|
Post by c4ndlelight on Dec 3, 2015 15:21:17 GMT -5
Oregon played 10 matches against the Top 12 (yes, TWELVE!) teams in the RPI, of course they are going to have a weaker W-L against the Top 100 compared to a team like the Beach that played 1 match against that level of competition. The Beach also played 8 of its 11 matches against the Top 100 at home, while Oregon only played 6 of its 16 at home. It's not interesting and curious that the NCAA Committee rewards teams for scheduling tough like this. It is known. The Beach knew what conference it was in, should have done better than the #93 non-conference SOS. again, NO you are absolutely incorrect, LB attempted to schedule a strong OOC schedule, it is not 'known' what the future is, LB scheduled fine, their OOC opponents winning % last year was 65%, they scheduled 9 OOC (bama, duke, Denver, Ore State, UCLA, LMU, SD, Oakland, Ill State) tourney teams from last year, there is NO WAY they could have predicted Bama (SEC pre-season #2), Duke (ACC pre-season #3), Oregon State would all go in the tank so bad sorry, but your comments are so indicative of how difficult in can be to 'game' RPI, and also that mid-majors get stuck with having to front-load their schedule how did Oregon do at the beginning - they couldn't beat a top 50 team it is what it is, but puuhhlleeeeze spare people this notion or lecturing teams about the 'ease' of getting your SOS up - it's not simple, and just trying to schedule a bunch of top 10 teams OOC would be ridiculous for any school Oregon St., Illinois St. - teams that reached historical program highs (by a mile!) and then graduated the majority of their lineup? I am SHOCKED to hear they didn't do as well this year. Bama and Duke underachieved expectations, but the effect of that on "scheduling intent" was cancelled out by the rest of that list all performing at above what they should have done - UCLA, LMU, USD all lost their best players and did a good deal better than last year - who woulda thought?
|
|
|
Post by BeachbytheBay on Dec 3, 2015 15:54:59 GMT -5
again, NO you are absolutely incorrect, LB attempted to schedule a strong OOC schedule, it is not 'known' what the future is, LB scheduled fine, their OOC opponents winning % last year was 65%, they scheduled 9 OOC (bama, duke, Denver, Ore State, UCLA, LMU, SD, Oakland, Ill State) tourney teams from last year, there is NO WAY they could have predicted Bama (SEC pre-season #2), Duke (ACC pre-season #3), Oregon State would all go in the tank so bad sorry, but your comments are so indicative of how difficult in can be to 'game' RPI, and also that mid-majors get stuck with having to front-load their schedule how did Oregon do at the beginning - they couldn't beat a top 50 team it is what it is, but puuhhlleeeeze spare people this notion or lecturing teams about the 'ease' of getting your SOS up - it's not simple, and just trying to schedule a bunch of top 10 teams OOC would be ridiculous for any school Oregon St., Illinois St. - teams that reached historical program highs (by a mile!) and then graduated the majority of their lineup? I am SHOCKED to hear they didn't do as well this year. Bama and Duke underachieved expectations, but the effect of that on "scheduling intent" was cancelled out by the rest of that list all performing at above what they should have done - UCLA, LMU, USD all lost their best players and did a good deal better than last year - who woulda thought? the scheduling intent was solid - and set up to be a level above the competition from the Big West, where Santa Barbara & Northridge compounded the SOS by having really bad years San Diego, UCLA do a good job retooling almost every year - the only one that looked obvious to go a step down was Oregon State scheduling is not that simplistic - going by the prior year is one of the best ways to plan - teams typically return a majority of their players - and then you can't necessarily schedule all the teams you might want to schedule if everyone 'chased' scheduling Florida State, North car, Kentucky, Kansas, etc., they couldn't accomadate everyone nor is it their job to
|
|
|
Post by BeachbytheBay on Dec 3, 2015 15:57:14 GMT -5
nope, best win LMU, on the road Still not comparable to the UCLA win, which was also in Oregon's last ten matches, while LBSU's road win over LMU was in their first ten. They had a chance to beat UCLA, at home, but didn't. absolutely right, and LB didn't get a 2nd shot at UCLA like Oregon did and Oregon didn't have two wins the equal of San Diego or LMU and LB had two shots at Hawaii Oregon also had 6 shots at USC, Stan, Wash we could go on and on bottom line is there's nothing Significant about what Oregon did compared to LB
|
|
|
Post by vbphilsdad on Dec 4, 2015 11:05:34 GMT -5
Yup. Except, that there is a bit of speculation as to who will win (I believe),
No. There is not "quite a bit of speculation." There is prediction based on Pablo, which has some variation for an individual team wins and losses, but works VERY well for opp and opp, because the sample sizes are vary large. This is the whole reason BluePenguin started doing it, because he realized that the opp and opp opp records are actually pretty easily determined once you get into the conference season. ? No, it predicts the end of the season RPI. And it does a great job at doing that. The variation you see is mostly based on teams performing above or below expectations. Win matches you should, you stay in place. Upset someone, you move up. Get upset, you move down. It's beautiful. How does "there is a bit of speculation as to who will win (I believe)" become "quite a bit of speculation". Quoting a misquote.. ummm
In any event, thanks for the clarifications on RPI futures, it is more what I'm looking for than NCAA RPI, since it accounts for future match-ups. I like it. As you describe it, though, it still requires a 'decision' on who will win those future match-ups. I was simply expressing the wish for an RPI that simply says an opp or opp opp is 12-6 rather than one that uses a predicted will be 22-10 at seasons end.
