How confident are we really about the Trakus time intervals? I know in the data description it says \"from what we can tell\" it's collected every 0.25 seconds. But from what I'm seeing, that doesn't seem quite right.
\nEvery race I look at closely, the apparent race time according to a 0.25s trakus interval (from after the run-up to the apparent instance of finish) is significantly less than the race time given for the winner in the public record. In fact, for some races, the entire trakus range is less than the winner's time, even including run-up and post-finish.
\nFor example, the fourth race - AQU, 2019-01-01, race 4. The publicly available video (from https://www.nyra.com/aqueduct/racing/replays) and record (https://www.equibase.com/premium/chartEmb.cfm?track=AQU&raceDate=01/01/19&cy=USA&rn=4) shows the winner with a time of 1:54.15.
\nHowever, my data for that race has a maximum \"trakus_index\" of 452 - equivalent to 112.75s (starting at index 1 = 0s), or 1:52.75. In addition, this clearly includes the run-up from velocity=0 (which I understand is NOT officially timed) and about 10s after the race, where all horses are clearly slowing.
\nIs anyone else seeing this? Am I missing something?
", "messages": [ { "id": "1939604", "postDate": "09/14/2022 21:08:57", "content": "How confident are we really about the Trakus time intervals? I know in the data description it says \"from what we can tell\" it's collected every 0.25 seconds. But from what I'm seeing, that doesn't seem quite right.
\nEvery race I look at closely, the apparent race time according to a 0.25s trakus interval (from after the run-up to the apparent instance of finish) is significantly less than the race time given for the winner in the public record. In fact, for some races, the entire trakus range is less than the winner's time, even including run-up and post-finish.
\nFor example, the fourth race - AQU, 2019-01-01, race 4. The publicly available video (from https://www.nyra.com/aqueduct/racing/replays) and record (https://www.equibase.com/premium/chartEmb.cfm?track=AQU&raceDate=01/01/19&cy=USA&rn=4) shows the winner with a time of 1:54.15.
\nHowever, my data for that race has a maximum \"trakus_index\" of 452 - equivalent to 112.75s (starting at index 1 = 0s), or 1:52.75. In addition, this clearly includes the run-up from velocity=0 (which I understand is NOT officially timed) and about 10s after the race, where all horses are clearly slowing.
\nIs anyone else seeing this? Am I missing something?
", "rawMarkdown": "How confident are we really about the Trakus time intervals? I know in the data description it says \"from what we can tell\" it's collected every 0.25 seconds. But from what I'm seeing, that doesn't seem quite right.\n\nEvery race I look at closely, the apparent race time according to a 0.25s trakus interval (from after the run-up to the apparent instance of finish) is significantly less than the race time given for the winner in the public record. In fact, for some races, the entire trakus range is less than the winner's time, even including run-up and post-finish. \n\nFor example, the fourth race - AQU, 2019-01-01, race 4. The publicly available video (from https://www.nyra.com/aqueduct/racing/replays) and record (https://www.equibase.com/premium/chartEmb.cfm?track=AQU&raceDate=01/01/19&cy=USA&rn=4) shows the winner with a time of 1:54.15.\n\nHowever, my data for that race has a maximum \"trakus_index\" of 452 - equivalent to 112.75s (starting at index 1 = 0s), or 1:52.75. In addition, this clearly includes the run-up from velocity=0 (which I understand is NOT officially timed) and about 10s after the race, where all horses are clearly slowing. \n\nIs anyone else seeing this? Am I missing something?", "votes": null }, { "id": "1944687", "postDate": "09/18/2022 14:00:16", "content": "Hello Jack:
\nI'm compiling the finish times for all of the races to see how consistent the trakus_index intervals may or may not be.
\nMy concern is two-fold:
\na) the index does not represent a uniform time interval between races
\nb) within the race itself does the duration between the trakus_index increments vary?
Without a solid handle on how reliable the trakus_index variable is, comparisons of velocities and accelerations--the most interesting topics to investigate with this granular GPS data--would be problematic.
\nI'll report back what I find.
