(09-21-2020 05:52 PM)jedclampett Wrote:  Here are the top 10 ranked teams in the current Massey composite rankings, which includes 3 B1G teams:

#8 UCF

#13 Memphis

#14 Cincinnati

#16 BYU

NOTE: #18 OHIO STATE (B1G)

#22 Appalachian State

NOTE: #23 (tie) PENN STATE (B1G)

#23 Louisiana

NOTE: #25 WISCONSIN (B1G)

#31 SMU

NOTE: #31 (tie) WISCONSIN (B1G)

NOTE: #33 (tie) IOWA (B1G)

#35 Marshall

#39 Army

NOTE: #40 (tie) MINNESOTA (B1G)

#40 FAU

...........................................

Next Six:

#42 Arkansas State

#43 Houston

#44 Navy

#46 Temple

#49 Tulane

#49 Louisiana Tech

(09-21-2020 06:32 PM)IWokeUpLikeThis Wrote:  Using Massey Composite when just 41 teams have played a game.

(09-21-2020 07:36 PM)jedclampett Wrote:  

(09-21-2020 06:32 PM)IWokeUpLikeThis Wrote:  Using Massey Composite when just 41 teams have played a game.

How many weeks would you reccommend waiting until checking the Massey Composite?

(09-21-2020 06:19 PM)quo vadis Wrote:  Computer rankings don't mean anything before around week 6 of the season because there are too few data points, so computers just fill in the gaps with last year's data, which is of course irrelevant.

And computers will be particularly useless this year because of so few OOC games being played.

(09-21-2020 07:36 PM)jedclampett Wrote:  

(09-21-2020 06:32 PM)IWokeUpLikeThis Wrote:  Using Massey Composite when just 41 teams have played a game.

How many weeks would you reccommend waiting until checking the Massey Composite?

(09-22-2020 08:58 AM)quo vadis Wrote:  

(09-21-2020 07:36 PM)jedclampett Wrote:  

(09-21-2020 06:32 PM)IWokeUpLikeThis Wrote:  Using Massey Composite when just 41 teams have played a game.

How many weeks would you reccommend waiting until checking the Massey Composite?

Even in a normal year, no computers are worth looking at before week 6 or so, that's because data points are too limited so in order to post something, they fill in the gaps with last year's data. I mean, how can Alabama be #3 when they haven't played a game?

And this year, the MC will be almost entirely useless, because of the very limited amount of OOC games. The SEC and B1G are playing none, while Clemson played The Citadel. What on earth can be learned from that? Nothing.

Granted, there are some web-interconnections. The SB laying the wood to the Big 12 could hurt the rankings of the top Big 12 teams, and rightly so. And the G5 conferences have played a number of OOC games. But that's about it.

No, this is the year for purely subjective "eye test" evaluations.

(09-22-2020 09:18 AM)ken d Wrote:  

(09-22-2020 08:58 AM)quo vadis Wrote:  

(09-21-2020 07:36 PM)jedclampett Wrote:  

(09-21-2020 06:32 PM)IWokeUpLikeThis Wrote:  Using Massey Composite when just 41 teams have played a game.

How many weeks would you reccommend waiting until checking the Massey Composite?

Even in a normal year, no computers are worth looking at before week 6 or so, that's because data points are too limited so in order to post something, they fill in the gaps with last year's data. I mean, how can Alabama be #3 when they haven't played a game?

And this year, the MC will be almost entirely useless, because of the very limited amount of OOC games. The SEC and B1G are playing none, while Clemson played The Citadel. What on earth can be learned from that? Nothing.

Granted, there are some web-interconnections. The SB laying the wood to the Big 12 could hurt the rankings of the top Big 12 teams, and rightly so. And the G5 conferences have played a number of OOC games. But that's about it.

No, this is the year for purely subjective "eye test" evaluations.

And that is the biggest problem with all computer algorithms. Kansas losing to Coastal Carolina really should't affect the rating for Oklahoma at all.

