nOf course the ball and players change in the z direction, but that is not captured in our data." }, "source": "meta" }