As it is already mentioned/discussed in various discussion threads and kernels that most reasonable reason of missing value is actually an event which is part of a game itself which also some what can be confirmed from missing position coordinates, velocity coordinates and player boost data in same proposition. So, keeping it in mind we can actually resolve the problem of missing value objectively instead on just dropping them which lead to some loss of data.
\nPlease feel free to share your idea or any improved we can do with above mentioned approaches.
", "messages": [ { "id": "1970620", "postDate": "10/04/2022 07:01:35", "content": "As it is already mentioned/discussed in various discussion threads and kernels that most reasonable reason of missing value is actually an event which is part of a game itself which also some what can be confirmed from missing position coordinates, velocity coordinates and player boost data in same proposition. So, keeping it in mind we can actually resolve the problem of missing value objectively instead on just dropping them which lead to some loss of data.
\nPlease feel free to share your idea or any improved we can do with above mentioned approaches.
", "rawMarkdown": "As it is already mentioned/discussed in various discussion threads and kernels that most reasonable reason of missing value is actually an event which is part of a game itself which also some what can be confirmed from missing position coordinates, velocity coordinates and player boost data in same proposition. So, keeping it in mind we can actually resolve the problem of missing value objectively instead on just dropping them which lead to some loss of data.\n\n1. **For player pos and vel coordinates** -: As after demolition car respawn near their team's goal so we can just assign random values which ranges near their team's goal i.e. For team A players (0,1,2) y range should be (-30, -70) and for team B players (3,4,5) y range should be (30, 70) and x range should be around (-60,60) and z will be 0 as player gonna start from ground itself. As in game too player is respawns randomly near to their team's goal.\n2. **For plater boost** -: it tricky to work with null values of player boost as when some player receives demolition, he/she respawn will less boost available than last time. So, we can fill the null values by reducing the player boost with 20% of previous value to null as train data is sequential but can't apply same for test data. But we can do one more thing take objectively good value from distribution of player boost and reduce it by 20-30% and assign to null values.\n\nPlease feel free to share your idea or any improved we can do with above mentioned approaches.", "votes": null }, { "id": "1992768", "postDate": "10/17/2022 22:30:23", "content": "Amazing insight. I think it's probably the best way to handle missing values in this dataset. What would you say it's a typical boost for a player recently destroy? also, do you think that been destroyed reduce the probabilities of your team to score and increase the probability of the other team to score?
", "rawMarkdown": "Amazing insight. I think it's probably the best way to handle missing values in this dataset. What would you say it's a typical boost for a player recently destroy? also, do you think that been destroyed reduce the probabilities of your team to score and increase the probability of the other team to score?", "votes": null } ], "comments": [ { "id": 1992768, "author_name": "pastorsoto", "author_url": "", "post_date": "10/17/2022 22:30:23", "content": "Amazing insight. I think it's probably the best way to handle missing values in this dataset. What would you say it's a typical boost for a player recently destroy? also, do you think that been destroyed reduce the probabilities of your team to score and increase the probability of the other team to score?
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "1970620": "As it is already mentioned/discussed in various discussion threads and kernels that most reasonable reason of missing value is actually an event which is part of a game itself which also some what can be confirmed from missing position coordinates, velocity coordinates and player boost data in same proposition. So, keeping it in mind we can actually resolve the problem of missing value objectively instead on just dropping them which lead to some loss of data.\n\n1. **For player pos and vel coordinates** -: As after demolition car respawn near their team's goal so we can just assign random values which ranges near their team's goal i.e. For team A players (0,1,2) y range should be (-30, -70) and for team B players (3,4,5) y range should be (30, 70) and x range should be around (-60,60) and z will be 0 as player gonna start from ground itself. As in game too player is respawns randomly near to their team's goal.\n2. **For plater boost** -: it tricky to work with null values of player boost as when some player receives demolition, he/she respawn will less boost available than last time. So, we can fill the null values by reducing the player boost with 20% of previous value to null as train data is sequential but can't apply same for test data. But we can do one more thing take objectively good value from distribution of player boost and reduce it by 20-30% and assign to null values.\n\nPlease feel free to share your idea or any improved we can do with above mentioned approaches.", "1992768": "Amazing insight. I think it's probably the best way to handle missing values in this dataset. What would you say it's a typical boost for a player recently destroy? also, do you think that been destroyed reduce the probabilities of your team to score and increase the probability of the other team to score?" }, "source": "meta" }