Re: NCAA Selection Discussion - Containment Thread
Posted: Wed Apr 19, 2023 3:17 pm
Same Party, Different House
https://fanlax.com/forum/
Hopkins was given some tournament bids and high seeds from the mid 2000s through 2019 based on their "SOS" and "good losses."wgdsr wrote: ↑Wed Apr 19, 2023 11:26 amwhen was the last time hopkins was in the nc$$?runrussellrun wrote: ↑Wed Apr 19, 2023 11:19 amArmy's problem is their not Hopkins.wgdsr wrote: ↑Wed Apr 19, 2023 10:32 ammy guy, the entire system is built on ooc. How so ?runrussellrun wrote: ↑Wed Apr 19, 2023 10:21 am Army, at 9-2, 6-0 in the Patriot league, a horrible conference full of horrible teams. (why independents like ACC avoid them, except for maybe Loyola.
If a "system" doesn't take into account OOC, what value is it, in the first place.
Sick world, when shoulders are shrugged, "it IS what it IS", and teams like Army, without the AQ, get left out.
Dudes.....th 80's is over....NO ONE cares about the Tar heels or Cuse/Hopkins juice.
again, if eye balls and tailgates in bucolic Connecticutt stadium lots ain't happening when High Point, Army, Maryland & Utah are playing.......we have a product problem.
Quinnapiac !
Anyone wanna guess the attendance for n$aa Frozen Four.
fun fact... every single conference has the exact same intra-conference record (%) as every other one. yeah, and
army's problem for ages has been their ooc scheduling. they make 'cuse lining up holy cross every year look like child's play.
the how so answer is directly below the question.
Head to head is part of the criteria. As is results against common opponents.Chousnake wrote: ↑Wed Apr 19, 2023 6:25 pm I don't think there is a mathematical formula that is going to be determinative and fair when teams play only 12-15 games. There is always going to need to be some subjectivity involved in the process. That is why, when it is close between two teams, there is no better criteria than head to head (when it exists). This is still a sport that is based on winning and losing and the standings and all post season tournament winners are based on winning games, not some algorithm. If team A and B have similar resumes, but team A beat team B, A should get the bid over B. There were years in the past where B got the bid over A based on some of the mathematical criteria we have been discussing, such as SOS, RPI, etc. I'm not sure I understand what the aversion is to settling these situations based on the results on the field rather than a formula.
I'm not saying head to head is the only criteria. And I understand you can get absurd results when you play the A beat B, B beat C, C beat D and D beat A scenario. But when it comes down to two teams and RPI and other criteria are close, the head to head winner should get the nod.
Why not have a formula that is used with the stipulation that if the first team out beat the last team in then the team that won that head to head matchup gets the bid?Chousnake wrote: ↑Wed Apr 19, 2023 6:25 pm
I'm not saying head to head is the only criteria. And I understand you can get absurd results when you play the A beat B, B beat C, C beat D and D beat A scenario. But when it comes down to two teams and RPI and other criteria are close, the head to head winner should get the nod.
are u adding something or "just throwing it out there"? i advocated... heavily... for cornell over umd and hop in 2019... maybe u remember? ultimately, the committee "chose" wins that year. and the ivy league let the big red down. & rpi was all the rage.Chousnake wrote: ↑Wed Apr 19, 2023 6:16 pmHopkins was given some tournament bids and high seeds from the mid 2000s through 2019 based on their "SOS" and "good losses."wgdsr wrote: ↑Wed Apr 19, 2023 11:26 amwhen was the last time hopkins was in the nc$$?runrussellrun wrote: ↑Wed Apr 19, 2023 11:19 amArmy's problem is their not Hopkins.wgdsr wrote: ↑Wed Apr 19, 2023 10:32 ammy guy, the entire system is built on ooc. How so ?runrussellrun wrote: ↑Wed Apr 19, 2023 10:21 am Army, at 9-2, 6-0 in the Patriot league, a horrible conference full of horrible teams. (why independents like ACC avoid them, except for maybe Loyola.
