RPI Futures - Week 3 (9/10/17)

[bluepenquin]

Sept 12, 2017 20:58:59 GMT -5 wiscvball said:

Sept 12, 2017 20:10:08 GMT -5 Fight On! said:

blue, can you explain something to me:
19. USC - Pac 12, Avg RPI - 25.271 (0) T4 - 1.0% T16 - 34.0% T45 - 88.1%
20. UCLA - Pac 12, Avg RPI - 25.732 (1) T4 - 2.4% T16 - 32.6% T45 - 88.9%

How does it work out that USC has a higher change of getting to the sweet 16 than ucla but a lower chance of getting to the final four?

Thanks for educating me Appreciate all you do to produce these analyses.

I'm sure BP will give a better explanation rooted in data, but a few quick thoughts:

* the %'s are not the chances of making a round, they are the % of times the team's final RPI ended up in that range given the 1000 different permutations of match outcomes.
* the reason UCLA has a higher probability to get to Top 4 RPI is probably driven by their strength of schedule (in the first post - UCLA is at 5, USC is at 23). UCLA has more upside potential given future match outcomes.

This is correct.

There is some *noise* when talking about 1/10th of % and only 1000 samples. 

UCLA's schedule has some interesting developments.  Currently 9 non conference matches and only 4 of them within the top 75 - so they miss out on the scheduling bonus.  Hawaii is #76 in unadjusted RPI Futures, so just a small positive shift for Hawaii will add points to UCLA (Top 75 add to scheduling bonus).  And then you have San Diego that is #26 in unadjusted RPI - so a small move to #25 or better and UCLA gets the Top 25 win bonus points instead of Top 50.  These are 'lost' points for UCLA when doing RPI Futures, but have a probability > 0 when looking at 1000 simulations.  UCLA's schedule has really taken a hit vs. preseason with Hawaii, Long Beach State, UNLV, & LMU being worse this year.

USC doesn't appear to have those same kinds of variability in gaining/losing points tied to what other teams do.  The big one for them is UNI.  If Northern Iowa can get back into the top 75 in unadjusted RPI than they will get the scheduling bonus - otherwise they will not.  UNI is currently #86 before last night's win.

The one that caught my eye was Illinois and Oregon.  #10 and 11 - yet Oregon is almost twice as likely to be in the Top 4 than Illinois and over 3X more likely to finish in the Top 3.  And yet the SOS between the 2 is almost the same.  Maybe because Illinois has more teams to contend for conference championship and thus a lower probability of winning their conference?

[The Bofa on the Sofa]

Sept 13, 2017 8:00:33 GMT -5 bluepenquin said:

USC doesn't appear to have those same kinds of variability in gaining/losing points tied to what other teams do.  The big one for them is UNI.  If Northern Iowa can get back into the top 75 in unadjusted RPI than they will get the scheduling bonus - otherwise they will not.  UNI is currently #86 before last night's win

UNI took a BIG hit due to their loss to UMKC. We'll see how Pablo responds to the win over ISU. However, I have noticed all year that UNI has been on non-firm foundation. Pablo just isn't sure.

[Fight On!]

Thanks blue. Qualitatively, how much impact does the scheduling bonus have on RPI? Does it bump a team up 2 spots? 4? Who knows?

[trollhunter]

Sept 13, 2017 9:56:31 GMT -5 Fight On! said:

Thanks blue. Qualitatively, how much impact does the scheduling bonus have on RPI? Does it bump a team up 2 spots? 4? Who knows?

Factor IV: Bonus: For wins against teams ranked 1 through 25 in the original RPI, teams receive a bonus that is
approximately two positions in the RPI. For wins against teams ranked 26 through 50 in the original RPI, teams receive
a bonus of approximately one position in the RPI.

Bonus: Teams receive a bonus of approximately two positions in the RPI for playing 50 percent of their
nonconference schedule against teams ranked 1 through 75 in the original RPI.

Penalty: For losses against teams ranked 285 through 309 in the original RPI, teams receive a penalty
of approximately one position in the RPI. For losses against teams ranked 310 through the remainder of the
original RPI, and against non-Division I teams, teams receive a penalty of approximately two positions in
the RPI. *still find this hilarious that they changed this from 150*

Penalty: Teams receive a penalty of approximately two positions in the RPI for playing 50 percent of
their nonconference schedule against teams ranked 260 through the remainder of the teams, and against
non-NCAA opponents.

