I will also say that it's apparent that no one is reading this, Joe.  And there will bet those who still believe the earth is flat (ie the best way to schedule is to schedule to go 8-2, which turns into 5-5 quickly and screws the rest of the conference.)

Posted by Rails on 12/11/2011 11:23:00 AM (view original):

Posted by girt25 on 12/9/2011 8:07:00 PM (view original):
Interesting stuff, rails. I'm pretty sure your hypothesis will end up being proven. Even in other seasons when the non-con performance hasn't been this spotless (well, almost spotless ... coughkylecough), I've watched as our SOS has gotten better and rpi lead widened. There also wasn't any kind of collaborative effort to schedule like this. I never did before getting to the ACC, but the conference is so brutal, I felt like it was the only strategy that made sense. We've had several times where a team that finished 4th or 5th in their own division went on to win it all.

Hopefully I'll remember to check back ...

Thanks girt.  You talk about the brutal conference schedule which it is.  Whether more legitimate or not, the thing is that any human filled conference willing to change their mindset could become a "brutal" conference by winning all of their OOC games.  The ACC has the #1 sos after just two conference games.

I agree that it's possible to make a team/conference look strong than it is through smart scheduling, etc. To an extent we've been able to do that in C-USA Rupp.

But the ACC Allen isn't even close to that. I'd say we easily have 10 of the top 20 most talented teams in Allen, and probably closer to 10 out of the top 15. The absolute dominance of the other BCS conferences each year in the NT further confirms that.

You know I didn't infer that the ACC wasn't deserving--you know that...just that  another conference could achieve similar results if everyone were rowing in the same direction.  And then eventually it might become self-fulfilling. 

Just peeked at your CUSA thread--it'd be really awesome if you all subscribed to this and gave it a shot one year--but it appears you've got a many who just don't have the capability to go 10-0 or would even subscribe to this strategy.  It'd be Awesome if the CUSA could buy into it, but reversing a team's thought process after 2-8 or 4-6 is the difficult part...   

From a few seasons ago, by conference:

Correlation certainly doesn't imply causation, and I don't have the stats knowledge to control that data for things like OOC SOS or team overall rating.  So there's a chance that the obvious statement of "the teams with higher noncon W/L are generally better than the teams with lower W/L records" is also driving final RPI.  But there's something to be said for the fact that non-con W/L and final RPI are correlated so highly.

Rails, what's interesting is that the ACC-Allen's conference rpi is pretty much in line now with where it's been the last several seasons. Not much difference.

Any thoughts/theories on that? At first glance, I think people scheduled weaker this season, which led to the silly non-con record. But the weaker competition seems to be holding us back enough to where we're still in line with previous seasons, where the records weren't quite as gaudy (although admittedly still quite strong), but the competition was tougher.

I've kept out of this thread but I have been following it.  Essentially I know just enough math to know that I don't know enough math to be part of the discussion.

I decided I'd come to post after looking at the conference RPI and noticing it wasn't anything special by recent ACC standards.  Essentially the same post that girt made yesterday.

I've kept an eye on D1 (and D2) the past few seasons to see if I could learn a trick or two.  And while I never bothered to keep records, I don't think this season's ACC non-conference mark is all that impressive relatively speaking.  It is if you focus on the 1 loss.  But I'm 95% sure that the ACC has had 100+ wins multiple times recently in non-conference play.  Without having the ability to do the math, how much difference is there for 12-18 more wins when it comes at the expense of terrible scheduling?  Given how the conference RPI isn't much different than it has been recently, I'd guess I'd argue there isn't much difference.

Ultimately I'm not sure the ACC is the best case study for this, although you got to take what you can get.  I think maybe the NAC from season 51 when we went 101-19 was a better case study.  Unlike the ACC where the 10th best team in the conference is one of the 20 best teams in D1, the NAC that season "only" had 8 teams make the NT and of the 8, I'd argue that only 4, maybe 5, of the teams were among the 20 best in D3 that season and those 101 wins probably got at least a 2-3 teams into the postseason that wouldn't have sniffed it in another conference.