If a "system" doesn't take into account OOC, what value is it, in the first place.
Sick world, when shoulders are shrugged, "it IS what it IS", and teams like Army, without the AQ, get left out.
Dudes.....th 80's is over....NO ONE cares about the Tar heels or Cuse/Hopkins juice.
again, if eye balls and tailgates in bucolic Connecticutt stadium lots ain't happening when High Point, Army, Maryland & Utah are playing.......we have a product problem.
Quinnapiac !
Anyone wanna guess the attendance for n$aa Frozen Four.
fun fact... every single conference has the exact same intra-conference record (%) as every other one. yeah, and
army's problem for ages has been their ooc scheduling. they make 'cuse lining up holy cross every year look like child's play.
the how so answer is directly below the question.
don't think we need it. last year, we had a team get granted a spot on head to head from a home game and liost nearly every other criteriia. seems like we're good.laxreference wrote: ↑Wed Apr 19, 2023 6:42 pmWhy not have a formula that is used with the stipulation that if the first team out beat the last team in then the team that won that head to head matchup gets the bid?Chousnake wrote: ↑Wed Apr 19, 2023 6:25 pm
I'm not saying head to head is the only criteria. And I understand you can get absurd results when you play the A beat B, B beat C, C beat D and D beat A scenario. But when it comes down to two teams and RPI and other criteria are close, the head to head winner should get the nod.
This is why Strength of record is such a powerful metric. Given your SOS, how impressive is your record. And it makes for a much easier comparison because it spits out a single number.
If it doesn't consider which of those teams on your schedule you've beaten and which ones have beaten you, it isn't a "powerful metric." Looking only at aggregates isn't enough.laxreference wrote: ↑Thu Apr 20, 2023 5:46 amThis is why Strength of record is such a powerful metric. Given your SOS, how impressive is your record. And it makes for a much easier comparison because it spits out a single number.
Strenth-of-Record does look at who you beat and who you lost to.Gobigred wrote: ↑Thu Apr 20, 2023 6:50 amIf it doesn't consider which of those teams on your schedule you've beaten and which ones have beaten you, it isn't a "powerful metric." Looking only at aggregates isn't enough.laxreference wrote: ↑Thu Apr 20, 2023 5:46 amThis is why Strength of record is such a powerful metric. Given your SOS, how impressive is your record. And it makes for a much easier comparison because it spits out a single number.
I’m not sure about the “weight” aspect to this. Seems arbitrary. Why not just rank each win and loss straight up?laxreference wrote: ↑Thu Apr 20, 2023 7:28 amStrenth-of-Record does look at who you beat and who you lost to.Gobigred wrote: ↑Thu Apr 20, 2023 6:50 amIf it doesn't consider which of those teams on your schedule you've beaten and which ones have beaten you, it isn't a "powerful metric." Looking only at aggregates isn't enough.laxreference wrote: ↑Thu Apr 20, 2023 5:46 amThis is why Strength of record is such a powerful metric. Given your SOS, how impressive is your record. And it makes for a much easier comparison because it spits out a single number.
Step 1: Assign points for each win based on the RPI (or other metric) of the opponent in reverse order (i.e. 75 points for beating #1)
Step 1a: Assign weight to victories depending on how notable they are; each victory is weighted (say 20%) less than the one before it
Step 2: Assign points for each loss based on the RPI (or other metric) of the opponent (i.e. 75 points for losing to #75)
Step 2a: Assign weight to losses depending on how notable they are; each loss is weighted (say 30%) less than the one before it
Step 3: Sum the points from step 1a (total win points / quality wins factor) and subtract the points from step 2a (total loss points / bad loss factor)
This gives you the total Strength-of-Record (if you use RPI as the input, it's called RPI SOR); it is the simplest method, but there are some ways to adjust it if you want to emphasize a team's best wins vs punish teams for their worst losses. Here's the profile for Cornell's season.