NCAA PRE-CHAMPIONSHIP MANUAL
DIVISION I WOMEN’S VOLLEYBALL

[bluepenquin]

Sept 13, 2017 9:56:31 GMT -5 Fight On! said:

Thanks blue. Qualitatively, how much impact does the scheduling bonus have on RPI? Does it bump a team up 2 spots? 4? Who knows?

From trollhunters's description below - 1 place is equal to .0014 and 2 places is .0028. Scheduling bonus is 2 places.  The difference in RPI rank depends... Larger impact for teams in the 20-60 range than those in the 10-20 range. You can look at the actual RPI score on the original post to see where .0014 and .0028 would impact in terms of rank.

For this run - it costs UCLA just 1 spot. For Michigan State it would be 3 spots and very close to being 5 spots.  For Kansas and Penn State - it wouldn't have changed their ranking.

BTW, I don't mess with penalty points in calculating RPI Futures.  It is unlikely that it would ever come into play for a team where RPI matters (bids and seeds).  Part of the reason I don't give the RPI rank of all 334 teams.

[c4ndlelight]

Sept 13, 2017 8:00:33 GMT -5 bluepenquin said:

Sept 12, 2017 20:58:59 GMT -5 wiscvball said:

I'm sure BP will give a better explanation rooted in data, but a few quick thoughts:

* the %'s are not the chances of making a round, they are the % of times the team's final RPI ended up in that range given the 1000 different permutations of match outcomes.
* the reason UCLA has a higher probability to get to Top 4 RPI is probably driven by their strength of schedule (in the first post - UCLA is at 5, USC is at 23). UCLA has more upside potential given future match outcomes.

This is correct.

There is some *noise* when talking about 1/10th of % and only 1000 samples. 

UCLA's schedule has some interesting developments.  Currently 9 non conference matches and only 4 of them within the top 75 - so they miss out on the scheduling bonus.  Hawaii is #76 in unadjusted RPI Futures, so just a small positive shift for Hawaii will add points to UCLA (Top 75 add to scheduling bonus).  And then you have San Diego that is #26 in unadjusted RPI - so a small move to #25 or better and UCLA gets the Top 25 win bonus points instead of Top 50.  These are 'lost' points for UCLA when doing RPI Futures, but have a probability > 0 when looking at 1000 simulations.  UCLA's schedule has really taken a hit vs. preseason with Hawaii, Long Beach State, UNLV, & LMU being worse this year.

USC doesn't appear to have those same kinds of variability in gaining/losing points tied to what other teams do.  The big one for them is UNI.  If Northern Iowa can get back into the top 75 in unadjusted RPI than they will get the scheduling bonus - otherwise they will not.  UNI is currently #86 before last night's win.

The one that caught my eye was Illinois and Oregon.  #10 and 11 - yet Oregon is almost twice as likely to be in the Top 4 than Illinois and over 3X more likely to finish in the Top 3.  And yet the SOS between the 2 is almost the same.  Maybe because Illinois has more teams to contend for conference championship and thus a lower probability of winning their conference?

What this says is that Illinois' results are more certain. This make sense as Oregon has a lot more matches that are closer to 50/50 left.
Oregon will play 12 matches with conference teams within 250 Pablo points, Illinois 6. Illinois has 15 matches with teams 500+ points below them, Oregon 8.

If you look at the 2 Pablo-ranked teams above Illinois (Nebraska/Wisky) and below them (MSU/Purdue) in conference, they're only playing each once. You also see the variability on the back end, with Oregon having a higher chance of not being T45.

[trollhunter]

Sept 13, 2017 10:50:00 GMT -5 bluepenquin said:

Sept 13, 2017 9:56:31 GMT -5 Fight On! said:

Thanks blue. Qualitatively, how much impact does the scheduling bonus have on RPI? Does it bump a team up 2 spots? 4? Who knows?

From trollhunters's description below - 1 place is equal to .0014 and 2 places is .0028. Scheduling bonus is 2 places.  The difference in RPI rank depends... Larger impact for teams in the 20-60 range than those in the 10-20 range. You can look at the actual RPI score on the original post to see where .0014 and .0028 would impact in terms of rank.

