Conference RPI

[RoyalShock]

A lot is made of conference RPI as a way to rank conferences by strength. There is also a myth I read from fans on a yearly basis that once conference season starts the league's RPI stays relatively constant since they are playing one another.

75% of a team's RPI is comprised of how their opponents (and opponents' opponents) do. Since roughly a third of a team's games are against non-conference foes, those team's performances in their respective leagues can have a significant impact on the conference RPI.

Case in point, even though there were no MVC games yesterday, the league nearly jumped up to 12th from 13th. (.5103 to the Ivy's .5107) The MVC is actually closer to 9th (Horizon .5226) than 14th (WCC .4939).

Should Valley foes do relatively well I could easily see the MVC jumping back into the top 10 of conference RPI.

[Capitol Shock]

Anything that will help the conference get two teams in.

[SHOXAAC]

Anything that will help the conference get two teams in.

Why stop at 2 teams Cap...don't you have any confidence in the ole' Mo Valley. What a pessimist. You're not a true Shocker fan. I think you've been a closet troll all these years. ;-) Just trying to supply some "cheap entertainment."

[shocker3]

According to Warren Nolan, the Valley has moved past the Ivy league for 12 th place and quickly moving up on 11th place Conference USA.

[Capitol Shock]

Anything but one...however, UNI laid an egg tonight so who knows.

[WstateU]

Anything but one...however, UNI laid an egg tonight so who knows.

... or got beat by a better team?

[SB Shock]

Re: Conference RPI

Should Valley foes do relatively well I could easily see the MVC jumping back into the top 10 of conference RPI.

RPIforcast.com has the MVC expected RPI to be at 10 at season end.

So if the MVC could do well in bracketbusters then the conference RPI might climb a even little higher.

[RoyalShock]

With conference season just now getting undeway, the RPI is still fairly fluid (MSU jumped nearly 20 spots after their win at UNI), making RPI forecasts unreliable. I'll give it more credence come late January.

That said, I think the forecast is probably correct, even if it is predicated on shaky (at this point) data.

Just last night the gap between the MVC and the 9th-place conference got significantly smaller.

[SB Shock]

With conference season just now getting undeway, the RPI is still fairly fluid (MSU jumped nearly 20 spots after their win at UNI), making RPI forecasts unreliable. I'll give it more credence come late January.

That said, I think the forecast is probably correct, even if it is predicated on shaky (at this point) data.

The http://www.rpiforecast.com/index2.html website has made the day-2-day RPI site meaningless for anything but day to day snapshot discussion.

The RPI forcecast is actually calculating the complete season interconnectedness that are not even being considered in the "real time RPI" and really eliminate alot of they day to day fluctuations you are going to see in the Warren Nolan or Live RPI sites.

Case in point look at WSU RPI over the season, the blue line is the day-2-day while the red line is EXPECTED. The blue line has ranged from both side of the 95% confidence limits, while the red has pretty much stayed fairly consistent.

I expect the blue line to bounce around the red line for the foreseeable future as they play some bad RPI teams countered with some good RPI teams.



The real unknowns is bracket buster effects and how those teams that make up WSU RPI do the rest of the way (209 teams). The Red Line is slighlty trending up because overall, WSU opponents and opponents of opponents have been doing slightly better than predicted.

In some ways it kind-of takes the fun out of looking at day-2-day RPI because it is pretty much determined what WSU needs to do to have a good RPI.

0 more losses - 14 RPI
1 more losses - 17 RPI
2 more losses - 23 RPI
3 more losses - 29 RPI
4 more losses - 37 RPI
5 more losses - 45 RPI
6 more losses - 56 RPI

[RoyalShock]

But isn't the forecast based on the results of games that haven't been played yet? I just don't see how, barely one-third of the way into the season, it can be close to accurate, unless a significant majority of games play out exactly as the RPI currently sits.

Look at how big UNI and MSU's RPI changed after last night's game. They were 40 spots apart and the cumulative change was about 50 spots (MSU moved up 30, UNI dropped 20).

That's a lot of fluidity for a predictive system based on current ratings to be relied upon.

And to keep to the subject of the thread, the MVC is back down to 13th so far tonight.

[SB Shock]

But isn't the forecast based on the results of games that haven't been played yet? I just don't see how, barely one-third of the way into the season, it can be close to accurate, unless a significant majority of games play out exactly as the RPI currently sits.

For real time RPI your going to see these fluctuations day to day because each time you play somebody you are adding them to the mix, plus their opponents. You are also seeing the weighting of each team effect on your RPI change.

But when you start looking at Conference RPI, pretty much the conference bed is made as 95% of the non-conference schedule is complete. The fluidity you will now will see is conference opponents playing conference opponents and getting the interconnectedness of opponents SOS and opponents opponents SOS.

MVC is going to climb over Horizon because their Out of Conference (OOC) W-L is better than Horizon even though MVC played a weaker schedule. Conference USA has the 28th rated SOS, but they are going to be higher as conference because they won 67% of their games.

MVC has a slight edge on the A-10 on because of their OOC W-L, but A-10 is projected to stay above them (at 9th) because they OOC SOS is rated 2nd while the MVC is 25th. But bracketbuster could give the conference just enough boost (or maybe not with the realities).

Look at how big UNI and MSU's RPI changed after last night's game. They were 40 spots apart and the cumulative change was about 50 spots (MSU moved up 30, UNI dropped 20).

This should be no suprise, MSU. UNI is their 2nd best current "Helper" opponent in their RPI. Of course your going to see them rise significantly, they finally got them in their RPI calculations now. Overall MSU real-time RPI is underpeforming where it should be at season end.

That's a lot of fluidity for a predictive system based on current ratings to be relied upon.

Of course there is going to be some fluidity in the predictive system. In case of WSU the system initial conditions were based I believe on where they were last year. But as the system gets additional data that shows WSU is better (the weight of those initial conditions is getting removed almost completely) you find that WSU expected RPI has trended up, meanwhile team that were better than last are trending down.

That is why the author runs 10,000 simulations and provides the probabilities and has a statistical basis. You can look at the histogram and see where you are most likely to be, or where you could be if things go for a "Cinderella" season.

Ultimately it is a "tool" for the fan to really see where their current RPI is where they could ultimately be. It also really shows what the true margin of error is for a team like WSU.

If WSU had lost to Evansville there would be some who would have said WSU is done for an at-large chance, but that wouldn't be true as WSU margin is greater than 1 game (to me it looks like WSU can lose up to 4 games and still have a shot at an at-large).

And to keep to the subject of the thread, the MVC is back down to 13th so far tonight.

Which is to me is meaningless, because I don't see