MSU loses at home to Little Rock on opening night. Yikes...
Announcement
Collapse
No announcement yet.
*2019-20 other games of interest*
Collapse
X
-
Originally posted by DownGoesKansas View Post
A loss that won't impact our RPI! Great news!
With Google Analytics AI, sometimes we legitimately don't know how the computer calculates what it knows.
For instance, a Google computer has learned to differentiate male and female retinas. And we don't know what it is seeing that tells us gender from retina photos.
Comment
-
Originally posted by sivert View Post
Do we know for sure that D-2 games don't get figured into NET?
With Google Analytics AI, sometimes we legitimately don't know how the computer calculates what it knows.
For instance, a Google computer has learned to differentiate male and female retinas. And we don't know what it is seeing that tells us gender from retina photos.
Why? Because you'd need to build every D2 into the system to account for how much that game should matter/impact the rank. On top of that should a D2 play a D3 or NAIA you'd need to build the NAIA and D3 into the system as well.
Hell, an NAIA beat Hampton last night, so the NAIA would have to be built in as well.
That's simply far more effort that is worth, especially for the selection committee. They are far better off making sure it's a highlighted note on the team sheet is far more useful. Afterall, a loss to a sub D1 is bad but is that loss in November the first game of the season still meaningful in March?
Comment
-
Originally posted by clenz View Post...That's simply far more effort that is worth, especially for the selection committee. ...
I have been suspicious that the obscurity in the description of what is in the NET was little more than, "Trust us, the P5 teams are the best." But the more I learn about how AI works, I am more leaning towards suspecting that the implementers and coders simply don't know what is in it.
Comment
-
Originally posted by sivert View Post
Normally, yes, but the data for D2 and D3 games are available from the NCAA. So while it would take people a LOT of effort, it would take the Google computers almost no time to include all of it.
I have been suspicious that the obscurity in the description of what is in the NET was little more than, "Trust us, the P5 teams are the best." But the more I learn about how AI works, I am more leaning towards suspecting that the implementers and coders simply don't know what is in it.
You think it's worth the effort to build/code another 1,0005 schools (D2, D3, NAIA 1&2) for the .0001% of the time a loss happens to a sub D1 in a regular season game?
Comment
-
Originally posted by sivert View Post"Is it worth it?" isn't really a relevant question. If the data is available to the AI software, then it decides what data is helpful. And it will do it without regard to effort or difficulty.
Honestly, .0001 might actually be high.
Last night there was 150 games. 2 of them resulted in a sub D1 win - that's .01%
Over the course of an entire OOC schedule which involves something like 3,500 OOC games, even if we keep the .01 as the mark that it happens at that is 35 games impacted. The reality is we might...maybe...maybe...get 10 total, which is .002% of games.
So no, it's not worth the time it takes to code them in.Last edited by clenz; 11-06-2019, 10:55 AM.
Comment
-
Originally posted by clenz View Post
Last night there was 150 games. 2 of them resulted in a sub D1 win - that's .01%
Won't matter too much because if GCU is getting beat by D2 schools, they're going to drop to plenty of D1 schools as well.
Comment
-
Originally posted by sivert View PostRedbirds got a nice win over Belmont tonight.
Comment
Comment