For this run - it costs UCLA just 1 spot. For Michigan State it would be 3 spots and very close to being 5 spots.  For Kansas and Penn State - it wouldn't have changed their ranking.

BTW, I don't mess with penalty points in calculating RPI Futures.  It is unlikely that it would ever come into play for a team where RPI matters (bids and seeds).  Part of the reason I don't give the RPI rank of all 334 teams.

I don't blame you for ignoring the penalty points now.

It used to matter as some bubblish teams around 50 RPI would occasionally lose to a 150 team. I thought it was a fair penalty for a "bad loss" in RPI. While there is a double penalty for it via "Significant Wins and Losses" category, there is also a chance for double bonus that same way. Seems silly to have changed it - either remove it entirely or leave it at 150 (or 168 for midpoint). Right now it just impacts team at very end of RPI (who really cares if you are 290 or 296 in RPI?)

[jayj79]

Sept 13, 2017 8:08:49 GMT -5 The Bofa on the Sofa said:

Sept 13, 2017 8:00:33 GMT -5 bluepenquin said:

USC doesn't appear to have those same kinds of variability in gaining/losing points tied to what other teams do.  The big one for them is UNI.  If Northern Iowa can get back into the top 75 in unadjusted RPI than they will get the scheduling bonus - otherwise they will not.  UNI is currently #86 before last night's win

UNI took a BIG hit due to their loss to UMKC. We'll see how Pablo responds to the win over ISU. However, I have noticed all year that UNI has been on non-firm foundation. Pablo just isn't sure.

Play a decently tough schedule, pull off a few top 25 upsets, lose to a couple other top 25 teams, and have one bad loss, and the ratings tank thanks to the MVC being full of dead weight.
Fabulous.

[bluepenquin]

Sept 13, 2017 8:08:49 GMT -5 The Bofa on the Sofa said:

Sept 13, 2017 8:00:33 GMT -5 bluepenquin said:

USC doesn't appear to have those same kinds of variability in gaining/losing points tied to what other teams do.  The big one for them is UNI.  If Northern Iowa can get back into the top 75 in unadjusted RPI than they will get the scheduling bonus - otherwise they will not.  UNI is currently #86 before last night's win

UNI took a BIG hit due to their loss to UMKC. We'll see how Pablo responds to the win over ISU. However, I have noticed all year that UNI has been on non-firm foundation. Pablo just isn't sure.

Since beating Iowa State following the unexplained loss to UMKC - UNI beats Nebraska then loses to Kansas State.  I assume they move up in Pablo this week - but they are a very hard team to understand right now?

[The Bofa on the Sofa]

Sept 18, 2017 11:45:10 GMT -5 bluepenquin said:

Sept 13, 2017 8:08:49 GMT -5 The Bofa on the Sofa said:

UNI took a BIG hit due to their loss to UMKC. We'll see how Pablo responds to the win over ISU. However, I have noticed all year that UNI has been on non-firm foundation. Pablo just isn't sure.

Since beating Iowa State following the unexplained loss to UMKC - UNI beats Nebraska then loses to Kansas State.  I assume they move up in Pablo this week - but they are a very hard team to understand right now?

They've moved up from last week, but not a lot. There is no resolution to it - they are all over the place. Pablo basically puts them in the middle of it all. Below USC, Nebraska, Iowa St and above KSU and UMKC. I don't know what else you can really do with them, if you don't ignore things. Then again, remember they also lost to Kentucky and Creighton.

Now, if I were doing the Ultimate Pablo Rankings, they'd be really high. Probably top 10. Although I'm not sure where Nebraska would be. Then again, they would also have to be ahead of Creighton and Kentucky, probably, so maybe not. Maybe they go the other direction?

[c4ndlelight]

That whole Creighton web - Kansas, Kentucky, USC, Iowa St., UNI, Purdue - with so many contradictory results makes it kind of impossible to rank the teams.

[c4ndlelight]

Sept 18, 2017 12:19:21 GMT -5 c4ndlelight said:

That whole Creighton web - Kansas, Kentucky, USC, Iowa St., UNI, Purdue - with so many contradictory results makes it kind of impossible to rank the teams.

I guess you can add Washington, Cal Poly, Wichita to the morass too... just for extra strands of confusion.