Here are the average RPI and SOS placements (relative to other teams, not the numerical formula).  I took the total and divided by 12.
 GAME                                                  AVG RPI          AVG SOS
After OOC:                                                24.2                130.6                                                                        
After 1 Conf game:                                   16.4                 82.6                       
After 5 Conf games:                                 14.8                 24.5
After 10 Conf games:                               17.5                 12.3                            
After Reg season:                                    19.1                 10.2
After Conf. Tourn:                                        ?                        ?

Average SOS continues to lower and has dropped 92% since OOC play ended.  The rate of decline is really slowing.  You're right Girt.  The average RPI is actually decreasing. 

It does appear that this and previous seasons have resulted in relatively the same rpi as figured by the actual rpi number for the ACC. Maybe it's because there is a limit to how high RPI can actually get (like the QB rating max of 158.2). Would be very interested in knowing what the GAP in RPI was in previous seasons though. That's really what matters, not the actual RPI number. The gap in RPI thru 16 conf games is .0439--the same exact gap after 10 conf. games. Of course the SOS gap is increasing (now a gap of .0359 after 16 conf games compared to .0298 after 10 conf. games). Interesting.

What I have seen and referenced in a previous post is that RPI seems to peak after around 7-9 conference games.  I earlier presumed that is because the 25% part from W% starts to shrink relatively speaking since all conferences have been going .500 for a while.  For example the 119-1 OOC record compared to 60-60 OOC record (a 59 game gap over 120 games) seems like a large gap.  But after 16 conf. games, all conferences have gone .500 and go 48-48 in conf. play.  Percentage-wise the that gap would be smaller than it once was.  For example, a record of 167-49 (119-1 OOC plus 48-48 in conference) compared to 108-108 (60-60 in OOC and 48-48 in conf. play) is a smaller gap because the 59 game gap is now over 216 games, not 120.

So in other words, the gap for the entire 25% (W%), of the RPI formula is shrinking relative to the lead they had after OOC play.  It might also be impacted by OWP and OOWP because after Conf play 65% of all teams' OWP and OOWP is now derived from conference play.  So that could be as much as 66% of RPI being calculated as a zero sum game and by a .500 record and trickeling down-----((.65*.25)+(.65*.50)+(.65*.25)).  Could that be right?  Brain hurts.       

You're right, Hawk.  For the most part, I'm looking at it from a standpoint of helping lower teams and lower conferences become a "power" conference.  In that regard I think it'd be easier for average teams to go 10-0 against weak sim teams than to go 8-2 against better competition.  There'd be too much room for error there. 
Rhetorical:  Because 65% of games are conference games, could it be that an even worse OOC SOS with the same W/L record could achieve the same postseason bids.  It appears that any team with 14 or more total wins will get into the NT.  This season one team, even though the bubble watch and RPI would have them there, will be left out due to only winning 13 and going 13-14.  So they will get 10 to the NT and 2 to the PIT.  And they would have had 12 to the NT if the bottom two teams would have been slightly better and the mid to high teams would be ever so slightly a little worse.  If the bottom two teams could have stolen a total of 2-3 games at the expense of the top to mid level teams it would have happened. 

I think it proves that a mediocre conference that doesn't have the extreme imbalance between the #1 team and #12 team in the conference could get them all to the NT even with a worse OOC SOS if they simply dominate OOC.  At the least, could produce similar results in getting teams to the postseason.  And they wouldn't need an average SOS of 10 between the twelve teams like the ACC will have.  I think this also shows that a high SOS ratings after OOC will quickly lower once conference play picks up, which would not prohibit teams with enough conference and overall wins from making the postseason due to a poor SOS.  Any team in the ACC with 14 total wins (4 wins in conference play or the CT) made the NT.  Even with a worse SOS, it would seem that any team with 5 or more wins would qualify or at least be in the running which is a lot better than most teams in most any conference with only 5 wins. 

{EDITED NOTE}:  The RPI records listed in the forums are after the NT, whereas this thread isn't concerned about postseason success.      

I think that's a big caveat at the end there, rails -- "that doesn't have the extreme imbalance between the #1 team and #12 team". I actually think there's less difference between the #1 team in the ACC (or, as it were, the top few, since they're basically indistinguishable) and #11 and 12 then there is in most other conferences. I think in the majority of conferences you'll find a bigger gap -- a team or two that are just clearly lagging behind. 

