Projecting the RPI

[GoneGolfin]

It is interesting to think about where the Cajuns' Softball RPI and RPI rank might be, depending on how they finish the season. There are many variables at play and you must choose which ones to hold constant in order to obtain a reasonable gauge. The following represents one approach, with assumptions documented. Once an Adjusted RPI figure is determined, you can then compare against the current Adjusted RPIs as well as last year's Adjusted RPIs at selection time to obtain an idea on ranking.

Assumptions:

I am starting with the Cajuns' WP (.9677) and OWP (.5467) through today's results (Sunday 4/1/12) and the OOWP (.5505) from last week (3/25). Once I have updated figures, I can calculate the new OOWP for 4/1/12.

I do not make any assumptions with respect to how non-conference Cajun opponents will play the remainder of their schedule. IOW, I am assuming the same individual OWPs from each team through NCAA selection time.

I am holding the OOWP constant at .5505.

All remaining games on the schedule are played.

For conference opponents, we know ... 1) How they will perform in games vs. the Cajuns as we are projecting those ... thus we include those results in the projections ... and 2) We know that outside of the games vs. the Cajuns, the other conference members will play .500 ball in aggregate against each other. Thus, I attempt to allocate these wins/losses as evenly as possible. How these games actually transpire in the end does matter (because percentages are calculated for each team ... not in the aggregate), but the difference from the projection should not be too significant.

I do not project any of the remaining non-conference games for the conference members.

I assume the RPI bonuses for the Cajuns are unchanged from this point foreward ... .0156 overall and .0130 for non-conference.

Projection:

The Cajuns have one remaining non-conference game, this Wednesday vs. McNeese State. For projection purposes I am tagging this as a win. This leaves six new (schools currently not participating in the Cajuns' OWP) schools, all conference games. I list all of the remaining games below along with the contributing OWP to the Cajuns' schedule.

McNeese State (.7000) - 1 game
South Alabama (.7188) - 3 games
Florida International (.5000) - 3 games
Middle Tennessee (.4000) - 3 games
Florida Atlantic (.3056) - 3 games
Western Kentucky (.5972) - 3 games
Louisiana-Monroe (.5517) - 3 games
---
Aggregate .5589 remaining OWP

However, since the Sun Belt in aggregate is an over .500 league, simply playing the remaining conference games will lower this aggregate OWP of Cajun conference opponents.

Obviously, it would be a very good thing if the Middle Tennessee and Florida Atlantic series were rained out. :-)

After applying the conference wins and losses evenly in the non-UL games (to all conference members), the aggregate OWP above is lowered to ... .5221. If we include Troy and North Texas ... teams the Cajuns have already played, this becomes .5346.

We now have everything we need for the various components of the formula.

All of the above (including the assumptions) leaves the Cajuns with a .5332 OWP, down from the current .5467.

So, we use ...

OWP = .5332
OOWP = .5505

We can then project whatever we want for wins/losses over the remaining conference games.

1) 18-0 -> 49-1 overall record -> WP = .9800

WP = .9400
OWP = .5332
OOWP = .5505
--
Base RPI = .6492
Adjusted RPI = .6648

2) 17-1 -> 48-2 overall record -> WP = .9600

WP = .9800
OWP = .5332
OOWP = .5505
--
Base RPI = .6442
Adjusted RPI = .6598

3) 16-2 -> 47-3 overall record -> WP = .9400

WP = .9400
OWP = .5332
OOWP = .5505
--
Base RPI = .6392
Adjusted RPI = .6548

4) 15-3 -> 46-4 overall record -> WP = .9200

WP = .9200
OWP = .5332
OOWP = .5505
--
Base RPI = .6342
Adjusted RPI = .6498

5) 14-4 -> 45-5 overall record -> WP = .9000

WP = .9000
OWP = .5332
OOWP = .5505
--
Base RPI = .6292
Adjusted RPI = .6448

6) 13-5 -> 44-6 overall record -> WP = .9000

WP = .8800
OWP = .5332
OOWP = .5505
--
Base RPI = .6242
Adjusted RPI = .6398

Obviously each loss costs the Cajuns .0050 in RPI since each loss lowers their WP by .02.

Now you can compare these Adjusted RPIs to 1) the current Adjusted RPIs and 2) the Adjusted RPIs at selection time in 2011 ...