\n-Brian
", "rawMarkdown": "Hello Jack:\n\nI'm compiling the finish times for all of the races to see how consistent the trakus_index intervals may or may not be.\n\nMy concern is two-fold:\na) the index does not represent a uniform time interval between races\nb) within the race itself does the duration between the trakus_index increments vary?\n\nWithout a solid handle on how reliable the trakus_index variable is, comparisons of velocities and accelerations--the most interesting topics to investigate with this granular GPS data--would be problematic.\n\nI'll report back what I find.\n\n-Brian", "votes": null }, { "id": "1946353", "postDate": "09/19/2022 19:10:13", "content": "Looking forward to it! Tough to have much confidence in an analysis with such questionable data. The relative positions seem to be accurate enough, so there may be some value if we can just assume the intervals are constant. But I'd rather not have to assume
", "rawMarkdown": "Looking forward to it! Tough to have much confidence in an analysis with such questionable data. The relative positions seem to be accurate enough, so there may be some value if we can just assume the intervals are constant. But I'd rather not have to assume", "votes": null }, { "id": "1947980", "postDate": "09/20/2022 18:20:18", "content": "At least from my experience the assumption that trakus indices are spaced in time by 0.25s in every race does provide realistic speed profiles, you can see an example of such profiles in my notebook. In my understanding the coordinate recording kicks in during run-up, therefore I refrain from deducing start positions or comparing profiles across races solely using trakus indices.
", "rawMarkdown": "At least from my experience the assumption that trakus indices are spaced in time by 0.25s in every race does provide realistic speed profiles, you can see an example of such profiles [in my notebook](https://www.kaggle.com/code/monogenea/big-data-derby-a-performance-proxy?scriptVersionId=105128325). In my understanding the coordinate recording kicks in during run-up, therefore I refrain from deducing start positions or comparing profiles across races solely using trakus indices.", "votes": null }, { "id": "1956615", "postDate": "09/26/2022 14:29:29", "content": "Hello Jack--I've verified that the Trakus devices don't stop logging until well after the finish line (during the \"gallop-out\")--see graphic attached from the Aqueduct races. So what we need at a minimum are the coordinates of the finish line so we can then determine the proper trakus_index value that represents when the first horse crossed the finish line. This will then allow us to correlate the finish time of record and help us determine reasonable time increments. I discussed the situation in greater detail in another thread. - BT
", "rawMarkdown": "Hello Jack--I've verified that the Trakus devices don't stop logging until well after the finish line (during the \"gallop-out\")--see graphic attached from the Aqueduct races. So what we need at a minimum are the coordinates of the finish line so we can then determine the proper trakus_index value that represents when the first horse crossed the finish line. This will then allow us to correlate the finish time of record and help us determine reasonable time increments. I discussed the situation in greater detail in another thread. - BT", "votes": null }, { "id": "1957258", "postDate": "09/26/2022 20:21:57", "content": "I noticed as well the trakus_index stops way after the end of race, some times. The finish line is not so simple to compute.
", "rawMarkdown": "I noticed as well the trakus_index stops way after the end of race, some times. The finish line is not so simple to compute.", "votes": null } ], "comments": [ { "id": 1944687, "author_name": "btimoney", "author_url": "", "post_date": "09/18/2022 14:00:16", "content": "Hello Jack:
\nI'm compiling the finish times for all of the races to see how consistent the trakus_index intervals may or may not be.
\nMy concern is two-fold:
\na) the index does not represent a uniform time interval between races
\nb) within the race itself does the duration between the trakus_index increments vary?
Without a solid handle on how reliable the trakus_index variable is, comparisons of velocities and accelerations--the most interesting topics to investigate with this granular GPS data--would be problematic.
\nI'll report back what I find.