NCAA RPI forecasts nothing from week to week, not even that a set schedule will be played out (ie. UF was not factored into LIU until they actually met). That's a weakness. RPI futures forecasts everything in the future. Not only that 2 teams will meet, but who will win. That's great. Better, even. Just wondering if there was a happy medium. Perhaps not.
By the way, isn't "expectations" + "predictions" = speculation? Particularly if performed in October and carried all the way out to the end of the season?
|
|
|
Post by The Bofa on the Sofa on Dec 4, 2015 11:17:54 GMT -5
No. There is not "quite a bit of speculation." There is prediction based on Pablo, which has some variation for an individual team wins and losses, but works VERY well for opp and opp, because the sample sizes are vary large. This is the whole reason BluePenguin started doing it, because he realized that the opp and opp opp records are actually pretty easily determined once you get into the conference season. ? No, it predicts the end of the season RPI. And it does a great job at doing that. The variation you see is mostly based on teams performing above or below expectations. Win matches you should, you stay in place. Upset someone, you move up. Get upset, you move down. It's beautiful. How does "there is a bit of speculation as to who will win (I believe)" become "quite a bit of speculation". Quoting a misquote.. ummm
In any event, thanks for the clarifications on RPI futures, it is more what I'm looking for than NCAA RPI, since it accounts for future match-ups. I like it. As you describe it, though, it still requires a 'decision' on who will win those future match -ups. But this isn't a problem in terms of what you want. As I said, the variation still comes down to your team winning or losing. It's just that there is a third option, "no change" if you win or lose as expected. Not really, although it depends on how you are using "speculation." For example, if you use the common definition (that pops up on an internet search) "the forming of a theory or conjecture without firm evidence." then not at all. By early October, there is absolutely firm evidence for the predictions for the rest of the year. Pablo correctly predicts more than 75% of the matches played for the rest of the year at that point. That's because there is very firm evidence for how good teams are. In fact, the prediction RATE does not increase all that much as the season progresses, when more evidence is available. So October ratings are really good. There is another definition of speculation, investment in stocks, property, or other ventures in the hope of gain but with the risk of loss. that doesn't really apply in this situation.
|
|
|
Post by vbphilsdad on Dec 4, 2015 11:36:43 GMT -5
"Pablo correctly predicts more than 75% of the matches played for the rest of the year at that point." Exactly. Pablo is MUCH a much better predictive tool that RPI -- and a much better ranking tool as well. Better than RPI in all aspects. There is no doubt that RPI future does a great job at predicting the final RPI. In fact, they converge exactly (one would hope) at season's end. Is there a Pablo futures? Unfortunately RPI is what we are stuck with (as the exercise of this thread has shown). So RPI futures perhaps gives a team an earlier indication that they don't have a hope of making the NCAA's and should focus on their conf tourney or, conversely, can play their conference season with without a care (hello gators) knowing they simply cannot fall out of the top-15.
BB RPI has had many tweaks over the years, due to need to seed 64 teams as well as the $$ involved in gertting a tourney bid. Any of these tweaks germane to the VB RPI? maybe requiring to seed more teams would push these tweaks ahead.
Bofa, I appreciate the clarifications you've given me from someone who has looked at this for some time, like yourself. Thanks.
|
|
|
Post by BeachbytheBay on Dec 4, 2015 12:05:08 GMT -5
my only problem with using Pablo in selection is that one should be careful as well. Do you really want a 'predictive' tool to be used for selection? versus a ranking tool, based more on wins. Pablo 'ranks' teams, but it is based on predictions, in and of itself that's fine, but still have to be cautious, cause Pablo also reports 'ranges'.
my big problem with RPI is that it is not statistical based in ranking. it's simply a formula, and then people (like BOFA) applied stats to it to analyze how it predicts.
