|
Post by ccman on Dec 6, 2010 0:11:32 GMT -5
I echo this statement. But I would add that the NCAA should use the coaches poll as a tool. If "the coaches knew that the NCAA was going to use their poll to help determine the top 16 seeds, they might pay closer attention."
I would disagree about using the coaches poll. The coaches poll is gonna get perverted if the coaches know it's used for selection. Football has already experienced this. In theory, I like the idea of averaging RPI (or some other measure) with the Coaches Poll. But in practice, the coaches are only human and the selection/seeding process creeps into their thinking. They have a conflict of interest.
|
|
|
Post by redbeard2008 on Dec 6, 2010 0:26:56 GMT -5
I like it. I like that someone's talking about actually doing something about it. Don't be discouraged if it doesn't happen right away -- the important thing is you're getting people to start thinking of concrete steps to take instead of just endlessly talking about it. I will say this, though. Don't be so halfway about the RPI. Get rid of it. At most, allow its use only in extraordinary circumstances such as if there are reasons to believe the other ranking system's integrity has been compromised. The RPI is fundamentally flawed. To keep it on as part of a hybrid-ranking system or to simply tweak it in some ways is to imply that it has a legitimacy that it simply does not deserve. Get rid of it. Choose between Pablo or a chess-rating (Elo) type of system. Or propose Pablo and offer an Elo system as a 2nd choice -- because the NCAA may well reject Pablo as an official ranking system on grounds that the NCAA may well be adamant in insisting are legitimate ones. I echo this statement. But I would add that the NCAA should use the coaches poll as a tool. If the coaches knew that the NCAA was going to use their poll to help determine the top 16 seeds, they might pay closer attention. Or it might corrupt some of the coaches, who might be tempted to vote to advantage their team or disadvantage another team in the selection process. For that reason, I wouldn't tie them to the coaches poll, any more than too RPI (or modified RPI). I think including ex-coaches on the selection committee might help - I don't see anything wrong with applying plain old human judgment to come up with a result which will pass both the "eye test" and the "smell test" (the Seattle regional fails the "eye test", while the Penn State regional fails the "smell test").
|
|
|
Post by baywatcher on Dec 6, 2010 0:54:37 GMT -5
I must apologize for being imprecise regarding the over 125 RPI comment. My point is that teams competing for the at large births generally have RPI of 60 or under. If you have a contest between a 125 or above RPI team and a 65 or under RPI team the under 65 team invariably wins. Yet the difference between the effect on the under 65 team to defeating a 275 RPI team, v. 125, would seemingly be immense, as a practical matter the winning percentage for the 275 team will be abysmal.
An example I have used elsewhere. Cal State Bakersfield is newly arrived on the Division 1 scene and has no conference. Lots of the Cal, St. Mary's, high level Division 1 teams have scheduled them. Fine, gives everybody a chance to play. Yet CS Bakersfield did not win a match, or only won one or two. Therefore a big fat 0 is calculated into 50% of your RPI if you play them; similar results with UC Riverside, San Jose State. The result screams; do not play these teams! The trouble is in the west there are fewer teams that you can play that you are sure to beat, but still have representative RPI. Also, with regional scheduling you want/need to play these teams. If RPI 275 is treated the same as RPI 125, it doesn't really affect winners v. losers, but doesn't disproportionately punish you for playing a team with a high RPI.
|
|
|
Post by baywatcher on Dec 6, 2010 0:56:28 GMT -5
Agree Coaches poll would be clearly conflict of interest; but who needs a weekly poll anyway. A portion of the Committee being retired coaches, lead by Pettit, Banachowski, Shaw would be great as far as selecting at large teams. They could watch teams during the year and come with real, good opinions.
|
|
|
Post by FreeBall on Dec 6, 2010 9:25:17 GMT -5
If RPI 275 is treated the same as RPI 125, it doesn't really affect winners v. losers, but doesn't disproportionately punish you for playing a team with a high RPI. I think I understand the point you are trying to make, but still think it is not supported by the reality of the RPI. Looking at the most recent RPI: ** There is a significant cluster of teams around the 125 level that have losing records. 120. Arkansas (14-17); 122. Kansas State (12-19); 125. N. Carolina St. (14-18); 126. VCU (13-20); 127. Pittsburgh (13-18); 129. Oregon State (9-23). ** While certainly fewer in number, there are also examples of teams around the 275 level that have winning records. 258. Delaware State (18-12); 268. Maryland Eastern Shore (20-12); 277. Mississippi Valley (23-11). Since there clearly is no absolute correlation between RPI rank and winning %, what winning % would you arbitrarily assign to all teams with RPI's above 125?
|
|
|
Post by bigpsufan on Dec 6, 2010 9:30:11 GMT -5
I don't know who "us" is. If it's Volleytalk that is meaningless as Volleytalk represents a small percentage of the volleyball world.
|
|
|
Post by austintatious on Dec 6, 2010 9:36:49 GMT -5
This might be the most sensible thread I have seen in quite a while. All good ideas and thoughts, no name calling, but seems to me some level headed discussion.