Capture.PNG
Let's say Cornell MLAX had beaten Harvard and lost to Yale instead of the reverse. Without steps 1a and 2a, it wouldn't affect their SOR. I agree with BigRed that any system used for Selection purposes should care who the wins and losses were against.rolldodge wrote: ↑Thu Apr 20, 2023 8:07 amI’m not sure about the “weight” aspect to this. Seems arbitrary. Why not just rank each win and loss straight up?laxreference wrote: ↑Thu Apr 20, 2023 7:28 amStrenth-of-Record does look at who you beat and who you lost to.
Step 1: Assign points for each win based on the RPI (or other metric) of the opponent in reverse order (i.e. 75 points for beating #1)
Step 1a: Assign weight to victories depending on how notable they are; each victory is weighted (say 20%) less than the one before it
Step 2: Assign points for each loss based on the RPI (or other metric) of the opponent (i.e. 75 points for losing to #75)
Step 2a: Assign weight to losses depending on how notable they are; each loss is weighted (say 30%) less than the one before it
Step 3: Sum the points from step 1a (total win points / quality wins factor) and subtract the points from step 2a (total loss points / bad loss factor)
This gives you the total Strength-of-Record (if you use RPI as the input, it's called RPI SOR); it is the simplest method, but there are some ways to adjust it if you want to emphasize a team's best wins vs punish teams for their worst losses. Here's the profile for Cornell's season.
Capture.PNG
It seems that the decay parameter is about time, not about who the win or loss was against. If it’s tied to time, why not make that explicit and set a decay based on actually how long ago the win/loss was? One thing a good system should be is easily grokable by althe average fan.laxreference wrote: ↑Thu Apr 20, 2023 8:15 amLet's say Cornell MLAX had beaten Harvard and lost to Yale instead of the reverse. Without steps 1a and 2a, it wouldn't affect their SOR. I agree with BigRed that any system used for Selection purposes should care who the wins and losses were against.rolldodge wrote: ↑Thu Apr 20, 2023 8:07 amI’m not sure about the “weight” aspect to this. Seems arbitrary. Why not just rank each win and loss straight up?laxreference wrote: ↑Thu Apr 20, 2023 7:28 amStrenth-of-Record does look at who you beat and who you lost to.
Step 1: Assign points for each win based on the RPI (or other metric) of the opponent in reverse order (i.e. 75 points for beating #1)
Step 1a: Assign weight to victories depending on how notable they are; each victory is weighted (say 20%) less than the one before it
Step 2: Assign points for each loss based on the RPI (or other metric) of the opponent (i.e. 75 points for losing to #75)
Step 2a: Assign weight to losses depending on how notable they are; each loss is weighted (say 30%) less than the one before it
Step 3: Sum the points from step 1a (total win points / quality wins factor) and subtract the points from step 2a (total loss points / bad loss factor)
This gives you the total Strength-of-Record (if you use RPI as the input, it's called RPI SOR); it is the simplest method, but there are some ways to adjust it if you want to emphasize a team's best wins vs punish teams for their worst losses. Here's the profile for Cornell's season.
Capture.PNG
But with the decay factors, their bad loss score is improved because the Harvard game has become a victory and their most weighted loss is now Yale. Every team's worst loss is weighted the same, so having a better "worst loss" improves their score.
Their quality wins score would be reduced because you replaced their best win with a less impressive victory, but since the Harvard game is being compared against every other team's 5th best victory, it doesn't go down as much. Net effect = higher RPI SOR.
You can say my 20/30 weights are arbitrary, and you wouldn't be wrong. I've always thought a team's best performances should be more important in NCAA selections than their worst losses, which is why I chose to decay wins more slowly. Other people may think different weightings make more sense.
If you want to see how this works in practice with every team in DI MLAX, you can see the calculation for each team here. You can even set your own weights and see how it affects the rankings.
