I'd like to get some feedback on the current logic for ranking teams for postseason purposes.  This is the ranking shown in the Projection Report, which is used for selecting and seeding for postseason tournaments.  Any changes will impact both the National Tournament and the Postseason Invitational.

The main components are:

• Result (win or loss)
• RPI rank of the opponent
• Top 25 rank of the opponent (this rank is actually calculated and used for all teams, even beyond the top 25)
• Score margin of the game
• Location (home, away, or neutral court)

Maybe not perfect, but I don't know how it could be changed to to make more people comfortable with it than are now.  Leave it alone and spend the time dealing with what seems to be the most importanat issue in the game. Fairness infiring/hiring policy.

aint broke, its fine

much better than the rpi centered scheme it replaced

I agree...a big improvement over the RPI formula

not to be a smartass - I seriously don't know how I can answer some of these without knowing how heavily they are currently weighted...

I dug through the FAQ, release notes and dev chats for info on how the process is explained to work: I'm assuming that the opponent rank and RPI are adjusted overall ranks and not ranks at the time the game was played?

---------

National Tournament selection and seeding both rely on the same logic.

The logic for ranking teams for the postseason essentially calculates a score for each game on a team's schedule. This game score is determined by the following components:

• Result (win or loss)
• RPI rank of the opponent
• Top 25 rank of the opponent (this rank is actually calculated and used for all teams, even beyond the top 25)
• Score margin of the game
• Location (home, away, or neutral court)

There is also consideration given to record over the final 10 games as well as conference tournament performance.

-----------------

11/29/11 dev chat

I'll give a quick recap of the changes coming to tournament selection. The new logic will go through each team's schedule game by game and score the results of each game based on game outcome (win/loss), opponent RPI, opponent rank, score margin, and game location (home/away/neutral). The overall score for a team will be the average of the individual game scores. I like this method better than just putting weights on RPI/rank/etc., because it allows me to give credit for a good win and take away value for a bad loss. For example a bunch of close losses to really good teams will give you more credit now than in the old logic. There will always be debates about which resume is really better, but I think this will be an improvement.

-------------------------

12/7/11 release notes

12/07/2011

• Made the following tweaks to the postseason selection logic: 
  
  - Increased value for winning the game 
  
  - Less weight on opponent rank 
  
  - Less weight on score margin 
  
  - Less weight on last 10 games 
  
  - Less weight on conference tournament results (those games are still included in the normal schedule results evaluation) 
  
  Also, score margin is now being adjusted for the pace of the game, so a 50-30 margin is considered slightly more impressive than a 90-70 margin. 

-------------

So, I get that these factors are the major players, but how much weight is given to each factor? How can I say if too much weight is given to the opponent RPI (for instance) when I don't know if the current weight given is 20% of the overall formula or 40%? Is there a downside to disclosing (at least roughly) the current weights (in that it might allow some to game the system)?

Posted by fd343ny on 5/17/2014 11:26:00 AM (view original):
aint broke, its fine

much better than the rpi centered scheme it replaced

I tend to agree with you, but there is like a 9 page thread about changing PIT selection out there...

I agree there is no doubt that the Postseason Projection Report is better than the RPI based system.

But there is no doubt in my mind that giving slightly more credit for wins will knock out some of those Big 6 bottom feeders with 15 losses, and will make it an even better system.

i said it when the projection report came out and will say it again now - doing these kinds of things well, for the whole gamut of cases, is very difficult. of all the rankings in WIS, this is by a massive margin the best, and i think its very well done, even by my standards.

pretty much every other big change done has some pill you have to swallow, to go along with the good. there's usually a good amount of upset people and often a correction will come out to make people generally happy with the changes. this is the cleanest big change in the last 7 years, as best i can tell, they nailed it on the first shot. is there anything else that went so smoothly?

so, this is probably obvious by now, but i'm not for changing it. i understand the complaint about the last teams in the PIT, but the lines are so muddled in the middle of the pack, the differences from one team to the next is minute. still - outside a few edge cases - i can't recall a single complaint about the ordering there. of course, the human eye is biased against teams with many losses, so there are some edge cases that don't look so good. if anything, add an edge case modifier, teams under .500 get a *small* penalty, something like that - this will even out for teams on the same level (W/L wise) and should have essentially zero effect on the ordering of the rest of the teams in the post season. the worst thing you can do, it doesn't matter if its here or anywhere else in the world of programming, is to break what works the vast majority of the time, to try to fix an edge case.

Posted by dacj501 on 5/17/2014 12:13:00 PM (view original):
not to be a smartass - I seriously don't know how I can answer some of these without knowing how heavily they are currently weighted...

I dug through the FAQ, release notes and dev chats for info on how the process is explained to work: I'm assuming that the opponent rank and RPI are adjusted overall ranks and not ranks at the time the game was played?

---------

National Tournament selection and seeding both rely on the same logic.

The logic for ranking teams for the postseason essentially calculates a score for each game on a team's schedule. This game score is determined by the following components:

• Result (win or loss)
• RPI rank of the opponent
• Top 25 rank of the opponent (this rank is actually calculated and used for all teams, even beyond the top 25)
• Score margin of the game
• Location (home, away, or neutral court)

There is also consideration given to record over the final 10 games as well as conference tournament performance.

-----------------

11/29/11 dev chat

I'll give a quick recap of the changes coming to tournament selection. The new logic will go through each team's schedule game by game and score the results of each game based on game outcome (win/loss), opponent RPI, opponent rank, score margin, and game location (home/away/neutral). The overall score for a team will be the average of the individual game scores. I like this method better than just putting weights on RPI/rank/etc., because it allows me to give credit for a good win and take away value for a bad loss. For example a bunch of close losses to really good teams will give you more credit now than in the old logic. There will always be debates about which resume is really better, but I think this will be an improvement.

-------------------------

12/7/11 release notes

12/07/2011

• Made the following tweaks to the postseason selection logic: 
  
  - Increased value for winning the game 
  
  - Less weight on opponent rank 
  
  - Less weight on score margin 
  
  - Less weight on last 10 games 
  
  - Less weight on conference tournament results (those games are still included in the normal schedule results evaluation) 
  
  Also, score margin is now being adjusted for the pace of the game, so a 50-30 margin is considered slightly more impressive than a 90-70 margin. 

-------------

So, I get that these factors are the major players, but how much weight is given to each factor? How can I say if too much weight is given to the opponent RPI (for instance) when I don't know if the current weight given is 20% of the overall formula or 40%? Is there a downside to disclosing (at least roughly) the current weights (in that it might allow some to game the system)?

The heaviest weight is on outcome (W/L).  There is also a lot of weight given to opponent strength.  Then margin has a bit less.  Location has the least weight, with road games given more value than home games.

Generally speaking, I'm trying to determine if a majority of coaches feel that a certain type of resume is ranked too high or too low.  Most of the comments I've seen are related to middling teams from big conferences being ranked higher than high winning percentage teams from lesser conferences.  That's always a tough comparison to make, but the system currently values a strong schedule enough to push those teams higher.

This is a very subjective area, so there isn't a "right" answer.  I feel like the current logic is doing a good job of putting the truly stronger teams higher, but it's my job to get a feel for the HD community as well.  That's what I'm trying to do here.  I can't react to feedback from a few users if that doesn't accurately reflect the community as a whole.

I think the projection report, as it is now, is one of the strengths of the game.