Thanks, I needed this.
|
|
|
Post by upyours on Dec 6, 2010 11:19:13 GMT -5
I know a lot of the discussion focuses on the models used to predict ranking. While that discussion is worthy, I think a lot of missing the substance of what is deficient in the tournament. In general, if the VB tournament could remove inequities and become more like the men's BB tournament, then I believe it would help to generate interest and fairness. I think the biggest problem right now is the use of 'host' sites for the tournament that for the most part go hand in hand with the seeding. The key is not to do what you don't need to do. #1) Seeding and hosting. Get the host sites set before the season starts. VB is biggest in the heartland (hard for me to state as a west coaster), so divide it into West, Mid-West, Mid-East & South-East. Hawaii, Northwest, Bay Area, SoCal, Utah/Colo., Ariz are natural sites to have 4 each year in the west as an example. There's a lot of worry about sub-regionals not attracting crowds, but basically you can't go wrong with a LA and SF sub-regional every year and then rotating the other 2. Or have 2 sub-regionals at one site like basketball. Yes, there will be a year where if a regional is in Salt Lake, and Utah/BYU don't make it, you worry about low attendance, but hey, USU made it this year!! You have the same problem with regionals anyway. Furthermore, there should be rule that a sub-regional and regional cannot be at the same site. #2) Don't seed 1-16 for the nation. Seed at least 1-8 for each regional. Instead the commitee should create trial seeds nationally & groups from 1-32. Not only does the RPI-centric seeding suck, it propagates criticism, it invites nudging the seeding to match hosting and hamstrings the committee. The committee can generate a rough seeding, but then they should have 4 #1s, 4#2s, 4 #3s,... etc. It doesn't really matter if Nebraska or Stanford is #2 or #3, nor do Stanford or Neb care about being #2 or #3. so just DON"T do it. And it really doesn't matter then if the #12 & 13 are backwards as a practical matter. By not having to force the system & seeds into host sites as well, it frees the committee to focus on spreading teams out. Duke is a prime example. Even if they still ended up a #12 and hosted, at least they should have had a #20ish team on their court. Imagine the benefit to the tournament if Duke & San Diego (or Duke & Cincinnatti) had played 2nd round - it would be better for Duke and the ACC, WCC/Big East and better for volleyball, and it would negate and self-correct any criticism of seeding. #3) Separate the top teams from multiple bid conferences. If a conference has 4 teams, they should all be in different regionals as a guideline. By doing #1 & #2, the committee can more easily do #3. And if 2 teams from a conference are top 4 seeds, then it forces the two teams to different regionals. What this does is that even there is some dispute over teams getting a wrong 'seed', it allows the competition between conferences to 'sort' it out on the court. #4) Use the travel & regionalization to benefit the tournament, not to discredit it. Look a the tournament this year. USC & Stanford are in Dayton. Mississippi and North Carolina in Berkeley - the NCAA has these travel goals, yet ends up with a tournament with teams traveling anyway that don't make sense. So let the traveling make sense to more equitably create brackets. #5) RPI and other models. I really don't care which model is used, but don't have the model run the tournament which is what happened this year. I don't care whether it's Pablo or RPI. The committee should clearly know that the computer model is to be used as a GUIDELINE (i.e., sanity check and to create seeds with results 1-100, etc.) but not to be used as excuse, answer to a FAQ, or rule-book. lastly, I don't want to hear about how 'time-limited' etc. problems for the committee. This is 2010, we have web-conference, telephone conferences, all sorts of methods, so that if the committee starts their seeding/grouping work even remotely a week before the tournament, they can get it done. Very good post ccman. It sounds like changes that could be made. I'd also like to hear karplets elobrate on the ELo-type rankings and the formula used for Pablo.
|
|
|
Post by karplets on Dec 6, 2010 11:23:06 GMT -5
2. When RPI is utilized numbers over a certain level, say 125, should all be considered the same. (ALLOWS MATCHES AGAINST LOWER LEVELS WITHOUT PUNISHMENT) In an Elo-like (chessrating-like) system, your rating is not skewed by playing drastically higher-rated or drastically lower-rated teams and I have reason to believe Pablo shares that characteristic. You cannot lose rating points simply by playing by a very low rated team (obviously assuming you win) -- you simply don't gain rating points either. Conversely, you cannot gain rating points simply by playing a very highly rated team (and losing as expected to them). Let me illustrate using a very crude Elo rating model - and using a rough approximation of Pablo's rating scale. You play 3 teams all rated 5000 and you win 2 out of 3 matches. The average rating of your opponents is (obviously) 5000. Your win pct is (.667). That corresponds roughly to a +300 rating differential. (I'm rounding off for ease of illustration - the actual win probability in Pablo for 300 points is .680) In other words a team rated 300 points higher is expected to win 2 out of 3 times. So crudely speaking based on this limited game sample, your rating is 5000 (the average of your opposition) + 300 = 5300 Now let's say you play and beat a really weak team, one rated only 1000. Now the average of your opponents' ratings over 4 games is only 4000. (add the opponents ratings and divide by the number of opponents, which is 5000+5000+5000+1000 divided by 4) Your record is now 3-1 for a win pct of (.750) which corresponds very roughly to +500 (actually +500 is .782). Calculating your rating this way would leave 4000 (the average of your opponents) + 500 = 4500. You would be losing rating points for playing a weak opponent. (Your win pct goes up and that increases your rating differential from +300 to +500 but obviously the weak opponent is dragging down the average opponent rating even faster) My understanding of Elo systems is that, in a rough manner of speaking, you would "throw out" the result against the weak opponent because it is skewing the result in a misleading way. So your rating (for all intents and purposes) would be based only on the 3 original games and your rating would remain 5300. Elo systems (and Pablo, I believe) take care of problems like this for you.