The trend that our conf RPI is following this season is the same as it's been -- peaks halfway through and then starts to drop for the same reasons you mentioned. Then, depending on exactly how well we do in the postseason, it starts to go back up again. If our postseason performance is similar to the last couple seasons, I do think that this season will set the new standard for conference RPI.

Great point.  The end of season RPIs listed in the forums have been after the NT.  The ones in this thread are not.  So we'd need to know what they were after the CT for things to be apples to apples. 

From a record standpoint, comparable to the top teams in its conference, they couldn't compete.  The 11th team actually would have made the NT with one more win.  The 13-14 mark was what held them back, not their bubblewatch position or RPI.  At quick glance it appears that there are many conferences with teams with only 2 conference wins.  Again not saying they aren't a good team (a 70 RPI isn't bad and obviously made the PIT).  It's just that getting only 1-2 wins in ACC conference play shows the same gap as any team in any conference with only 1-2 conference wins so record-wise it still shows a pretty big gap between #1 and #12.

free bump for this thread

bump

Bump not sure this got fixed interesting read though

m

RPI is just a very simple equation, so lets look at Conference RPI ... first off, unless the FAQ is wrong, WIS uses 1,2 and 0.8 and not 1.4 and 0.6 for away and home games.  Here is what the FAQ says:

"The RPI (Rating Percentage Index) is a measure of strength of schedule and how a team does against that schedule. It does not consider margin of victory. It is used by the WCAA as one of their factors in deciding which teams to invite to the National Tournament and where to seed them.

The basic formula is 25% team winning percentage (WP), 50% opponents' average winning percentage (OWP), and 25% opponents' opponents' average winning percentage (OOWP). When calculating winning percentage (WP), the RPI weights a road win as 1.2 x a normal win, a home loss as 1.2 x a normal loss, a road loss as 0.8 x a normal loss and a home win as 0.8 x a normal win - neutral site games are not adjusted (weighted at 1.0)"

=======================================
I will use a 28 game schedule.

There are 12 teams in a conference .. and each team will play 16 conference games .. the winning percentage of the conference will always be .500 .. so, the number of games played (in conference) for all teams is: 16 x 12 = 192 and therefore the conference record is 96 W - 96 L.  Then in the CT, there will be 6 wins and 6 losses.  That means every season, every conference will finish with a 102 - 102 record.

Lets say the conference goes 120-0 out of conference (scheduling crappy SIMs).  That makes the absolute best conference record (not including NT or PIT), 222 - 102 and that means that the best winning percentage of any conference is:

Best (Win all OOC):
222 / (222+102) = 0.6852

Worst (Win 50% OOC):
162/(222+102) = 0.500

The problem is that 75% of RPI is made up from your opponents record and your opponents opponents records.

For the sake of argument, I will take the 150 through 159 teams (10 teams for 10 games) and say that every team in the conference scheduled teams of that caliber for all 10 OOC games, I will use Naismith as the example .. the records for 150 to 160 are:

11-16
14-13
14-13
11-17
11-16
11-16
16-12
11-17
14-14
12-16
_____
124-150 = 124 / (124+150) = 0.4526 (for light schedule)

Now lets do the best 10 teams (teams 1-10)and get their Percentage:

26-1
21-6
23-4
26-1
19-8
26-1
25-2
21-6
23-4
25-2
____
235-35 = 235 / (235+35) = 0.8703

For these calculations, lets assume that the opponents opponents are .500 for each group ..

so, lets figure out how much conference games are worth compared to non-conference .. 204 games (102-102)  / 12 = 17 games and there are 10 non conference games.

So for Opponents we would take 10/27*(OOC Win%) + 17/27*(Conf Win%) or :

120 OOC Wins, played teams 150 to 159:

(10/27)*0.4526 + (17/27)*0.6852 =

0.1676 + 0.4314 = 0.599 %

and for worse case 0 OOC wins, top 10 teams:

(10/27)*0.8703 + (17/27)*.500 =

0.3223 + 0.3148 = 0.637 %

Now this makes the overall math do this:

0.685*0.25 + 0.599*0.50 + .500*0.25 = 0.596

or

0.500*0.25 + 0.637*0.50 + 0.500*0.25 = 0.567

So, the bottom line is that going 120 - 0 against terrible teams is not significantly different than going 60-60 against good teams.  And neither seems to work better is better than the top 2 - 3 conferences right now.