3/25/12 Adjusted RPIs

(Adjusted RPI in parentheses)
#1 Alabama (.7000)
#2 Florida (.6879)
#3 California (.6878)
#4 Tennessee (.6845)
#5 Arizona State (.6790)
#6 Louisville (.6777)
#7 Louisiana (.6654)
#8 Texas (.6615)
#9 Washington (.6613)
#10 Oregon (.6561)

2011 Adjusted RPIs at NCAA Tournament Selection Time

#1 Arizona State (.7080)
#2 Georgia (.7059)
#3 Texas (.7054)
#4 Alabama (.7014)
#5 Missouri (.6978)
#6 Florida (.6962)
#7 California (.6946)
#8 Arizona (.6876)
#9 Oklahoma (.6743)
#10 Michigan (.6673)
#11 Baylor (.6658)
#12 Tennessee (.6601)
#13 Washington (.6563)
#14 Texas A&M (.6482)
#15 Oregon (.6470)
#16 Nebraska (.6407)

The significantly higher Adjusted RPIs at selection time in 2011 vs. 3/25/12 are due to the RPI bias I have mentioned (both in the base RPI formula and the awarding of bonuses). The elite conferences feast on the RPI during conference play as their RPI is enhanced due to the heavy weighting of the OWP (and SOS) and the bonuses awarded for all wins against the Base RPI Top 25/50 ... including conference games and home games ... not just non-conference road games. Arizona State above had 21 bonus games at selection time in 2011 and Georgia had 24. How many schools outsides of the Pac-12, SEC, Big XII, and Big Ten (four conferences) do you see represented above in the Top 16? How about zero. In fact, 15 of the 16 schools above came from three conferences.

Brian

[Turbine]

Good stuff Brian.

Great way to wake up on a Monday (or in your case, finally fall asleep)

Thanks

[nathancpike]

Good stuff Brian.

Great way to wake up on a Monday (or in your case, finally fall asleep)

Thanks

My head hurts now.

[GoneGolfin]

Part II
I did not want to inundate folks with a five page post, do I decided to break it up. This section focuses on the Non-Conference RPI.

The same assumptions apply as with the overall RPI, except .5444 is the OOWP that is used. The Cajuns' non-conference OWP is currently .5316. As stated earlier, I am using the projection of a win over McNeese State this Wednesday. This puts the Cajuns at 26-0 non-conference and a perfect WP component (1.0000).

WP = 1.0000
OWP = .5381
OOWP = .5444
--
Base RPI = .6552
Adjusted RPI = .6682

Now you can compare these Non-Conference Adjusted RPIs to 1) the current Non-Conference Adjusted RPIs and 2) the Non-Conference Adjusted RPIs at selection time in 2011 ...

3/25/12 Non-Conference Adjusted RPIs

(Adjusted RPI in parentheses)
#1 Alabama (.7018)
#2 Louisville (.6677)
#3 California (.6661)
#4 Hawaii (.6650)
#5 Louisiana (.6615) would now be .6682, good for #2
#6 Arizona State (.6581)
#7 Tennessee (.6578)
#8 Florida (.6566)
#9 Texas (.6540)
#10 Oregon (.6498)

2011 Non-Conference Adjusted RPIs at NCAA Tournament Selection Time

#1 Alabama (.7181)
#2 Florida (.6952)
#3 Georgia (.6872)
#4 Arizona (.6677)
#5 Michigan (.6610)
#6 Texas (.6596)
#7 Washington (.6479)
#8 Oklahoma (.6451)
#9 Arizona State (.6422)
#10 Missouri (.6409)
#11 California (.6371)
#12 Baylor (.6281)
#13 Louisville (.6258)
#14 Long Beach State (.6209)
#15 Houston (.6207)
#16 Georgia Tech (.6196)

Using last year's list at selection time, the Cajuns would have finished with a Non-Conference RPI rank of 4 and the first undefeated non-conference record since Tennessee in 2007.

Does everyone notice the difference between the Top 16 non-conference RPI list above and the Top 16 RPI list (all games below)? Look at the schools that fell out and the ones that moved in. Also look at the differentials between the actual Adjusted RPI in All Games vs. just Non-Conference games. The RPI bias in action.

Of all 16 schools below, the only school that had a higher Non-Conference RPI vs. All Games Adjusted RPI was Alabama. Meanwhile, the gap between these two RPIs was quite significant for most schools, especially the Pac-12 schools.

2011 Adjusted RPIs at NCAA Tournament Selection Time

#1 Arizona State (.7080)
#2 Georgia (.7059)
#3 Texas (.7054)
#4 Alabama (.7014)
#5 Missouri (.6978)
#6 Florida (.6962)
#7 California (.6946)
#8 Arizona (.6876)
#9 Oklahoma (.6743)
#10 Michigan (.6673)
#11 Baylor (.6658)
#12 Tennessee (.6601)
#13 Washington (.6563)
#14 Texas A&M (.6482)
#15 Oregon (.6470)
#16 Nebraska (.6407)

Brian

[CajunAmos]

Brian, I know it would probably be useless, but I was thinking.......This is Scott Farmers last year on the NCAA softball committee. Couldn't he present this obvious unfairness to the committee?