Pablo & Massey are statistical based. But the goal in selection isn't necessarily to be a 'predictor' of results. because there are other factors - some in the selection guidelines - that are the main source of consternation - principally, how 'quality' wins-losses are applied and the weight given.
Plus weight given to recent - last 10, quality road wins, injuries...., which is what makes the Arizona State selection so mind-boggling - they didn't pass the secondary tests of what constitutes a tourney team.
IMO, the simple tweeks would be:
to use a composite, hell keep RPI in there as one of 3-4 indexes that could be averaged.
throw out worst loss (toss out the potential 'flukes'), apply weight for the average of the top 3 or 4 wins (and especially the top 2 road/neutral wins) a team has to better see the 'quality' - to supplement the current 25/50/100 cut-offs that leave the committee members 'blind' to rounding errors.
and then apply quality win/losses to the indexed rating - IMO a 'bad loss' should be about above 125 (approximately top 1/3 of the NCAA)
I think if the NCAA did those simple things it would remove a huge component of the RPI & selection complaints
|
|
bluepenquin
Hall of Fame
4-Time VolleyTalk Poster of the Year (2019, 2018, 2017, 2016), All-VolleyTalk 1st Team (2021, 2020, 2019, 2018, 2017, 2016)
Posts: 12,423
|
Post by bluepenquin on Dec 4, 2015 12:14:20 GMT -5
"Pablo correctly predicts more than 75% of the matches played for the rest of the year at that point." Exactly. Pablo is MUCH a much better predictive tool that RPI -- and a much better ranking tool as well. Better than RPI in all aspects. There is no doubt that RPI future does a great job at predicting the final RPI. In fact, they converge exactly (one would hope) at season's end. Is there a Pablo futures? Unfortunately RPI is what we are stuck with (as the exercise of this thread has shown). So RPI futures perhaps gives a team an earlier indication that they don't have a hope of making the NCAA's and should focus on their conf tourney or, conversely, can play their conference season with without a care (hello gators) knowing they simply cannot fall out of the top-15.
RPI Futures allows teams to know very early how many wins with their schedule will be required to achieve a desired RPI. Particularly a total win total to either be in the discussion or safe for a seed and in the discussion or safe for a bid. This can be done because the OPP w/l% and OPP/OPP w/l% don't change much after the 1st couple weeks (3rd/4th week) of the season.
RPI Futures is also a projected RPI - however this is only as accurate as the final record for the team in question.
A couple examples:
The Week 3 RPI Futures had Oregon projected to go 16-13 and finish #47 in the RPI. Oregon ended up going 16-13 and finished #48 in the RPI.
The Week 3 RPI Futures had Minnesota projected to go 19-11 and finish #22 in the RPI. Minnesota ended up going 26-4 and finished #2. If Minnesota had finished 19-11, I suspect they would have been in the 20's for RPI. Their OPP and OPP/OPP didn't change (much), but they exceeded their own wins.
Essentially, teams can mostly only improve their final RPI from the RPI Futures (projection) by exceeded their own projected win total, not by having the teams they played win more games.
|
|
|
Post by vbphilsdad on Dec 4, 2015 12:15:33 GMT -5
BB RPI has had many tweaks over the years, due to need to seed 64 teams as well as the $$ involved in getting a tourney bid. Any of these tweaks germane to the VB RPI? maybe requiring to seed more teams would push these tweaks ahead.
Is BB RPI considered a better RPI that VB? If so, that would be ironic, since the reliance on RPI in VB is based in part on the inability to have a large enough pool of "eyes" who would have seen a large number of the top 100 so the committee relies on RPI. Basketball doesn't have that issue, yet they've tweaked their RPI to make it 'better".
So. BB downplays a 'good' RPI system, and VB relies heavily on a 'bad' one. Is that a correct assessment?
|
|
|
Post by vbphilsdad on Dec 4, 2015 12:26:18 GMT -5
Essentially, teams can mostly only improve their final RPI from the RPI Futures (projection) by exceeded their own projected win total, not by having the teams they played win more games.
