Question for Doc Trivia - Football Ratings - All - Football Ratings - Message Board - Yuku

RSS
Email

0 Kudo

<< Previous Topic Next Topic >>

Question for Doc Trivia

Author	Comment
kycat1a	Question for Doc Trivia	Lead	[-]
Posts: 2 (09/26/06 09:57:42) Reply Quote More My Recent Posts Message Me Connection Blocking Invite Ignore User's Posts	Tags : None Doc, I've also been tracking predictions and point spreads on my own this season. I've recently added the average predictions that appear on this site to my analysis. My question is about your previous comments. You seem to find a pattern of games favoring a certain predictor when there is a wide margin between the prediction and the Vegas line. Correct? I've also been tracking the margins of prediction, but have not found such a correlation. Could you explain again how you're using your analysis to predict winners at such a high degree of accuracy? If the prediction favors a team by 20, while the Vegas line favors the same team by 10, are you saying that the prediction tends to be more accurate? I'm not seeing that in my data, but perhaps I need larger samples. Thanks!
PerformanZ	prediction averages	#1	[-]
Posts: 40 (09/27/06 12:21:35) Reply Quote More My Recent Posts Message Me Connection Blocking Invite Ignore User's Posts	I can't always tell myself from what people are saying exactly what they are doing with the numbers. There was a guy here once that was getting some great numbers and ended up giving me some good publicity but what he was doing was never that clear. My original thought when posting these numbers was to look at (prediction average - line) / standard deviation. If the predictions are an accurate reflection of game outcomes this seemed like a good idea. From a statistical standpoint you would prefer to see this number >2 or <-2 but in practice you get only a handfull of games each week that give values >1. And in the NFL the prediciton average is very close to the line by the end of the season. I have had decent results posting these games on my weblog page until last season being the first they didn't really pan out. I may start posting these game again just to see what happens it's been a mixed bag so far this season. And you might even expect that because there is a good change the reason the line is different than the computer predicitons is because the line includes information the average computer system does not have.
kycat1a	interesting	#2	[-]
Posts: 2 (09/28/06 10:12:30) Reply Quote More My Recent Posts Message Me Connection Blocking Invite Ignore User's Posts	Thanks for your insight there. This is a great exercise. A couple questions: 1) Is the prediction average listed now an average of all the systems listed on the page? 2) Does that average also include the Vegas line number? A comment: I've gone to a couple of the actual sites you pull data from and done some of the calculations via their instructions, and get a different number than you have listed as their prediction. Thompson STATS, for example. Are your scripts up to date? I keep track of the margins of prediction for a few of them, trying to find a pattern. My original hypothesis was something like: True score = some weighted average of Vegas - Predictor for games where the difference between the predictions was quite large. In the fraction you just listed, would you use the std. deviation of the prediction average in the denominator, or the standard deviation of all the (line- predictor) sums?
PerformanZ	Re: interesting	#3	[-]
Posts: 40 (09/29/06 06:39:33) Reply Quote More My Recent Posts Message Me Connection Blocking Invite Ignore User's Posts	The average listed is not the average of all the systems listed. There are a few systems that I do not include because I feel that they are not really in tune with where the predictions should be. The average does not include the vegas line. Hopefully the predictions on my page are correct. If you thing some are off let me know. Thompson for example sends me his predictions directly each week so I use them rather than his ratings. Dunkel is another example where the predictions on his own page do not match his ratings. I would use the std listed next to the average as the denominator. it is the standard devation of the average predicition.

<< Previous Topic Next Topic >>

Quick Reply