|
Post by badgerbreath on Oct 22, 2018 11:20:04 GMT -5
When we gauge the toughness of schedule by eye, we often only pay attention to the very top tier teams -- of which there are 5(!) this year, Minn, UW, Ill, PSU, Neb -- because we tend to think that matches between these teams will determine the champion(s). Be interesting to know if that idea has been true through the years.
|
|
|
Post by FreeBall on Oct 22, 2018 11:22:15 GMT -5
To all those whining about their team's RPI, there's a really simple answer. If you want your team to have a decent RPI convince your coach to schedule stronger teams in the non-conference part of the season. This isn't rocket science. And frankly, your coach already knows this. Pitt and Illinois, for example, chose to really challenge their teams with tough competition at the start of the season. And by going unbeaten then they have earned choice positioning for the playoffs. Other teams chose to compete against much weaker opponents for their own reasons. Too bad. Choose tougher competition next time. In a post later in this thread you stated the following: However, your post about scheduling "tougher" teams as the key to attaining a decent RPI indicates that you really don't understand the RPI all that well. It's not about scheduling "tougher" teams. It's about scheduling teams that will end up with a good overall record. Teams that dominate weaker conferences are not generally "tougher" teams, but they are quite often the best type of team to schedule in order to enhance your own RPI.
|
|
|
Post by vbkahuna on Oct 22, 2018 12:19:44 GMT -5
To all those whining about their team's RPI, there's a really simple answer. If you want your team to have a decent RPI convince your coach to schedule stronger teams in the non-conference part of the season. This isn't rocket science. And frankly, your coach already knows this. Pitt and Illinois, for example, chose to really challenge their teams with tough competition at the start of the season. And by going unbeaten then they have earned choice positioning for the playoffs. Other teams chose to compete against much weaker opponents for their own reasons. Too bad. Choose tougher competition next time. In a post later in this thread you stated the following: However, your post about scheduling "tougher" teams as the key to attaining a decent RPI indicates that you really don't understand the RPI all that well. It's not about scheduling "tougher" teams. It's about scheduling teams that will end up with a good overall record. Teams that dominate weaker conferences are not generally "tougher" teams, but they are quite often the best type of team to schedule in order to enhance your own RPI. On the contrary, you apparently are unaware that the RPI includes not just your team's winning percentage and your opponents' winning percentage but also your opponents' opponents' winning percentage. If you only play teams that play against only weak teams, your RPI will be adversely impacted. You need to schedule teams that will likely have winning records against other teams that also have winning records. That translates conversationally into "tougher teams". Do I need to spell out the entire formula for you?
|
|
|
Post by trollhunter on Oct 22, 2018 12:23:43 GMT -5
In a post later in this thread you stated the following: However, your post about scheduling "tougher" teams as the key to attaining a decent RPI indicates that you really don't understand the RPI all that well. It's not about scheduling "tougher" teams. It's about scheduling teams that will end up with a good overall record. Teams that dominate weaker conferences are not generally "tougher" teams, but they are quite often the best type of team to schedule in order to enhance your own RPI. On the contrary, you apparently are unaware that the RPI includes not just your team's winning percentage and your opponents' winning percentage but also your opponents' opponents' winning percentage. If you only play teams that play against only weak teams, your RPI will be adversely impacted. You need to schedule teams that will likely have winning records against other teams that also have winning records. That translates conversationally into "tougher teams". Do I need to spell out the entire formula for you? You may want to review prior years discussions of RPI. It comes up every year. The RANGE of OOWP is very small so has much less impact on RPI than the other two components. Don't get confused with the 3 components percentage weighting - it is that there is very little difference in the range of OOWP that it has a small effect.
|
|
|
Post by The Bofa on the Sofa on Oct 22, 2018 12:35:35 GMT -5
On the contrary, you apparently are unaware that the RPI includes not just your team's winning percentage and your opponents' winning percentage but also your opponents' opponents' winning percentage. If you only play teams that play against only weak teams, your RPI will be adversely impacted. You need to schedule teams that will likely have winning records against other teams that also have winning records. That translates conversationally into "tougher teams". Do I need to spell out the entire formula for you? You may want to review prior years discussions of RPI. It comes up every year. The RANGE of OOWP is very small so has much less impact on RPI than the other two components. Don't get confused with the 3 components percentage weighting - it is that there is very little difference in the range of OOWP that it has a small effect. In terms of relative differences, the OOWP accounts for about 10% of the range. OWP is about 40%, and winning percentage is about 50%.
|
|
|
Post by trollhunter on Oct 22, 2018 12:35:53 GMT -5
It was Mike Garrison that explained it well: The confusion is between the amount of weight in the formula and the amount of variation in the different components. OOWP, for instance, is weighted at 25%, but for most teams the actual OOWP is pretty close to 0.500. So because the variation in OOWP is small, then the effect of it on the formula will be small, even though it is weighted at 25%. A fictitious example: let's say the typical team's OOWP is 0.500. But your team's OOWP is 0.505. Even though the formula gives it a weighting of 25%, that's still only 25% of 0.005 in terms of separating you away from the average team. Now let's say your WP is .825 and the average team's WP is 0.500. Even though this is weighted the same as the OOWP (25%), it is 25% of 0.325 that separates you from average, versus being 25% of 0.005. So even though the two factors are weighted the same, the greater variation in the WP compared to the OOWP means that the WP drives the results more. And Bofa did even better in post above.