... nor by who they win or lose to. There is no "strength of victory" component, to borrow an NFL phrase, in the RPI. Maybe more weight to the opp % and opp opp % of the teams you actually beat instead of merely play?
|
|
|
Post by BeachbytheBay on Dec 4, 2015 12:26:29 GMT -5
BB uses RPI much better - it provides a general '1st cut' ranking
the selection is so much more sophisticated, and I'd say not really relying on eyes so much except for the top 12-16 national seeds
since assuming VB doesn't have the proper 'eyes', I certainly wouldn't count on VB committee members being expert enough, you can get better sophistication by tweeking the criteria to be more sophisticated, namely accounting for 'quality' better
I would also add the 'qaulity' W-L should also be reported as W-L percentages to better gauge teams
any top 50 team could play 10 top 10 teams and come away with 1 win - it's simple statistical odds - so have the committee look at that W-L % and the average of the top wins to better compare teams from disparate conferences
and yes, you have to have some threshold, a team that has only 1-2 top 50 matches (or in Hawaii's case this year, 2 top 25 matches) is gonna incur some scheduling penalty - but they should provide some counter-balance for the RPI SOS that really can add bias
|
|
|
Post by vbphilsdad on Dec 4, 2015 12:37:48 GMT -5
BB uses RPI much better - it provides a general '1st cut' ranking the selection is so much more sophisticated, and I'd say not really relying on eyes so much except for the top 12-16 national seeds since assuming VB doesn't have the proper 'eyes', I certainly wouldn't count on VB committee members being expert enough, you can get better sophistication by tweeking the criteria to be more sophisticated, namely accounting for 'quality' better I would also add the 'qaulity' W-L should also be reported as W-L percentages to better gauge teams any top 50 team could play 10 top 10 teams and come away with 1 win - it's simple statistical odds - so have the committee look at that W-L % and the average of the top wins to better compare teams from disparate conferences and yes, you have to have some threshold, a team that has only 1-2 top 50 matches (or in Hawaii's case this year, 2 top 25 matches) is gonna incur some scheduling penalty - but they should provide some counter-balance for the RPI SOS that really can add bias So, perhaps like I said, a bigger weight assigned to the match-ups you won rather than lost. the RPI components assigned to OPP and OPP OPP should weighted 60-40 or 75-25 team you beat?
|
|
|
Post by BeachbytheBay on Dec 4, 2015 12:53:52 GMT -5
BB uses RPI much better - it provides a general '1st cut' ranking the selection is so much more sophisticated, and I'd say not really relying on eyes so much except for the top 12-16 national seeds since assuming VB doesn't have the proper 'eyes', I certainly wouldn't count on VB committee members being expert enough, you can get better sophistication by tweeking the criteria to be more sophisticated, namely accounting for 'quality' better I would also add the 'qaulity' W-L should also be reported as W-L percentages to better gauge teams any top 50 team could play 10 top 10 teams and come away with 1 win - it's simple statistical odds - so have the committee look at that W-L % and the average of the top wins to better compare teams from disparate conferences and yes, you have to have some threshold, a team that has only 1-2 top 50 matches (or in Hawaii's case this year, 2 top 25 matches) is gonna incur some scheduling penalty - but they should provide some counter-balance for the RPI SOS that really can add bias So, perhaps like I said, a bigger weight assigned to the match-ups you won rather than lost. the RPI components assigned to OPP and OPP OPP should weighted 60-40 or 75-25 team you beat? I would disagree. I wouldn't bother changing the formula for RPI - it is what it is. And changing it I'm not sure would eliminate the bias inherent in it. It just should be applied differently in terms of 'quality' W-L, and in conjunction with another index or multiple indices. RPI is ok to make general groupings - but once you get to those groupings (1-25ish for national seeds, & 30-70ish for bubble teams), the application of RPI in the W-L categories in it's current form compounds the flaws and bias inherent in it - I doubt tweeking those percentages would really prevent the compounding, you have to have other measures to prevent the compounding flaws
|
|
|
Post by vbphilsdad on Dec 4, 2015 13:14:56 GMT -5
Doesn't weighting who you beat objectively result in determining quality wins? Beating USC adds to your OPP % more than beating Utah. Right now going 1-1 against makes no difference who you beat.
You are saying don't fix the RPI, but come up with other indices. Why not both? BB did/does it.\
Good discussion!
|
|
|
Post by The Bofa on the Sofa on Dec 4, 2015 13:21:43 GMT -5
my only problem with using Pablo in selection is that one should be careful as well. Do you really want a 'predictive' tool to be used for selection? versus a ranking tool, based more on wins. REMINDER: Pablo is based 100% on what a team has done. The only question is, what part of "what a team has done" matters? Pablo looks at the things that a team has done that translates into future success. Why wouldn't you want to use them for picking teams to play for the championship? And your comment about a ranking tool based "more on wins" is similarly misguided. Pablo actually reflects wins and losses BETTER than RPI does (at least, RPI that doesn't have an HCA component - by adding an HCA to RPI, it can do AS WELL as Pablo, but that is it). And if you really want to use wins, why not use a method that is much better in wins? I have that, too. However, I'd advise against them. So any realistic approach is not going to be solely about wins and losses, and you are going to have some other factors come into play. In that case, why not look at things that successful teams do?
|
|