Rules:

Rule 1.
Please use a "*" or "nm" in the subject line to denote topic only posts.
Any post or attempt to take things "off the board" will result in a ban. Do not use other poster's real names or personal info.
Don't be a Tom Izzo.

Admins - Rich, BBA, Nikos
Ceremonial Moderator - the wino
Moderator Emeritus - Beave
Honorary Moderators - BigLake, Blue Man

POTY - Craig

📌 🏈 Michigan at Minnesota*
📌 To register, email victardgoblue@gmail *

The Victors

MGoBlog || UMHoops || BB90 || Scores || Cards || Chat || HOF || Games

Post a New Message

Home | Search | Login | Register | Settings

Yes, he tests that constantly.

Posted on August 17, 2026 at 01:48:40 PM by GritsGoBlue

One thing that he has found (and this also goes to gbmcq's point) is that keeping the preseason predictions in the mix until around week eight yields a more predictive formula than just going off of what is happening in the season itself. The preseason predictions rely on track record and talent on hand, which retain a lot of value for much of the season before they are ultimately phased out.

Back to the board

Thread:

Is there any back testing done on SP+ projections over the years?* - By Daytripper August 17, 2026 at 12:42:51 PM
- Yes, he tests that constantly. - By GritsGoBlue August 17, 2026 at 01:48:40 PM
  - interesting ty* - By Daytripper August 17, 2026 at 02:00:24 PM
- My opinion, it's still one of my favorite data metrics but it's borderline pointless until he has current season data.* - By gbmcq August 17, 2026 at 12:43:52 PM
  - The test is whether his projections are relatively close to the betting lines, as the sites with skin in the game have a strong interest in assessing team strength accurately. - By GritsGoBlue August 17, 2026 at 01:59:01 PM
    - last season he was wondering how SP+ predictions would fare in light of much higher degrees of roster turnover than we had seen previously - By DoctorD August 17, 2026 at 06:42:11 PM
  - Uncertainty lurks in any sort of preseason forecast. (nm) - By suede August 17, 2026 at 12:47:09 PM
    - He relies a lot on returning production and I don't think his formulas properly factor the transfer portal yet. He includes those players obviously, but I don't think Marsh returning at Michigan is the same as if Marsh had transferred to USC for instance.* - By gbmcq August 17, 2026 at 01:20:58 PM
      - I dunno, it seems pretty high on Indiana, TTU etc. teams without the HS composite talent but tons of successful portals * - By Nikos August 17, 2026 at 01:22:19 PM
        
        I'm not saying he doesn't include production from transfer portal, I just don't think it's properly weighted or I think it's harder to project.* - By gbmcq August 17, 2026 at 01:27:16 PM
    - PECOTA used to do the percentile projections and that's helpful like you had on yours. Like 20% of the time we're projected to win 10 and so on* - By Daytripper August 17, 2026 at 12:58:51 PM
  - Confirms my bias* - By Daytripper August 17, 2026 at 12:46:00 PM
  - I agree with this.* - By AuntLilly August 17, 2026 at 12:45:58 PM

Log in to reply.

Probation ends …