|
|
|
Post by The Bofa on the Sofa on Oct 22, 2018 12:43:30 GMT -5
It was Mike Garrison that explained it well: The confusion is between the amount of weight in the formula and the amount of variation in the different components. OOWP, for instance, is weighted at 25%, but for most teams the actual OOWP is pretty close to 0.500. So because the variation in OOWP is small, then the effect of it on the formula will be small, even though it is weighted at 25%. A fictitious example: let's say the typical team's OOWP is 0.500. But your team's OOWP is 0.505. Even though the formula gives it a weighting of 25%, that's still only 25% of 0.005 in terms of separating you away from the average team. Now let's say your WP is .825 and the average team's WP is 0.500. Even though this is weighted the same as the OOWP (25%), it is 25% of 0.325 that separates you from average, versus being 25% of 0.005. So even though the two factors are weighted the same, the greater variation in the WP compared to the OOWP means that the WP drives the results more. And Bofa did even better in post above. I like Mike's example. Although it would be useful to put some context to those pulled out of thin air numbers. Let me see what I can find.
|
|
|
Post by FreeBall on Oct 22, 2018 12:46:27 GMT -5
In a post later in this thread you stated the following: However, your post about scheduling "tougher" teams as the key to attaining a decent RPI indicates that you really don't understand the RPI all that well. It's not about scheduling "tougher" teams. It's about scheduling teams that will end up with a good overall record. Teams that dominate weaker conferences are not generally "tougher" teams, but they are quite often the best type of team to schedule in order to enhance your own RPI. On the contrary, you apparently are unaware that the RPI includes not just your team's winning percentage and your opponents' winning percentage but also your opponents' opponents' winning percentage. If you only play teams that play against only weak teams, your RPI will be adversely impacted. You need to schedule teams that will likely have winning records against other teams that also have winning records. That translates conversationally into "tougher teams". Do I need to spell out the entire formula for you? Let's look at two hypothetical non-conference schedules for the 2018 volleyball season: Schedule #1 | Schedule #2 | Kennesaw St. | Wichita St. | Radford | Baylor | James Madison | Iowa St. | Yale | Hawaii | Miami (OH) | San Diego | Austin Peay | Colorado St. | Lehigh | Oregon | East Tennessee St. | UCLA | Stephen F. Austin | USC | Denver | Utah |
Now let me ask you two questions: 1. Which hypothetical schedule includes the "tougher" teams? and, 2. Which hypothetical schedule would give a team a better RPI at this point in the 2018 season?
|
|
|
Post by trollhunter on Oct 22, 2018 12:51:31 GMT -5
Bofa actually already did the work previously on ranges: "Although the actual formula weights opp% at 50%, that does NOT mean that 50% of the differences in RPI are due to opp%. In fact, it is closer to the case that 50% of the difference in RPI is due to your own win%, opp% is about 40%, and oppopp% is about 10. The reason is because of the actual ranges. For example, the difference between the top team in winning% to the bottom is the difference basically between 100% and 0%. Since winning % accounts for 25% of rpi, that means the difference between top winning% and bottom winning% is 0.25. OTOH, the range between the top and bottom opp% is from about .700 to .300. Since RPI takes half of that, the difference in RPI is about .200. Finally, the difference between top and bottom OppOpp% is from .600 to .400. Therefore, the difference in the RPI contribution is only 0.05. Thus, the difference between top and bottom is .25 due to win%, .20 due to Opp%, and .05 in OppOpp%. There will be variations in any given year, but this is close to what it is. Therefore, the main factor that separates you from others is your win%, which is slightly more important than opp%. OppOpp% makes very little difference overall." Read more: volleytalk.proboards.com/search/results?captcha_id=captcha_search&what_at_least_one=oppopp&who_only_made_by=0&display_as=0&search=Search#ixzz5UgQtwuMb
|
|
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: 12,446
|
Post by bluepenquin on Oct 22, 2018 12:54:53 GMT -5
Okay - I shouldn't have been trying to do this at work - and then find I had made an error while stuck in a meeting...
The numbers sent earlier on the nonconference schedule was correct, but something went wrong on the conference schedule. Has to do with my dependent formulas I use for the simulations and didn't correct for this...
If we make Minnesota more similar to the other 4 so that we don't get the - but Minnesota doesn't have to play themselves comment, here is the expected w/l% in conference. There really isn't much difference in the conference schedules.