It's not based on time. You sort the wins by how notable they are (i.e. best teams beaten are ordered first and given the most weight). So the least notable wins get the least weight and count the least toward your quality wins score.rolldodge wrote: ↑Thu Apr 20, 2023 8:32 amIt seems that the decay parameter is about time, not about who the win or loss was against. If it’s tied to time, why not make that explicit and set a decay based on actually how long ago the win/loss was? One thing a good system should be is easily grokable by althe average fan.laxreference wrote: ↑Thu Apr 20, 2023 8:15 amLet's say Cornell MLAX had beaten Harvard and lost to Yale instead of the reverse. Without steps 1a and 2a, it wouldn't affect their SOR. I agree with BigRed that any system used for Selection purposes should care who the wins and losses were against.rolldodge wrote: ↑Thu Apr 20, 2023 8:07 amI’m not sure about the “weight” aspect to this. Seems arbitrary. Why not just rank each win and loss straight up?laxreference wrote: ↑Thu Apr 20, 2023 7:28 amStrenth-of-Record does look at who you beat and who you lost to.
Step 1: Assign points for each win based on the RPI (or other metric) of the opponent in reverse order (i.e. 75 points for beating #1)
Step 1a: Assign weight to victories depending on how notable they are; each victory is weighted (say 20%) less than the one before it
Step 2: Assign points for each loss based on the RPI (or other metric) of the opponent (i.e. 75 points for losing to #75)
Step 2a: Assign weight to losses depending on how notable they are; each loss is weighted (say 30%) less than the one before it
Step 3: Sum the points from step 1a (total win points / quality wins factor) and subtract the points from step 2a (total loss points / bad loss factor)
This gives you the total Strength-of-Record (if you use RPI as the input, it's called RPI SOR); it is the simplest method, but there are some ways to adjust it if you want to emphasize a team's best wins vs punish teams for their worst losses. Here's the profile for Cornell's season.
Capture.PNG
But with the decay factors, their bad loss score is improved because the Harvard game has become a victory and their most weighted loss is now Yale. Every team's worst loss is weighted the same, so having a better "worst loss" improves their score.
Their quality wins score would be reduced because you replaced their best win with a less impressive victory, but since the Harvard game is being compared against every other team's 5th best victory, it doesn't go down as much. Net effect = higher RPI SOR.
You can say my 20/30 weights are arbitrary, and you wouldn't be wrong. I've always thought a team's best performances should be more important in NCAA selections than their worst losses, which is why I chose to decay wins more slowly. Other people may think different weightings make more sense.
If you want to see how this works in practice with every team in DI MLAX, you can see the calculation for each team here. You can even set your own weights and see how it affects the rankings.
Doesn’t the RPI already determine the ordering and weight of the notable wins/losses? I still don’t understand the need for the decay.laxreference wrote: ↑Thu Apr 20, 2023 8:36 amIt's not based on time. You sort the wins by how notable they are (i.e. best teams beaten are ordered first and given the most weight). So the least notable wins get the least weight and count the least toward your quality wins score.rolldodge wrote: ↑Thu Apr 20, 2023 8:32 amIt seems that the decay parameter is about time, not about who the win or loss was against. If it’s tied to time, why not make that explicit and set a decay based on actually how long ago the win/loss was? One thing a good system should be is easily grokable by althe average fan.laxreference wrote: ↑Thu Apr 20, 2023 8:15 amLet's say Cornell MLAX had beaten Harvard and lost to Yale instead of the reverse. Without steps 1a and 2a, it wouldn't affect their SOR. I agree with BigRed that any system used for Selection purposes should care who the wins and losses were against.rolldodge wrote: ↑Thu Apr 20, 2023 8:07 amI’m not sure about the “weight” aspect to this. Seems arbitrary. Why not just rank each win and loss straight up?laxreference wrote: ↑Thu Apr 20, 2023 7:28 amStrenth-of-Record does look at who you beat and who you lost to.
