Our own RPI oddities (long)

[kwyjibo]

In order to figure out what SLU's NCAA chances are (the usual disclaimer about not PREDICTING anything here just exploring the mathematical possibilities) I have been playing around with forecasting our RPI.

Now, I have to start off with a few caveats though. With a balanced schedule and no non-conference games you can forecast pretty accurately what final record is necessary for a NCAA bid. (Although non-conference opponents going on hot streaks in their own conferences like Kennesaw St and UWM can change things a bit, but it usually washes out).

However, the A-10 schedule is not balanced and more importantly the RPI adjusts by 40% the win/loss of home/road. Both of these things makes it much more difficult computationally to figure out exactly what record is necessary. So much so it may not be worth doing.

Yet I decided to do a rough calculation (at least for myself) and what I found out was interesting. One of the reason different sites has SLU at different RPI's is that some sites do not make the home/away adjustments (Ken Pomeroy does this correctly in theory but you'll see in a second a problem). Also, some might not compute them correctly (Ken Pomeroy computes correctly) so that would explain some differences (for example, one differnce is that some sites use games against provisional D-1 teams incorrectly). But Ken Pomeroy gets his raw data from other places and in the Billikens case he may be artificially boosting our RPI.

He has included our two Chuck's Family Arena games as "neutral" games (i.e. like a tournament game). This means that our two victories count 40% more than they probably should in Pomeroy's calculation. This will be fine if the NCAA statmeisters do the same but I get nervous that they will do a little more work (this is not a criticism of Pomeroy as he is dealing with A LOT of data and does not have time to do this kind of little stuff).

So, my calculations keep the SCFA games as neutral but remember that may be optimistic. I assumed for simplicity that our SOS will not change. This may be a stupid or smart assumption based on how the rest of the year works out but I cannot really do any more detailed work on it. Casually, we have a good SOS now and will likely decline slightly through the rest of the year (although Dayton/X/Rhody or any non-conf opp hot streaks will definetly help).

Yet another caveat is that I am not looking at "profile" details. SLU already has two top 50 victories (one on the road) and I assume we get at least 2 more top 100 victories.

The other obvious caveat is that the conference tournament can really help a team that runs to the final and loses. I will not forecast that at all.

With all the many caveats in mind I figured that going 9-2 or better will put us in with ease. Also 6-5 will keep us in the NIT (more on the NIT "Jurassic Park" selection committee later but basically we have a near lock in the NIT even with a 15-14 record).

OK things get more interesting with 8-3 and 7-4 depending on the road home split. What is counterintuitive is that HOME wins are better GIVEN a fixed record. That is, if you had could pick 7 Bills wins for the rest of the season you would be better off picking HOME games (it is complicated but it boils down to the fact that your penalized more for losing at home than rewarded for winning on the road IF you are winning team). Road victories do count more in the "profile" of "good wins", particularly against good teams.

SO, 8-3 with 6 home wins is progressively better than with 5 homers, 4 homers, and then 3 home wins by enough that I would forecast 8-3 with 6/5/4 wins as tournament material but 3 would put us on more on very iffy ground with a tourney run necessary.

7-4 with 6 home wins is the only one I would say gets us in. I would also say that 7-4 with 2 home wins keeps us out. 7 wins/5 home puts on high bubble and 7 wins/3 home means we need a run and other profile elements to be in place.

So, the easiest road to the tournament will be completely defending at Savvis and winning one more on the road. That and no big changes in the "other factors". Go Bills.

[kshoe]

for all the work.

A small corection to what you said, Pomeroy has been treating the two St. Charles games as "semi-home" games. not sure what that means as far as the 1.4/0.6 goes but I know thats not the same as a neutral game.

You said: I assumed for simplicity that our SOS will not change. This may be a stupid or smart assumption based on how the rest of the year works out but I cannot really do any more detailed work on it."

I think this is the biggest and most dangerous assumption you make. While you are right that the remainder of the A-10 schedule won't really decrease us much when you compare to the non-confernce and beginning of A-10, where the difference will be is all the other teams around the country will be improving dramatically. All the MU/Pitt/Syracuses of the world that played soft non-conference schedules will see their SOS increase as they play out the tougher portion of their scheule. Our SOS will most likely go down as compared to other schools.

I still think 9-2 plus a decent showing in the A-10 gets it done. Those that say our ONLY chance is to win the A-10 tourney aren't really thinking it out.

[kwyjibo]

>A small corection to what you said, Pomeroy has been

>treating the two St. Charles games as "semi-home" games. not

>sure what that means as far as the 1.4/0.6 goes but I know

>thats not the same as a neutral game.

What I said above is correct. A win counts .6/1/1.4 depending on site (home/neutral/away). While he marks it as "semi-home" in the text mathematically he weighted it as 1 (one of the things I had to compute was our current win percentage and the only way to square with Pomeroy's number is to weight the two SC Family Arena games as one.

You are of course right in some sense that the ranking of many teams will move around but I did not use a "relative ranking" estimation method. For simplicity and increased accuracy (I have done this for 6 years and one of the things I have learned is that .586 is more accurate than projecting into the current distribution, probably for the reasons you mention). I used a fixed RPI number as the goal (.586) which has been amazingly stable over the course of the RPI (it is almost always in the 38-44 range at the end of the year and I assume the same this year).

The other thing is that I was much fuzzier with my description than usual to reflect the fuzziness of the estimate. However, the one thing I want to get across is that we do not need to get to 9-2. In fact 7-4 will get us a look but then a lot of other things have to go our way as well. Obviously every victory is a good thing but defending home court, winning one on the road and one in the conference tourney stands a pretty good chance of being enough.

We will be in the NIT if we do not make the dance and finish over .500.

[kwyjibo]

I went back and checked my work and noticed a small mistake that makes me a smidgeon less optimistic. I think 7-4 gets a look but we would need some other things (like an A-10 tourney run of 2-1, 3-1). Essentially I would like to be a tad less optimistic than above.

This also means that 8-3 gets a good hard look but in the unlikely event we do this while going 3-3 at home (5-0 road) I would not like our chances unless we got a deep tourney run and other breaks.

[courtside]

There is a lot of assumption that SLU will get a fair honest look vs a bigger conference, bigger name school. Looking at SLU's schedule...16-12, based on 7-4...finishing worse than top 2 in conference....and somehow you think SLU has a chance to dance without winning conference tourney? Based on conference strength, non-conference results, other intangibles...I don't think so...just my opinion. We agree to disagre.. NIT with winning record...I'll buy that possibility.