tag:blogger.com,1999:blog-645749475857315124.post2424048136109382586..comments2016-02-04T10:28:40.892-05:00Comments on The 14-Game Tournament: The 2012-13 Ivy Basketball Projection PostMichael Jameshttp://www.blogger.com/profile/06461922638043769098noreply@blogger.comBlogger7125tag:blogger.com,1999:blog-645749475857315124.post-26873801160438967582012-11-30T07:59:57.204-05:002012-11-30T07:59:57.204-05:00The proposal to scrap AI averaging would involve a...The proposal to scrap AI averaging would involve an AI floor of 180 for all schools. There would be no differences in what one school could accept versus another. (Now, schools would obviously have their own internal standards of their choosing, so this wouldn't mean that every 180+ would automatically get accepted, but at least that decision would be completely up to the school in question).<br /><br />As for Nate Silver, from a forecast perspective, he was very close to 100% right (great HSAC post on this... http://harvardsportsanalysis.wordpress.com/2012/11/08/nate-silver-and-forecasting-as-an-imperfect-science/), but more importantly, the mechanism for building the model revealed (amazingly, in my view) that it is the national polls which were quite biased. Building up from the state polls, something which I felt would increase the noisiness of the projections, was actually the way to get to a more accurate number. Very interesting stuff.Michael Jameshttps://www.blogger.com/profile/06461922638043769098noreply@blogger.comtag:blogger.com,1999:blog-645749475857315124.post-29238752839507307802012-11-30T00:13:04.352-05:002012-11-30T00:13:04.352-05:00In your proposal to scrap AI averages for each Ivy...In your proposal to scrap AI averages for each Ivy in favor of higher AI floors, are you suggesting that the differential between various Ivies be maintained, in particular that Yale, Princeton and Harvard be held to a higher AI threshold?<br /><br />Parenthetically, I'm sure you've noticed that Nate Silver's 538 forecast was 100% vindicated on Election Day. The electoral race was not at all close and Silver called the component state races perfectly. His underlying assumptions on the turnout demographics must be right on target.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-645749475857315124.post-24066049492304674312012-11-04T17:59:00.032-05:002012-11-04T17:59:00.032-05:00Thanks for bringing this up, because it's a qu...Thanks for bringing this up, because it's a question I get a lot.<br /><br />I am not at all dismissive of Nate's 538 forecast. He has a model which has been built to fit historical data and to use that to project the future, just like mine. And as I'm sure he does with his, I constantly re-evaluate the assumptions underlying my model. For instance, this offseason, I did a ton of work on home court advantage to replace a previous assumption that all HCA is created equal (it isn't). I've also done some tweaking to the pace at which my preseason projections are filtered out of model and the way I use similarity scores to project player performance for preseason rankings.<br /><br />Models undergo a constant evolution, in which prior assumptions are tested and shred and new data restructures the weighting of the inputs.<br /><br />My points about Nate's model are not meant to say that data-driven analysis of political outcomes is wrong as a concept. I'm merely hypothesizing about assumptions backing the model - primarily that if there is a difference between a build-up from state models and national models, the difference might arise from different composite turnout models. Nate's a pro and has way more data than I do, so it's entirely possible that his assumption will be right. But even in the unlikely event that my assumption is right, that a) would have to be pretty severe to change the result and b) wouldn't be an argument against using data-driven analysis to predict the outcome, just an argument to tweak the assumptions.<br /><br />Hopefully, that makes sense.<br /><br />(Also, as a further note, I'm pretty ardently against both parties for a variety of reasons, so there isn't really an outcome Tuesday for me to like. I'm just fascinated by the numbers and the predictive process - the outcome is only really interesting to me in so much as it's another data point.)Michael Jameshttps://www.blogger.com/profile/06461922638043769098noreply@blogger.comtag:blogger.com,1999:blog-645749475857315124.post-83381967905806695382012-11-04T14:09:27.171-05:002012-11-04T14:09:27.171-05:00I find it more than a little ironic that your dism...I find it more than a little ironic that your dismissiveness of Nate Silver's 538 forecast is very comparable to the objection of the Cornell Basketball Blog to your fine analysis of Ivy basketball. I would summarize both of your criticisms as, "Don't bother me with all of that math and statistics mumbo jumbo. Your conclusions just don't *feel* right. The race is a lot closer than you think it is."<br /><br />It seems that many observers are more comfortable with the approach of data-driven analysis when they like the output. Fortunately, there will soon be an empiricial confirmation or refutation of both of your models. I, for one, expect both you and Nate Silver to be proven correct.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-645749475857315124.post-14375922455853751592012-10-13T12:01:58.365-04:002012-10-13T12:01:58.365-04:00This is really helpful. I have read this and achi...This is really helpful. I have read this and achieved a lot of information.<br /><br /><a href="http://www.assetline.com/en/used-construction-equipment.php" rel="nofollow">Used Roller</a><br />Anonymoushttps://www.blogger.com/profile/16303919845816853713noreply@blogger.comtag:blogger.com,1999:blog-645749475857315124.post-86734734921880307042012-10-05T23:59:18.645-04:002012-10-05T23:59:18.645-04:00Absolutely. I'll try to post that information ...Absolutely. I'll try to post that information here, but if I don't, it will be on the Basketball-U Ivy Message Boards, for certain.<br /><br />I continually re-run the model after each game, gradually phasing out the preseason projections in favor of regular season actuals. But I'll probably highlight where things stand on the eve of league play in a separate piece.<br /><br />The ebb and flow of the rankings and projections is fun to keep an eye on as the season progresses.Michael Jameshttps://www.blogger.com/profile/06461922638043769098noreply@blogger.comtag:blogger.com,1999:blog-645749475857315124.post-49706871933315761552012-10-05T17:44:07.318-04:002012-10-05T17:44:07.318-04:00This is an always interesting read and I think I s...This is an always interesting read and I think I speak for many Ivy fans when I thank you for providing it. I wonder if you would consider running your model again on the eve of conference play in January. There's a lot of new information to be gained before then, especially this year with all the talent losses around the league.<br /><br />In 2010-11, for example, your model liked Princeton and Harvard but, by the time conference play began, the Crimson seemed like a clear favorite, making the Tigers' eventual NCAA bid seem a bit more like overachieving, even though P and H seemed even at season's start.Anonymousnoreply@blogger.com