(09-22-2020 09:31 AM)quo vadis Wrote:  

(09-22-2020 09:18 AM)ken d Wrote:  

(09-22-2020 08:58 AM)quo vadis Wrote:  

(09-21-2020 07:36 PM)jedclampett Wrote:  

(09-21-2020 06:32 PM)IWokeUpLikeThis Wrote:  Using Massey Composite when just 41 teams have played a game.

How many weeks would you reccommend waiting until checking the Massey Composite?

Even in a normal year, no computers are worth looking at before week 6 or so, that's because data points are too limited so in order to post something, they fill in the gaps with last year's data. I mean, how can Alabama be #3 when they haven't played a game?

And this year, the MC will be almost entirely useless, because of the very limited amount of OOC games. The SEC and B1G are playing none, while Clemson played The Citadel. What on earth can be learned from that? Nothing.

Granted, there are some web-interconnections. The SB laying the wood to the Big 12 could hurt the rankings of the top Big 12 teams, and rightly so. And the G5 conferences have played a number of OOC games. But that's about it.

No, this is the year for purely subjective "eye test" evaluations.

And that is the biggest problem with all computer algorithms. Kansas losing to Coastal Carolina really should't affect the rating for Oklahoma at all.

It operates on the X played Y who played Z who played R and therefore we know something about how to relate X and R principle. Now does that linkage tell us that X would beat R or vice-versa? No, but it tells us *something*, and so i think it makes sense for computers to use that information.

Of course, no computer is perfect. Heck we don't even know what perfect means. They reflect the beliefs of their programmers. That's why sometimes a single computer produces what most regard as a screwy result. I mean last year, there was a computer that ranked LSU #4 overall, behind Ohio State, Clemson, and Georgia, even though LSU had a better record than Clemson and Georgia and beat both head to head. That's why it's best to look at multiple computers and when you do that, the rankings tend to pass the sniff test.

(09-22-2020 10:26 AM)ken d Wrote:  

(09-22-2020 09:31 AM)quo vadis Wrote:  

(09-22-2020 09:18 AM)ken d Wrote:  

(09-22-2020 08:58 AM)quo vadis Wrote:  

(09-21-2020 07:36 PM)jedclampett Wrote:  How many weeks would you reccommend waiting until checking the Massey Composite?

Even in a normal year, no computers are worth looking at before week 6 or so, that's because data points are too limited so in order to post something, they fill in the gaps with last year's data. I mean, how can Alabama be #3 when they haven't played a game?

And this year, the MC will be almost entirely useless, because of the very limited amount of OOC games. The SEC and B1G are playing none, while Clemson played The Citadel. What on earth can be learned from that? Nothing.

Granted, there are some web-interconnections. The SB laying the wood to the Big 12 could hurt the rankings of the top Big 12 teams, and rightly so. And the G5 conferences have played a number of OOC games. But that's about it.

No, this is the year for purely subjective "eye test" evaluations.

And that is the biggest problem with all computer algorithms. Kansas losing to Coastal Carolina really should't affect the rating for Oklahoma at all.

It operates on the X played Y who played Z who played R and therefore we know something about how to relate X and R principle. Now does that linkage tell us that X would beat R or vice-versa? No, but it tells us *something*, and so i think it makes sense for computers to use that information.

Of course, no computer is perfect. Heck we don't even know what perfect means. They reflect the beliefs of their programmers. That's why sometimes a single computer produces what most regard as a screwy result. I mean last year, there was a computer that ranked LSU #4 overall, behind Ohio State, Clemson, and Georgia, even though LSU had a better record than Clemson and Georgia and beat both head to head. That's why it's best to look at multiple computers and when you do that, the rankings tend to pass the sniff test.

And overwhelmingly, those composite rankings look very much like the 3 human polls (AP, Coaches and CFP). So, why do we need them, other than to have a larger sample from which fans can cherry pick the results that "prove" why their team is better than your team?

