New changes to the Nerd engine

New changes to the Nerd engine

September 10, 2008 4 Total Comments

I've disabled a number of the sites that the Nerd searches based upon their performance projections from last week. I'll continue watching their projections and can very easily add them back. There's one more that I may disable depending on how it's projections go this week. I've reviewed today's search index and they're not that far off so I'll give them another week.

This is also the first week where the individual source projections get weighted based on last week's performance. This will continue throughout the season.

I had to do a double take when I saw today's search index results. Seattle's Defense ranked #1? From 15 last week to number 1? Against Gore and the remaining 49'ers? I think Seattle will have a good week on D, but I'm not sure I would rank them #1.

As I looked at the individual source projections, I noticed that FOX ranks Seattle at #3 and CBS ranks them #1. CBS projects the Seattle D to have 2 interceptions, 1 forced fumble, 5 sacks, 1 defensive TD, and 368 total yards allowed. FOX projects 3 sacks, 1 interception, 0 TD's, and 293 yards allowed.

The other sites weren't that far off, all listing Seattle in the Top 8. So...Seattle it is. I'll be watching tomorrow to see how often changes occur during the week leading up to Sunday.

Comments

great site! but why hasn't colston disappeared from the rankings yet?

Submitted by VitaminJ on September 11, 2008

Thanks VitaminJ. Garbage in...garbage out. I noticed that as well and am working on some more sanitizing of the data. I have a current list of injuries that I'm working on and will be bringing in shortly. More to follow!

Submitted by Nerd on September 12, 2008

Removing sites for one week of bad projections seems drastic and seems to me like it could be counterproductive if the site just happened to have a bad week. I realize you only temporarily remove inaccurate sites until they prove themselves worthy, but I would guess that Week 1 projections are, in general, wildly inaccurate. What about introducing a weighting system were the more accurate sites are more heavily weighted in the overall ranking than less accurate sites. As the year-to-date accuracy changes per site the weightings could and should change.

Submitted by bigdrew34 on September 12, 2008

Never mind my previous comment. I just re-read the post and you appear to have already implemented exactly what I just suggested


Submitted by bigdrew34 on September 12, 2008

Add Your Comments

You must be signed in to post your comments