|
|
|
Post by pogoball on Dec 6, 2010 12:16:17 GMT -5
I brought the point up in another thread, but it is better here:
Choosing which teams deserve to be invited to the tournament is one set of criteria.
Choosing how to seed/place those teams in the tournament doesn't have to be and, IMO, shouldn't be the same criteria.
We could use RPI for placing teams in the tournament, but we could use a Pablo system for seeding and placing them.
|
|
|
Post by karplets on Dec 6, 2010 12:25:44 GMT -5
I brought the point up in another thread, but it is better here: Choosing which teams deserve to be invited to the tournament is one set of criteria. Choosing how to seed/place those teams in the tournament doesn't have to be and, IMO, shouldn't be the same criteria. We could use RPI for placing teams in the tournament, but we could use a Pablo system for seeding and placing them. Conceivably it would be appropriate to have different or slightly different criteria for selecting the tournament vs seeding the tournament.* But why should RPI be one of them? If it's fundamentally flawed (and it is), get rid of it. Let's not make different criteria just for the sake of having different criteria. * For example (not that I necessarily support the idea) you could use the non-weighted version of Pablo (or an Elo system) for selecting who's invited and the weighted version for determining the seeds. That would be more reasonable than using the RPI for one and Pablo (or an Elo system) for the other.
|
|
|
Post by karplets on Dec 6, 2010 18:14:49 GMT -5
btw, I hope most people aren't seriously interested in using polls to determine the bracket.
|
|
|
Post by karplets on Dec 7, 2010 10:11:59 GMT -5
Very good post ccman. It sounds like changes that could be made. I'd also like to hear karplets elobrate on the ELo-type rankings and the formula used for Pablo. Maybe that post I made above gives you a flavor of how an Elo-type of rating works. I have reason to think there are key similarities with Pablo -- and a couple key differences. For those of you familiar with Pablo, I would tentatively think of an Elo system as similar to Pablo except that an Elo system only uses game results (won-losses in volleyball since there are no ties) and not points scored. An Elo system can either weight recent games more heavily than earlier ones or not. Pablo weights recent games more heavily but it doesn't have to -- a Pablo variation could be done that weights games equally no matter when in the season they are played. In both Elo systems and Pablo (from what I can tell), a team's rating tells you as a matter of probability how likely they are to win or lose to a team with a different rating. (That's the basic concept underlying both, I believe, and is used in calculating the ratings - with the difference that an Elo system is calculating solely on the probability of winning while Pablo is calculating on the probability of a certain number of point scored) The absolute numbers in themselves don't mean anything. It's the difference between the numbers that means something. Pablo for instance uses a scale where a difference of 100 points means the higher rated team is expected to win 56.2% of the time. And a 200 point difference means they should win 62.2%. (And the median is set at 5000) That's just the scale the creator of Pablo chose. She could've chosen a different scale if she wanted to. In chess for example, a 100 point difference corresponds to a 62.5% win probability for the higher-rated team, and a 200 pt difference corresponds to a 75% probability. In women's soccer I'm used to Prof Albyn Jones scale where a 100 pt difference corresponds to a 67% win probability and a 200 pt difference corresponds with an 80% win probability. (Median usually around 1350)
|
|
|
Post by big10+4 fan on Dec 7, 2010 11:27:25 GMT -5
The ex-coach idea is a good one except it has already been tried. Cook said that after last year's mess up the Big12 conference suggested to the NCAA that they have someone like Petit(yes he is an ex-neb coach but they said any ex-coach would do) to come in and look over the brackets and see if anything was out of whack. Of course they didn't listen very well did they?
|
|
|
Post by The Bofa on the Sofa on Dec 7, 2010 11:50:44 GMT -5
The ex-coach idea is a good one except it has already been tried. Cook said that after last year's mess up the Big12 conference suggested to the NCAA that they have someone like Petit(yes he is an ex-neb coach but they said any ex-coach would do) to come in and look over the brackets and see if anything was out of whack. Of course they didn't listen very well did they? I thought there were committee members who are either currently or former coaches? UPDATE: There are three former coaches on the committee, including Nona Richardson. So what would be the point of bringing in another former coach to look over the brackets? BTW, they didn't need a coach to look at the brackets last year to notice the Big 12 regional matchups. That was a mistake, but didn't happen because of lack of coaches on the committee.
|
|