It's seeming more and more likely that the Coronavirus outbreak in the US could impact the tournament in some fashion. Has anyone though about how this could effect results or if it would change your modeling?
\n\nThe most obvious to me is removing any impact of the crowd, though cities may still be a useful feature in some other way.
\n\nhttps://www.wbrc.com/2020/03/07/nba-tells-teams-plan-empty-arenas-lebron-says-no-way/
", "messages": [ { "id": "765986", "postDate": "03/07/2020 13:43:20", "content": "It's seeming more and more likely that the Coronavirus outbreak in the US could impact the tournament in some fashion. Has anyone though about how this could effect results or if it would change your modeling?
\n\nThe most obvious to me is removing any impact of the crowd, though cities may still be a useful feature in some other way.
\n\nhttps://www.wbrc.com/2020/03/07/nba-tells-teams-plan-empty-arenas-lebron-says-no-way/
", "rawMarkdown": "It's seeming more and more likely that the Coronavirus outbreak in the US could impact the tournament in some fashion. Has anyone though about how this could effect results or if it would change your modeling?\n\nThe most obvious to me is removing any impact of the crowd, though cities may still be a useful feature in some other way.\n\nhttps://www.wbrc.com/2020/03/07/nba-tells-teams-plan-empty-arenas-lebron-says-no-way/", "votes": null }, { "id": "766152", "postDate": "03/07/2020 18:31:28", "content": "There have been attempts in years past to use the cities data to develop features based on travel distance or deviation from home city altitude. Those are both effects that could reasonably take a physical toll on the players, though I've never had such features increase performance of my models.
\n\nModeling effects of the virus itself would be difficult given that it's an unprecedented event and there are no training data that could be used to develop the model. I suppose you could, based on the location of virus outbreaks, assign each college a probability that a player becomes infected and use that to calculate an expected number of players who have to miss playing for their team in the tournament (due to being quarantined). That would diminish team strength. You might be able to get such numbers from websites that publish location-based Coronavirus forecasts.
", "rawMarkdown": "There have been attempts in years past to use the cities data to develop features based on travel distance or deviation from home city altitude. Those are both effects that could reasonably take a physical toll on the players, though I've never had such features increase performance of my models. \n\nModeling effects of the virus itself would be difficult given that it's an unprecedented event and there are no training data that could be used to develop the model. I suppose you could, based on the location of virus outbreaks, assign each college a probability that a player becomes infected and use that to calculate an expected number of players who have to miss playing for their team in the tournament (due to being quarantined). That would diminish team strength. You might be able to get such numbers from websites that publish location-based Coronavirus forecasts.", "votes": null }, { "id": "766247", "postDate": "03/07/2020 22:45:37", "content": "In the case that players are actually sick, a model that has been built to account for significant injuries might actually be relevant. That's an interesting point.
\n\nIt is interesting that with no precedent, modeling the problem becomes effectively impossible. My gut instinct is that the games may actually become more predictable in the case that stadiums are empty, which might make more aggressive models perform better... but then again. Who knows!
", "rawMarkdown": "In the case that players are actually sick, a model that has been built to account for significant injuries might actually be relevant. That's an interesting point.\n\nIt is interesting that with no precedent, modeling the problem becomes effectively impossible. My gut instinct is that the games may actually become more predictable in the case that stadiums are empty, which might make more aggressive models perform better... but then again. Who knows!", "votes": null }, { "id": "769426", "postDate": "03/11/2020 22:23:50", "content": "I was thinking that same thing @rsa013 (empty stadiums may be more predictable games). But then I thought it could go the other way too ... some \"better\" teams may play down to the level of their opponents or get lazy without the hype. Then the lack of a crowd could lead to an upset ... this will be interesting anyhow!
", "rawMarkdown": "I was thinking that same thing @rsa013 (empty stadiums may be more predictable games). But then I thought it could go the other way too ... some \"better\" teams may play down to the level of their opponents or get lazy without the hype. Then the lack of a crowd could lead to an upset ... this will be interesting anyhow!", "votes": null } ], "comments": [ { "id": 766152, "author_name": "goodspellr", "author_url": "", "post_date": "03/07/2020 18:31:28", "content": "There have been attempts in years past to use the cities data to develop features based on travel distance or deviation from home city altitude. Those are both effects that could reasonably take a physical toll on the players, though I've never had such features increase performance of my models.
\n\nModeling effects of the virus itself would be difficult given that it's an unprecedented event and there are no training data that could be used to develop the model. I suppose you could, based on the location of virus outbreaks, assign each college a probability that a player becomes infected and use that to calculate an expected number of players who have to miss playing for their team in the tournament (due to being quarantined). That would diminish team strength. You might be able to get such numbers from websites that publish location-based Coronavirus forecasts.
", "votes": null, "replies": [ { "id": 766247, "author_name": "rsa013", "author_url": "", "post_date": "03/07/2020 22:45:37", "content": "In the case that players are actually sick, a model that has been built to account for significant injuries might actually be relevant. That's an interesting point.
\n\nIt is interesting that with no precedent, modeling the problem becomes effectively impossible. My gut instinct is that the games may actually become more predictable in the case that stadiums are empty, which might make more aggressive models perform better... but then again. Who knows!
", "votes": null, "replies": [] } ] }, { "id": 769426, "author_name": "thawatt", "author_url": "", "post_date": "03/11/2020 22:23:50", "content": "I was thinking that same thing @rsa013 (empty stadiums may be more predictable games). But then I thought it could go the other way too ... some \"better\" teams may play down to the level of their opponents or get lazy without the hype. Then the lack of a crowd could lead to an upset ... this will be interesting anyhow!
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "765986": "It's seeming more and more likely that the Coronavirus outbreak in the US could impact the tournament in some fashion. Has anyone though about how this could effect results or if it would change your modeling?\n\nThe most obvious to me is removing any impact of the crowd, though cities may still be a useful feature in some other way.\n\nhttps://www.wbrc.com/2020/03/07/nba-tells-teams-plan-empty-arenas-lebron-says-no-way/", "766152": "There have been attempts in years past to use the cities data to develop features based on travel distance or deviation from home city altitude. Those are both effects that could reasonably take a physical toll on the players, though I've never had such features increase performance of my models. \n\nModeling effects of the virus itself would be difficult given that it's an unprecedented event and there are no training data that could be used to develop the model. I suppose you could, based on the location of virus outbreaks, assign each college a probability that a player becomes infected and use that to calculate an expected number of players who have to miss playing for their team in the tournament (due to being quarantined). That would diminish team strength. You might be able to get such numbers from websites that publish location-based Coronavirus forecasts.", "766247": "In the case that players are actually sick, a model that has been built to account for significant injuries might actually be relevant. That's an interesting point.\n\nIt is interesting that with no precedent, modeling the problem becomes effectively impossible. My gut instinct is that the games may actually become more predictable in the case that stadiums are empty, which might make more aggressive models perform better... but then again. Who knows!", "769426": "I was thinking that same thing @rsa013 (empty stadiums may be more predictable games). But then I thought it could go the other way too ... some \"better\" teams may play down to the level of their opponents or get lazy without the hype. Then the lack of a crowd could lead to an upset ... this will be interesting anyhow!" }, "source": "meta" }