Stat Allocation key?Question for all

[carwash 0]

Just curious as to what everyone uses for the stat allocation, FFtoday, or some mix of sites or past years?

 

Thanx

[Mike MacGregor 20]

Ok, I'll start. Honestly I pretty much just use the default projections 100%. The only thing I will do is alter a number of projections right in there if I want to move some players up or down.

 

I figure the default projections already account for all offseason news and prospects of players for this year. Sometimes working in the past stats projected on to this year can skew things.

 

What I will do however is use the past stats at 100% for each year or average a few of them together to see how the points and value shake out based on actual history, rather than projections. It works as a good check on the current year projections, and gives a true assessment of the value of each position based on your scoring. Plus it can help identify players who might be good sleeper prospects for this year but were overlooked by the current year projections.

[zippychi 0]

i have been using the fftoday and fbg at 50% each.

 

i tried using another site but they didnt have enough players projected to make it very helpful.

[ICETIME25 0]

Last year, I used 33% FFToday, 33% CBS and 33% 4for4. This post made me curious about which sites projections would have yielded the most accurate drafts based on actual results versus projected for 2002.

 

Our 12 owner league starts 1 QB, 2 RB, 2 WR, 1 TE, 1 K, 1 D and drafts 15 players. My analysis evaluated each sites ability to find the stud (i.e. top 3 QB/TE/K/D and top 6 RB/WR), find the top half starters (i.e. top 6 QB/TE/K/D and top 12 RB/WR), and finally find my viable staters (i.e. top 12 QB/TE/K/D and top 24 RB/WR).

 

Here are the results:

 

4for4 yielded the best results with:

8 of 24 studs (33%)

26 of 48 top half (54%)

65 of 96 starters (68%)

 

FFToday yielded the next best results with:

8 of 24 studs (33%)

24 of 48 top half (50%)

65 of 96 starters (68%)

 

CBS yielded the worst results with:

7 of 24 studs (29%)

21 of 48 top half (44%)

62 of 96 starters (65%)

 

To determine if a combination of the top 2 (4for4 and FFToday) would yield better results, I evaluated projected versus actual with 50% each. The results were:

8 of 24 studs (33%)

24 of 48 top half (50%)

65 of 96 starters (68%)

 

or the same as FFToday alone. I was hoping that the combination of the two would create move value, but it appears that the slightly lesser accuracy of FFToday diluted the combination of the two.

 

What I found interesting about this analysis is that even with the best tools and best projections, predicting optimal roster configurations is very difficult.

 

Paying top dollar (auctions) or using a high draft picks for Kurt Warner, Ahman Green, David Boston had devastating results for their owners, while missing out on value sleepers like Drew Bledsoe, Clinton Portis and Hines Ward.

 

In the end, it appears that diversity and depth with low risk for studs and a flyer on a couple of sleepers might yield the optimal draft. (Sorry for overstating the obvious).

As such, I won't be overpaying for 1-2 key players and risk the lack of depth needed to compensate for a 46% chance I don't have a top tier player or a 34% chance my cheat sheets don't identify a solid starter.

 

Just one "geeks" perspective....

:cool:

 

If anyone else has found a site (free or pay)that yields better results, I would appreciate the tip!!!

[Stockyard 0]

Ice, etc-

 

Last year I used FFtoday, Huddle, and Tuffs. I think Tuffs went under as I can't find them this year. I anticipate using football guys as a replacement this year. Huddle is a pay service, but as far as I can tell, Football guys projections are free.

 

We run our league on CBSsportsline and although I am very pleased with their site, I have never liked their commentary or projections (IMHO).

 

The problem with averaging (as stated above) is that it dilutes the numbers. My hope is that it will lower some of the crazy projections, and the agreeable solid players will rise to the top. No one set of projections is perfect from year to year so I don't like to put all my eggs in one basket.

 

Good luck this year.

Stock

[Stockyard 0]

Update from previous post--

 

I started loading some QB data yesterday to get my process down. I realized that the Huddle stats don't include att/comp so if your scoring system uses more data, you may need to look somewhere else. Also Huddle only supplies data on about the top 45 QBs compared to 65+ with Football Guys and even more in FFToday. I imagine its the same for other positions. The missing player stats can mess up your projections if you don't add something to fill the holes. I wind up copying over the stats from FFToday. All the missing players are from the bottom of the list anyway, so it shouldn't matter too much unless you are in a huge league.

 

Later,

Stock