Topic: What the hell happend!!!!

[greenblazes]

Olen is dead on. The computer numbers were figured based on a total possible 500 points instead of 100 the bcs uses (because it's only designed to rank 25 teams). But they didn't also tweak % of points methodology for the polls so therefore, it almost totally eliminated any effect from votes in coaches and harris polls had on final calculation. Basically, they have a better computer average.

Palm's rankings are pretty much worthless past the teams who recd votes. After looking at this for past 2 hours, I can't believe cbs actually publishes his "rankings."  He just takes all teams who recd any votes and ranks them in order based on the bcs formula. Then he takes remainder who recd no votes and puts them in order behind the "vote getters" based on computer average.

[_sturt_]

The weighting is catty-wampus - MU was 29th out of 125 in the Coaches Poll, and 31st out of 125 in the Harris Poll, and probably mid-60s on the computer average, Rice was shut out of both polls and had a 52/125 computer average, but based on CUSA's formula, Rice rates higher than Marshall.  It doesn't add up.

See the problems?  The retort is that the coaches and Harris polls don't rank each position 1-125, while the computers do -- but that is the big problem, there is not a one-to-one correlation, and therefore the weighting is thrown off.  MU's 29th and 31st positions out distances Rice's best computer ratings by at least 12-15 positions (I think), yet Rice's computer ratings accounted for more. 

Screwy.

See, there's where... I think... we just don't have an adequate way of knowing how they treated that... you're supposing 12-15 positions, but it's a supposition. My guess would be that the 29 and 31 aren't actually the numbers employed, but rather they're inputting share of total number of votes. In other words, #1 and #2 aren't simply ordinal 1 and 2, but rather, the number inputted into the formula for the number one team is their share of total votes.

[_sturt_]

Olen is dead on. The computer numbers were figured based on a total possible 500 points instead of 100 the bcs uses (because it's only designed to rank 25 teams). But they didn't also tweak % of points methodology for the polls so therefore, it almost totally eliminated any effect from votes in coaches and harris polls had on final calculation. Basically, they have a better computer average.

Um, well... if they're ranking 125 teams instead of 25, then it would make sense, would it not, that they'd figure at 500 possible instead of 100? (i.e., 25 is 20% of 125, as 100 is 20% of 500)

What do you mean by "didn't also tweak % of points methodology for the polls"...? Not following you.

[Olen]

See, there's where... I think... we just don't have an adequate way of knowing how they treated that... you're supposing 12-15 positions, but it's a supposition. My guess would be that the 29 and 31 aren't actually the numbers employed, but rather they're inputting share of total number of votes. In other words, #1 and #2 aren't simply ordinal 1 and 2, but rather, the number inputted into the formula for the number one team is their share of total votes.

I was close -- Rice's four computer ratings (after tossing high/low) are:  43, 45, 49, and 56 (an average of 48.25).

The difference between MU's poll rankings and Rice's computer ratings are:  +14, +16, +20, and +27; +12, +14, +18, and +25, or an average difference of +19 and +17.

Your statement highlights the broader point - once CUSA expanded the computer rating beyond the BCS formula, there is no appropriate way to maintain the weighting between the human polls and the computer rating averages, or at least not one I have figured out.

[greenblazes]

Um, well... if they're ranking 125 teams instead of 25, then it would make sense, would it not, that they'd figure at 500 possible instead of 100? (i.e., 25 is 20% of 125, as 100 is 20% of 500)

What do you mean by "didn't also tweak % of points methodology for the polls"...? Not following you.

The bcs attributes points for CPU averages the same as the human polls. A 1st place ranking = 25 points, 2nd = 24 and so on. (A ranking of 25th = 1 point). Since you only use 4 averages, you take the total and divide by 100. This gives you your total % points and is your 1/3 number to be averaged with coaches and Harris. The bcs was only designed to rank 25 teams so a ranking past 25 receives no points. Thus to rank teams all the way down to 125, you have to change the methodology. Cusa it appears gave points from 125-1 based on ranking and changed the divider to 500 (4 total rankings counted x 125 for a 1st place vote). Sounds reasonable up to that point.

However, the system breaks down if you don't also adjust for the coaches and Harris. Again (using coaches as example) teams get points from 25-1 based on where they are voted. With 62 voters, there is a total possible of 1550 point. Since we recd 13 points, that's where they get the 0.0084 number (13/1550).  For it to still be a true 1/3 weight for each, you would have to attribute points like above from 125-1 and then divide by 7750 which would be total number of points after tweaking (125 x 62 total voters).

Example, we got 13 points so say 13 voters out of the 62 voters put us at #25. That would (under a tweaked methodology equal to how they tweaked the CPU methodology) mean we would have 1300 total points. Divide that by 7750 which would be the total possible points and you have 0.168 % of total points (obviously much higher than the 0.0084 they factored in. It's impossible to accurately calculate b/c we don't know 2 things: 1. How many voters voted for us and 2. Where the put us.  If you had that info, you could tweak the human poll averages and then truly give all 3 1/3 equal weighting.

Kind of confusing and maybe rambling. Bottom line is that with cusa' method, the human polls basically counted for squat and all that mattered was the CPU average which rice clearly was higher.

[HaveUherd]

The human polls are weighted fairly. We just needed more votes for it to make a difference for us to offset the computers. The unfair thing to me is not being able to use margin of victory for the computer rankings. That's a bigger argument than this though.

[field pass]

Wow...the fact that Sagrin has Rice ranked higher suprises me.  I guess our schedule strength is suspect slightly.

[herdadreh]

Well said Olen, here on this thread and elsewhere.

CUSA - "We don't want to overweight the poll results since Rice didn't get poll votes. (However, apparently we are perfectly OK with overweighting the computer results.)"

Whether intentional or not (whether they began with the end - a Rice home game - in mind or not), this is just selective points-adjustment. There is no mathematical, statistical, logical legitimacy for what was done.

Arbitrary and possibly capricious.

I mean, this is the kind of stuff of which lawyers dream. I think Marshall would win a (very) temporary injunction at the very least. Not saying I want them to file for one. There is a serious time constraint on the remedy.