KevinL Posted December 9, 2003 Share Posted December 9, 2003 I posted this in the USC got screwed thread but it is worth repeating and having it's own thread. I have not heard anyone mention this yet. The BCS is a mathematical formula, where a low score is better (like golf). USC got 1 point from the poll average component, LSU got 2. The prior week USC had 2 and LSU had 3. That is a 1 point difference and whether the points are 1 & 2 or 2 & 3 has no effect. However, looking at the computer rankings, LSU got 1.83 and USC got 2.67. Oklahoma's loss has no effect if Oklahoma is ahead of both or behind both in a computer rank. But the Kenneth Massey poll put Oklahoma in between, with LSU at 1 & USC at 3. Had Oklahoma won and stayed ahead of LSU, that is a 0.17 point difference. LSU's BCS Score was 5.99 and USC's was 6.15, a difference of 0.16 ! Quote Link to comment Share on other sites More sharing options...
Kansas State 2000 Posted December 9, 2003 Share Posted December 9, 2003 Dear LSU, Send check to the following address: Kansas State Athletic Department 35-7 Buttkicking Drive Manhappening, Kansas 66502 You're welcome, Kansas State Quote Link to comment Share on other sites More sharing options...
davearm Posted December 9, 2003 Share Posted December 9, 2003 Great point Kevin. I agree that most folks (myself included) seem to have overlooked this. You can take it a step further though: Even with OU's loss, we had 5 of the 7 computers keep the Sooners #1 (KM had them 2, and NYT had them 5), followed by LSU #2. Had the Kenneth Massey computer followed suit and kept the Sooners 1 and LSU 2, the same outcome would have been achieved: USC 6.15 => Sugar Bowl LSU 6.16 => Pissed Quote Link to comment Share on other sites More sharing options...
davearm Posted December 9, 2003 Share Posted December 9, 2003 Of course the flipside is, if the NYT computer was taken out back and shot -- as it should be, for putting Texas ahead of Oklahoma -- then LSU's margin over USC would be even larger. Quote Link to comment Share on other sites More sharing options...
muck Posted December 9, 2003 Share Posted December 9, 2003 Hey, dave...they drop the lowest score kicked out by a computer... Quote Link to comment Share on other sites More sharing options...
davearm Posted December 10, 2003 Share Posted December 10, 2003 muck: Hey, dave...they drop the lowest score kicked out by a computer... Yes I'm aware of that. NYT is the only of the 7 computers that had USC #1, so if the NYT's output was awarded its rightful spot in the circular file, then USC's final points would be higher, and thus the gap between LSU and USC would be larger than 0.16. Quote Link to comment Share on other sites More sharing options...
KevinL Posted December 10, 2003 Author Share Posted December 10, 2003 I set out to prove Sarge wrong but I learned yet another thing. Let me try to explain this again. The computer ranking component uses 7 computer rankings and takes the average of the best 6 rankings for each team (low score is dropped). Okl: 1,1,1,2,5,1,1 = 7/6 = 1.17 LSU: 2,2,2,1,2,2,2 = 11/6 = 1.83 USC: 3,3,3,3,1,4,3 = 16/6 = 2.67 Difference = 0.84 Now had Oklahoma won the rankings would be Okl: 1,1,1,1,1,1,1 = 6/6 = 1.0 LSU: 2,2,2,2,3,2,2 = 12/6 = 2.0 USC: 3,3,3,3,2,4,3 = 17/6 = 2.83 Difference = 0.83 So even though Oklahoma splitting LSU & USC in the 5th computer ranking gives LSU a 0.17 benefit over USC, LSU's highest ranking (which is dropped) going from 3 to 2 balances it out and the net benefit to LSU on Oklahoma's win is 0.01, not enough to make a difference. LSU would have been in the title game regardless of Oklahoma's result. Quote Link to comment Share on other sites More sharing options...
davearm Posted December 10, 2003 Share Posted December 10, 2003 Yeah it really seems to come down to this: * NYT gave the Trojans a chance by bucking the trend and putting USC ahead of LSU, and * Kenneth Massey took it away by also bucking the trend and putting LSU ahead of OU. Quote Link to comment Share on other sites More sharing options...
davearm Posted December 10, 2003 Share Posted December 10, 2003 Disclaimer: If you're not a numbers geek like me, this post is going to bore you. This whole discussion brings up an interesting issue (maybe -- perhaps we don't even want to go here): there's a whole other unrevealed layer of detail in all of these computer rankings. Naturally, each team is given some sort of numerical score by each computer, and those raw scores are what give rise to the 1-2-3 rankings that plug into the BCS formula. All the public sees are the rankings, not the raw scores. I wonder how close together some of these numbers were for OU, LSU, and USC? It's very conceivable that just like the BCS itself, only fractions of points separated the teams, and any little obscure change could have altered the rankings and swayed the final outcome. While I'm at it here, why not ask this question: what if, instead of churning out rankings, the computers kicked out their raw scores instead, and then the BCS applied a normalization to them? Would that have changed the outcome? It seems logical to believe it very well could have. Quick example to illustrate my point: imagine the davearm supercomputer spits out power rankings, and has: OU: 125 LSU: 115 USC: 110 If we rank based on these scores, we obviously get OU: 1 LSU: 2 USC: 3 If instead we normalize by 100, we get OU: 1.25 LSU: 1.15 USC: 1.10 The teams end up in the same order, but much closer together with the latter method, no? Only 0.15 separate the three, versus a full 2.00 under the ranking method. OK I'm sure that's more than enough on this tangent. Quote Link to comment Share on other sites More sharing options...
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.