(09-22-2020 10:47 AM)quo vadis Wrote:  

(09-22-2020 10:26 AM)ken d Wrote:  

(09-22-2020 09:31 AM)quo vadis Wrote:  

(09-22-2020 09:18 AM)ken d Wrote:  

(09-22-2020 08:58 AM)quo vadis Wrote:  Even in a normal year, no computers are worth looking at before week 6 or so, that's because data points are too limited so in order to post something, they fill in the gaps with last year's data. I mean, how can Alabama be #3 when they haven't played a game?

And this year, the MC will be almost entirely useless, because of the very limited amount of OOC games. The SEC and B1G are playing none, while Clemson played The Citadel. What on earth can be learned from that? Nothing.

Granted, there are some web-interconnections. The SB laying the wood to the Big 12 could hurt the rankings of the top Big 12 teams, and rightly so. And the G5 conferences have played a number of OOC games. But that's about it.

No, this is the year for purely subjective "eye test" evaluations.

And that is the biggest problem with all computer algorithms. Kansas losing to Coastal Carolina really should't affect the rating for Oklahoma at all.

It operates on the X played Y who played Z who played R and therefore we know something about how to relate X and R principle. Now does that linkage tell us that X would beat R or vice-versa? No, but it tells us *something*, and so i think it makes sense for computers to use that information.

Of course, no computer is perfect. Heck we don't even know what perfect means. They reflect the beliefs of their programmers. That's why sometimes a single computer produces what most regard as a screwy result. I mean last year, there was a computer that ranked LSU #4 overall, behind Ohio State, Clemson, and Georgia, even though LSU had a better record than Clemson and Georgia and beat both head to head. That's why it's best to look at multiple computers and when you do that, the rankings tend to pass the sniff test.

And overwhelmingly, those composite rankings look very much like the 3 human polls (AP, Coaches and CFP). So, why do we need them, other than to have a larger sample from which fans can cherry pick the results that "prove" why their team is better than your team?

Bottom line is, all collectively unbiased systems are likely to produce similar results, because this isn't rocket science. You don't need a degree from Harvard to be able to tell that Clemson is better than Louisiana-Monroe.

As for "needing" the computers, well I guess it was decided in 2012 that we don't, because the CFP system did away with them. Personally though, I like them, and I suspect the CFP members tend to rely on them too.

(09-22-2020 10:57 AM)ken d Wrote:  

(09-22-2020 10:47 AM)quo vadis Wrote:  

(09-22-2020 10:26 AM)ken d Wrote:  

(09-22-2020 09:31 AM)quo vadis Wrote:  

(09-22-2020 09:18 AM)ken d Wrote:  And that is the biggest problem with all computer algorithms. Kansas losing to Coastal Carolina really should't affect the rating for Oklahoma at all.

It operates on the X played Y who played Z who played R and therefore we know something about how to relate X and R principle. Now does that linkage tell us that X would beat R or vice-versa? No, but it tells us *something*, and so i think it makes sense for computers to use that information.

Of course, no computer is perfect. Heck we don't even know what perfect means. They reflect the beliefs of their programmers. That's why sometimes a single computer produces what most regard as a screwy result. I mean last year, there was a computer that ranked LSU #4 overall, behind Ohio State, Clemson, and Georgia, even though LSU had a better record than Clemson and Georgia and beat both head to head. That's why it's best to look at multiple computers and when you do that, the rankings tend to pass the sniff test.

And overwhelmingly, those composite rankings look very much like the 3 human polls (AP, Coaches and CFP). So, why do we need them, other than to have a larger sample from which fans can cherry pick the results that "prove" why their team is better than your team?

Bottom line is, all collectively unbiased systems are likely to produce similar results, because this isn't rocket science. You don't need a degree from Harvard to be able to tell that Clemson is better than Louisiana-Monroe.

As for "needing" the computers, well I guess it was decided in 2012 that we don't, because the CFP system did away with them. Personally though, I like them, and I suspect the CFP members tend to rely on them too.

What the computers (or hybrids like the Massey Composite) do that the polls don't is rank teams below those who get votes. That lets you approximate which conferences are deeper than others, if that matters to you.