Step 1: Assign points for each win based on the RPI (or other metric) of the opponent in reverse order (i.e. 75 points for beating #1)
Step 1a: Assign weight to victories depending on how notable they are; each victory is weighted (say 20%) less than the one before it
Step 2: Assign points for each loss based on the RPI (or other metric) of the opponent (i.e. 75 points for losing to #75)
Step 2a: Assign weight to losses depending on how notable they are; each loss is weighted (say 30%) less than the one before it
Step 3: Sum the points from step 1a (total win points / quality wins factor) and subtract the points from step 2a (total loss points / bad loss factor)
This gives you the total Strength-of-Record (if you use RPI as the input, it's called RPI SOR); it is the simplest method, but there are some ways to adjust it if you want to emphasize a team's best wins vs punish teams for their worst losses. Here's the profile for Cornell's season.
Capture.PNG
But with the decay factors, their bad loss score is improved because the Harvard game has become a victory and their most weighted loss is now Yale. Every team's worst loss is weighted the same, so having a better "worst loss" improves their score.
Their quality wins score would be reduced because you replaced their best win with a less impressive victory, but since the Harvard game is being compared against every other team's 5th best victory, it doesn't go down as much. Net effect = higher RPI SOR.
You can say my 20/30 weights are arbitrary, and you wouldn't be wrong. I've always thought a team's best performances should be more important in NCAA selections than their worst losses, which is why I chose to decay wins more slowly. Other people may think different weightings make more sense.
If you want to see how this works in practice with every team in DI MLAX, you can see the calculation for each team here. You can even set your own weights and see how it affects the rankings.
In the same way, you sort the losses by how notable they are (so worst teams who beat you come first).
We are already used to comparing teams' best wins vs other teams' best wins and their worst losses against other teams' worst losses, so I assumed that the average fan would be able to grasp it.
What IS the end game of all of this ?wgdsr wrote: ↑Wed Apr 19, 2023 11:09 amcan you name anyone here that likes the system? yesrunrussellrun wrote: ↑Wed Apr 19, 2023 10:45 am Army, as history has shown us based on moronic math, can ONLY make the playoffs if they win the AQ. Huh....
Same for Binghamton, Vermont, Richmond, Utah, Deleware, Drexel, Hampton, UMBC, Sienna, Manhattan....even the Big East is a one bid league, unless the moronic math kicks Denver to the curb.
So, why not try new defensive schemes, like Loyola did, last night. YOU....the ones that love the "system", know this, in your heart of hearts, to be true. Losing by 1, or 11, matters not. Trying different players. Slide packages. Rides. etc.
Even good ole sandbagging.......you AIN'T gonna be invited without that Wonka Bar golden ticket...the AQ. Why not roll with some "different" stuff. Loyola, if they had beaten Cornell, wouldn't have mattered. Outcomes don't matter. results, neither. Just play teams with good records, .....got it.
guess using an rpi formula that only counts game in which you won......IS.....silly and dumb.
Syracuse's rpi wouldn't look so great, now would it, (8-27 = 50% of Cuses rpi, based on the 3 early season wins )
....and, do games against NON-n$aa eligible teams, like Merrimack, Hampton, etc. count for/against your "system". Should they? Weird, Furman didn't have a "waiting period" to gain n$aa entry. How did THAT happen
how are those army wins vs hc, colgate, lafayette and bucknell, not to mention wagner and mercer? how do they do in your system with those 6 (out of 9) w's? all that matters is winning the AQ
'cuse's hc and bonnies games are actually murdering their rpi right now.
about every patriot team's problem besides loyola is every year they already have 3 or 4 tackling dummies in the conference... and then they add 2 or 3 more. you can't do the latter. it's impossible to make a system that works for you if you:
wow, just wow. Colgate a "tackling dummy". Bucknell? Wouldn't part of "understanding the system" include the REALITY that, with rare exceptions, the league/conference you coach in is a ONE bid n$aa likely outcome?
- don't understand the system (should "search committees" have this question at the ready for potential coaching hires? )
- purposely disadvantage yourself (purposely ? dude, your Wahoos have they ever played a regular season game in New England ? )