\n-Brian
", "votes": null, "replies": [ { "id": 1946353, "author_name": "jackvance", "author_url": "", "post_date": "09/19/2022 19:10:13", "content": "Looking forward to it! Tough to have much confidence in an analysis with such questionable data. The relative positions seem to be accurate enough, so there may be some value if we can just assume the intervals are constant. But I'd rather not have to assume
", "votes": null, "replies": [] } ] }, { "id": 1947980, "author_name": "monogenea", "author_url": "", "post_date": "09/20/2022 18:20:18", "content": "At least from my experience the assumption that trakus indices are spaced in time by 0.25s in every race does provide realistic speed profiles, you can see an example of such profiles in my notebook. In my understanding the coordinate recording kicks in during run-up, therefore I refrain from deducing start positions or comparing profiles across races solely using trakus indices.
", "votes": null, "replies": [] }, { "id": 1956615, "author_name": "btimoney", "author_url": "", "post_date": "09/26/2022 14:29:29", "content": "Hello Jack--I've verified that the Trakus devices don't stop logging until well after the finish line (during the \"gallop-out\")--see graphic attached from the Aqueduct races. So what we need at a minimum are the coordinates of the finish line so we can then determine the proper trakus_index value that represents when the first horse crossed the finish line. This will then allow us to correlate the finish time of record and help us determine reasonable time increments. I discussed the situation in greater detail in another thread. - BT
", "votes": null, "replies": [] }, { "id": 1957258, "author_name": "georgeserbanut", "author_url": "", "post_date": "09/26/2022 20:21:57", "content": "I noticed as well the trakus_index stops way after the end of race, some times. The finish line is not so simple to compute.
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "1939604": "How confident are we really about the Trakus time intervals? I know in the data description it says \"from what we can tell\" it's collected every 0.25 seconds. But from what I'm seeing, that doesn't seem quite right.\n\nEvery race I look at closely, the apparent race time according to a 0.25s trakus interval (from after the run-up to the apparent instance of finish) is significantly less than the race time given for the winner in the public record. In fact, for some races, the entire trakus range is less than the winner's time, even including run-up and post-finish. \n\nFor example, the fourth race - AQU, 2019-01-01, race 4. The publicly available video (from https://www.nyra.com/aqueduct/racing/replays) and record (https://www.equibase.com/premium/chartEmb.cfm?track=AQU&raceDate=01/01/19&cy=USA&rn=4) shows the winner with a time of 1:54.15.\n\nHowever, my data for that race has a maximum \"trakus_index\" of 452 - equivalent to 112.75s (starting at index 1 = 0s), or 1:52.75. In addition, this clearly includes the run-up from velocity=0 (which I understand is NOT officially timed) and about 10s after the race, where all horses are clearly slowing. \n\nIs anyone else seeing this? Am I missing something?", "1944687": "Hello Jack:\n\nI'm compiling the finish times for all of the races to see how consistent the trakus_index intervals may or may not be.\n\nMy concern is two-fold:\na) the index does not represent a uniform time interval between races\nb) within the race itself does the duration between the trakus_index increments vary?\n\nWithout a solid handle on how reliable the trakus_index variable is, comparisons of velocities and accelerations--the most interesting topics to investigate with this granular GPS data--would be problematic.\n\nI'll report back what I find.\n\n-Brian", "1946353": "Looking forward to it! Tough to have much confidence in an analysis with such questionable data. The relative positions seem to be accurate enough, so there may be some value if we can just assume the intervals are constant. But I'd rather not have to assume", "1947980": "At least from my experience the assumption that trakus indices are spaced in time by 0.25s in every race does provide realistic speed profiles, you can see an example of such profiles [in my notebook](https://www.kaggle.com/code/monogenea/big-data-derby-a-performance-proxy?scriptVersionId=105128325). In my understanding the coordinate recording kicks in during run-up, therefore I refrain from deducing start positions or comparing profiles across races solely using trakus indices.", "1956615": "Hello Jack--I've verified that the Trakus devices don't stop logging until well after the finish line (during the \"gallop-out\")--see graphic attached from the Aqueduct races. So what we need at a minimum are the coordinates of the finish line so we can then determine the proper trakus_index value that represents when the first horse crossed the finish line. This will then allow us to correlate the finish time of record and help us determine reasonable time increments. I discussed the situation in greater detail in another thread. - BT", "1957258": "I noticed as well the trakus_index stops way after the end of race, some times. The finish line is not so simple to compute." }, "source": "meta" }