[GoneGolfin]

Brian, I know it would probably be useless, but I was thinking.......This is Scott Farmers last year on the NCAA softball committee. Couldn't he present this obvious unfairness to the committee?

I do not see why not. But he needs to be willing to dig and understand the problem, such that it could be presented properly.

Brian

[CajunAmos]

I do not see why not. But he needs to be willing to dig and understand the problem, such that it could be presented properly.

Brian

I kind of felt the same way a couple of years ago when illegal bats were "grandfathered" into the new bat standards, allowing the same conferences you referenced above the advantage of a "hot" bat and an unfair advantage. This finally got changed so all are on the same playing field. I'm just hoping.

[CajunRebel]

I nominate GoneGolfin for King of raginpagin.

[CajunEXPRESS]

Brian, I would think the RPI could be fine tuned to be a better tool. Wins should count more period. Just a guess, but move wins from 25% to 30% and move that second group from 50% to 45% there would be an important improvement. If that does not cull out the middle fo the pack losers from the power conference sweet spot move it a little more.

I just cannot ever get over a team moving up the RPI while losing. That is just crazy.

[GoneGolfin]

Part II
Does everyone notice the difference between the Top 16 non-conference RPI list above and the Top 16 RPI list (all games below)? Look at the schools that fell out and the ones that moved in. Also look at the differentials between the actual Adjusted RPI in All Games vs. just Non-Conference games. The RPI bias in action.

Here is a more clear breakdown of the differential between the "All Games Adjusted RPI" and the "Non-Conference Adjusted RPI" at NCAA selection time in 2011. The rankings below are for "All Games Adjusted RPI", not "Non-Conference Adjusted RPI".

2011 Adjusted RPIs at NCAA Tournament Selection Time (Overall) vs. (Non-Conference)

#1 Arizona State (.7080) vs. (.6422)
#2 Georgia (.7059) vs. (.6872)
#3 Texas (.7054) vs. (.6596)
#4 Alabama (.7014) vs. (.7181)
#5 Missouri (.6978) vs. (.6409)
#6 Florida (.6962) vs. (.6952)
#7 California (.6946) vs. (.6371)
#8 Arizona (.6876) vs. (.6677)
#9 Oklahoma (.6743) vs. (.6451)
#10 Michigan (.6673) vs. (.6610)
#11 Baylor (.6658) vs. (.6281)
#12 Tennessee (.6601) vs. (.6146)
#13 Washington (.6563) vs. (.6479)
#14 Texas A&M (.6482) vs. (.5877)
#15 Oregon (.6470) vs. (.6108)
#16 Nebraska (.6407) vs. (.6161)

Brian

[wcd35]

That's pretty ridiculous. Some of them have a chasm between NC and Overall. Its as major conference-leaning as the BCS

[GoneGolfin]

Brian, I would think the RPI could be fine tuned to be a better tool. Wins should count more period. Just a guess, but move wins from 25% to 30% and move that second group from 50% to 45% there would be an important improvement. If that does not cull out the middle fo the pack losers from the power conference sweet spot move it a little more.

While I agree that winning must count for more, fixing the RPI is much more complicated than "fine tuning" it. Fine tuning it will not get you the results needed. Other than the problems with component weighting, the lack of distinction between home/road, and an arbitrary bonus system, you have a very significant flaw in how you calculate the strength of schedule component (SOS) ... that together with the WP (winning component) forms the RPI.

This flaw is what I termed the "arbitrary levels of recursion" used to calculate the SOS component. The discussion that made CajunRed puke. Two levels is not nearly enough (OWP and OOWP). This significant shortcoming gets exposed in many ways. But just one example is ...

You have an opponent on your schedule (let's call them South Alabama for simplicity) that is obviously a component of your OWP (and thus South Alabama is very important). Whom South Alabama plays on their schedule is of less consequence to you (this represents your OOWP, while it represents South Alabama's OWP). But that opponent's SOS is not represented in your RPI. It is only represented partly in South Alabama's OOWP. But South Alabama's performance against that team will certainly affect your RPI. So, what we have is a win/loss result that affects your RPI without providing any input as to the real strength of that opponent (as measured by their SOS).

Once you identify the above as a fundamental flaw of the RPI ... and then set out to implement a real fix ... you end up with an entirely different system that bears little resemblance to the RPI. So let's call it like it is. A new system is needed.

Could you make it more tolerable by making some tweaks? Yes. But it would still be a bad an unfair system. Why not fix it right the first time?

I just cannot ever get over a team moving up the RPI while losing. That is just crazy.

Yes, this is where the RPI shows its a$s.

Brian (I found a hole in Turbine's language filter)