1) Nebraska - .751 2) Penn State - .764 3) Wisconsin - .769 4) Minnesota - .768 5) Illinois - .783
I started looking specifically at the difference between Wisconsin and Minnesota, and didn't see anything close to what I was writing before. Then realized my formula error.
|
|
|
Post by c4ndlelight on Oct 22, 2018 12:56:30 GMT -5
It was Mike Garrison that explained it well: And Bofa did even better in post above. I like Mike's example. Although it would be useful to put some context to those pulled out of thin air numbers. Let me see what I can find. I did this a few years ago, and IIRC, at the most extremes, OWP only ranged from about .70 (a VERY select few with pretty large gaps) to .30, with basically everyone in the .38-.63 range. OOWP had a spread of well less than three tenths. Most of the bottom of both of those ranges were teams that not only weren't even close to NCAA contention but also barely played any Top 100 RPI teams so in practice the spreads were smaller than that.
|
|
|
Post by The Bofa on the Sofa on Oct 22, 2018 13:07:09 GMT -5
It was Mike Garrison that explained it well: The confusion is between the amount of weight in the formula and the amount of variation in the different components. OOWP, for instance, is weighted at 25%, but for most teams the actual OOWP is pretty close to 0.500. So because the variation in OOWP is small, then the effect of it on the formula will be small, even though it is weighted at 25%. A fictitious example: let's say the typical team's OOWP is 0.500. But your team's OOWP is 0.505. Even though the formula gives it a weighting of 25%, that's still only 25% of 0.005 in terms of separating you away from the average team. Now let's say your WP is .825 and the average team's WP is 0.500. Even though this is weighted the same as the OOWP (25%), it is 25% of 0.325 that separates you from average, versus being 25% of 0.005. So even though the two factors are weighted the same, the greater variation in the WP compared to the OOWP means that the WP drives the results more. And Bofa did even better in post above. So let's consider a team who is in the 80th percentile in everything. 80th percentile in winning pct, 80th in OWP and 80th in OOWP. According to the current numbers on RichKern.com, that means the winning pct is .704, OWP = .5586 and OOWP is .5400 That team's raw RPI is 0.5903, which is 0.0903 higher than an average team (with .500 across). Of that 0.0903 difference, 0.0510 (56%) is due to the difference in winning pct 0.0293 (32%) is due to the difference in OWP, and 0.0100 (11%) is due to the difference in OOWP BTW, 80% in WP, OWP and OOWP gets you an RPI more like in the 85th percentile (the 80th percentile in raw RPI is 0.569) When you understand the nature of RPI like this, it puts a different perspective on how you view those percentages in the formula. I used to think it was more or less made up, but these days, I tend to think they were chosen with this in mind - these percentages give you a ranking that is basically half your record and half your opponents, broken down as 80% your OPP and 20% your OPPOPP.
|
|
|
Post by The Bofa on the Sofa on Oct 22, 2018 13:14:16 GMT -5
I like Mike's example. Although it would be useful to put some context to those pulled out of thin air numbers. Let me see what I can find. I did this a few years ago, and IIRC, at the most extremes, OWP only ranged from about .70 (a VERY select few with pretty large gaps) to .30, with basically everyone in the .38-.63 range. OOWP had a spread of well less than three tenths. Most of the bottom of both of those ranges were teams that not only weren't even close to NCAA contention but also barely played any Top 100 RPI teams so in practice the spreads were smaller than that. This year, the range for OWP is, ignoring outliers, about 0.66 to 0.35 For OOWP, it's about 0.6 to 0.4 That is consistent with my analysis above. If we use the range in record of 0 - 0.93, that gives 0.2325 for record (53%) 0.155 for OPP (35%) 0.05 for OPPOPP (11%) ----- .4375
|
|
|
Post by c4ndlelight on Oct 22, 2018 13:16:07 GMT -5
I did this a few years ago, and IIRC, at the most extremes, OWP only ranged from about .70 (a VERY select few with pretty large gaps) to .30, with basically everyone in the .38-.63 range. OOWP had a spread of well less than three tenths. Most of the bottom of both of those ranges were teams that not only weren't even close to NCAA contention but also barely played any Top 100 RPI teams so in practice the spreads were smaller than that. This year, the range for OWP is, ignoring outliers, about 0.66 to 0.35 For OOWP, it's about 0.6 to 0.4 That is consistent with my analysis above. If we use the range in record of 0 - 0.93, that gives 0.2325 for record (53%) 0.155 for OPP (35%) 0.05 for OPPOPP (11%) ----- .4375 What are the outliers at, if you already have that data up?
|
|
|
Post by The Bofa on the Sofa on Oct 23, 2018 16:31:40 GMT -5
What are the outliers at, if you already have that data up? Bah. That is all messed up. I was using last year's data. Try it again. If I use the full range, I get Rec 0.952 - 0.042 (the distribution is a little light in the high end, but I took out the two undefeateds - the list is 1.00, 1.00, 0.952, 0.950, 0.947, 0.92...) Opp 0.725 - 0.393 (no outliers) OppOpp 0.621 - .394 (no outliers) OK, so that means the range is .2275 for Rec (50.7%) .166 for Opp (36.9%) .057 for OppOpp (12.7%) Now for the 80th percentile Rec .6818 Opp .5648 OppOPP .5459 Total RPI .5893 So this team has an RPI that is 0.0893 higher than a team that is all 0.5 That difference is due to Rec .045 (50.9%) Opp .032 (36.3%) OppOpp 0.11 (12.8%) Wow, that is right on the same.
|
|