tag:blogger.com,1999:blog-38600807.post1787524777409167125..comments2023-09-09T10:36:20.114-04:00Comments on Advanced Football Analytics (formerly Advanced NFL Stats): Fantasy Draft Strategy Part 2Unknownnoreply@blogger.comBlogger10125tag:blogger.com,1999:blog-38600807.post-82437355691962273362012-04-02T14:24:33.196-04:002012-04-02T14:24:33.196-04:00How high was your correlation? Because I did the s...How high was your correlation? Because I did the same thing as you (plotting point values v. rank) for the past 3 years and found a much smaller correlation than you seemed to. How many years worth of data did you use? Also what's your regression function?Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-64334785146794848122009-09-02T10:09:00.331-04:002009-09-02T10:09:00.331-04:00what about auction leagues? there you have no draf...what about auction leagues? there you have no draft order just cash to spend.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-22861977757489206712008-08-24T01:16:00.000-04:002008-08-24T01:16:00.000-04:00Yea, it definitely matters if you are in a 1 QB le...Yea, it definitely matters if you are in a 1 QB league as opposed to a 2 QB league. In a 2 QB league, the scarcity measure makes drafting QBs much more important. In this scenario, you can easily justify taking Brady and Peyton over most running backs save for the top 2. This all depends on the scoring system of course, but I have found this generally to be true.Nitinhttps://www.blogger.com/profile/01490190407967107464noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-20760650970467074592008-08-21T16:04:00.000-04:002008-08-21T16:04:00.000-04:00Does the # of roster spots matter? With only 1 QB...Does the # of roster spots matter? With only 1 QB and TE spot for most formats, who cares about QB and TE past rank 20? However, RB and WR in that range really matter. I'd guess TE is still better than WR, but RB jumps even more important early.<BR/><BR/>BTW, great articles. Really got me thinking.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-29959273384400101362008-08-19T11:04:00.000-04:002008-08-19T11:04:00.000-04:00i love this articlei love this articleAnonymousnoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-5352290783939472092008-08-18T17:21:00.000-04:002008-08-18T17:21:00.000-04:00My guess is that correlations are higher as you ex...My guess is that correlations are higher as you expand the number of WRs included because a large source of variability is TDs. Third and 4th WRs have low TD totals that don't vary much. Perhaps a look at consistency of yardage vs. consistency of TDs(which I'm guessing is low) would be useful.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-75066879803642089542008-08-18T16:48:00.000-04:002008-08-18T16:48:00.000-04:00Actually, I plotted year n Rank vs. year n-1 fanta...Actually, I plotted year n Rank vs. year n-1 fantasy point production, but the information I'm trying to glean is the same - I'm trying to see how consistent fantasy production is by rank. The reason I did it this way is because this is an approximation of how people make their fantasy drafting decisions (which RB should I take first? The RB with the highest production last year...etc). Thus, I found that, on average from 1988 to 2007, the difference in REALIZED value from the PROJECTED #1 TE and the REALIZED value of the PROJECTED #2 TE (I'll call that "slope at #1") is about the same as slope at #1 for WRs. In fact, the slope at almost all points are pretty equal. I think this is saying that, although ex post it would be much better to pick the best TE than the best WR, ex ante you have no way of knowing who that will be, whereas you might have a better idea who the best WR will be.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-38600807.post-48610844496259386832008-08-18T16:28:00.000-04:002008-08-18T16:28:00.000-04:00Now I think I get what you did. I had to think abo...Now I think I get what you did. I had to think about it a little longer. You plotted Yr n+1 fantasy performance vs. previous yr rank. That's clever. That way you could see how consistency levels change throughout the depth chart of the league.Brian Burkehttps://www.blogger.com/profile/12371470711365236987noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-17803730703166624252008-08-18T16:21:00.000-04:002008-08-18T16:21:00.000-04:00So you did the top 10 WRs instead of the top 32, o...So you did the top 10 WRs instead of the top 32, or something similar? And they were as consistent as TE's? I used points rather than rank, but that's interesting, because the larger I made the samples for WRs, the higher the correlation. Top 64 was higher than for top 32, and top 96 was even higher. <BR/><BR/>I don't see why you say I'd need to know the outcome of players' stats before the season. It sounds like we did the same thing, I just used points and you used rank.<BR/><BR/>Perhaps I have confused the 2 main points of my article. The analysis was just to get a feel for how projections should be made, or how they should be weighted according to scarcity and consistency considerations. I'm just quantifying what many people already know intuitively. <BR/><BR/>The analysis led me to a system, which is the second main point--the opportunity cost system. You can use whatever projections you want under that system. I wouldn't suggest using a purely linear regression of last year's stats as a projection for this year.<BR/>(But it is an excellent starting point.) <BR/><BR/>My next post will illustrate exactly what I did and give an example of how it works.Brian Burkehttps://www.blogger.com/profile/12371470711365236987noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-26817458054102534762008-08-18T15:45:00.000-04:002008-08-18T15:45:00.000-04:00Brian - This is a very interesting analysis - howe...Brian - This is a very interesting analysis - however, you are assuming that we know what the stats of each player will be at the beginning of the season. I charted the Position Rank (ex post) vs. the average of those players fantasy production for the year before (as a proxy for "projected" stats). Basically it shows the same as your graph, except that the slopes of the WR and TE are almost the same. I think this is explained by the fact that, while WR stats are not very auto-correlated overall, "top" WRs tend to be more a lot more consistent than average WRs and about as consistent as the top TEs. I can send you the spreadsheet if you are interested.Anonymousnoreply@blogger.com