Historical RPI vs KenPom on Selection Sunday

[Jamar Howard 4 President]

This topic came up in another thread. I put together more data and found the following.

Here are the last 10 years of RPI and KenPom ranks as of Selection Sunday. Includes all MVC teams who finished top 100 in both ranking systems.

Year	Team	RPI	KenPom	Difference	% Difference
2015	UNI	14	13	-1	-8%
2015	WSU	17	14	-3	-20%
2015	Illinois St	62	60	-2	-3%
2014	WSU	4	5	1	20%
2013	Creighton	24	17	-7	-41%
2013	WSU	37	35	-2	-6%
2013	Indiana St	72	98	26	27%
2013	UNI	83	70	-13	-19%
2012	WSU	12	9	-3	-33%
2012	Creighton	24	38	14	37%
2012	UNI	72	91	19	21%
2011	Missouri St	43	77	34	44%
2011	WSU	60	56	-4	-7%
2010	UNI	17	21	4	19%
2010	WSU	43	67	24	36%
2010	Illinois St	76	82	6	7%
2010	Missouri St	90	86	-4	-5%
2009	Creighton	40	62	22	36%
2009	Illinois St	47	56	9	16%
2009	UNI	59	76	17	22%
2008	Drake	12	12	0	0%
2008	Illinois St	37	43	6	14%
2008	Creighton	20	23	3	13%
2008	SIU	60	56	-4	-7%
2007	SIU	7	16	9	56%
2007	Creighton	20	23	3	13%
2007	Missouri St	36	36	0	0%
2007	Bradley	39	54	15	28%
2007	UNI	86	74	-12	-16%
2007	WSU	100	72	-28	-39%
2006	WSU	19	37	18	49%
2006	Missouri St	24	28	4	14%
2006	UNI	30	31	1	3%
2006	SIU	31	35	4	11%
2006	Bradley	32	23	-9	-39%
2006	Creighton	41	51	10	20%
	Avg	41.4	45.8	4.4	7%

I find it interesting that the RPI tends to give a more favorable rating than KenPom. Note that these systems utilize totally different formulas. The RPI is totally wins and losses based. KenPom ignores wins and losses and calculates offensive and defensive efficiency based on number of points scored, allowed, number of possessions, and quality of opponent.

The argument elsewhere was that WSU would have better RPI's if they played in better conferences. Based on this data, this seems doubtful, unless you think that KenPom not only also penalizes MVC teams, but that it is even more penalizing than the RPI.

[WUpigsooie]

Where did you find this historical info? In have a hypothesis but don't want to burden anyone else with the research required.

[Jamar Howard 4 President]

http://realtimerpi.com/2005-2006/rpi_Men.html

Just change out the years in the link to find what you are looking for.

For KenPom.com, at the top of the page he has a link to data in spreadsheet format. You can get this info for end of season, or in the case of this thread, "pre-tourney", which means as of selection sunday.

[ShockBand]

Not related to your question, but man did we have a snootfull of good RPI's in 2006 or what? Wasn't that back in the day when the whiney P5 conferences accused the Valley of "gaming" the system?

[jdshock]

For KenPom.com, at the top of the page he has a link to data in spreadsheet format. You can get this info for end of season, or in the case of this thread, "pre-tourney", which means as of selection sunday.

Do you have to subscribe to kenpom to see that? I still can't find it.

[Jamar Howard 4 President]

@jdshock and @WUpigsooie:, unfortunately it appears you need to be a member to see the links to the spreadsheet data. I logged out and the links disappeared.

Here is what I see as a member.

[Kung Wu]

The average distance isn't that telling. Calculate the average "error" per team.

Error = (RPI - KenPom) / KenPom

Then you can average that over the whole set.

[Cdizzle]

Now we know. It's official. Both data sets are rigged to screw WSU.


I knew it.

[Jamar Howard 4 President]

The average distance isn't that telling. Calculate the average "error" per team.

Error = (RPI - KenPom) / KenPom

Then you can average that over the whole set.

Done. Original post now has that data. Doesn't change my conclusions one bit.

[Kung Wu]

Done. Original post now has that data. Doesn't change my conclusions one bit.

No, but it will be more "correct".

[Jamar Howard 4 President]

No, but it will be more "correct".

If you rank a team 20, and I rank them 30, we have a 50% difference.
If you rank a team 2, and I rank them 3, we still have a 50% difference.
However, I would argue that the 20vs30 dispute is much more substantial than the 2vs3 dispute.

I agree that there are also problems with focusing purely on difference, as the difference between 2 and 22 seems much greater than between 102 and 122.

Ideally, you would probably find a formula to blend the two comparison methods. For now, having them both side by side is probably good enough.

[Cdizzle]

Done. Original post now has that data. Doesn't change my conclusions one bit.

Mine either, for the record. In fact, now it looks even more fancier. When I added it all up, it showed that I was 289% correct.

Seemed high initially, but then I got to thinking about just how right I often am, and I decided it was probably pretty close.

[Cdizzle]

It would be interesting if we had the S-Curve data from tournament seeding to compare this against. Just using the seeds could be somewhat useful, but would include a pretty high variance wherein the difference between a #5 and a #6 could be adjacent teams (#20 and #21) or teams nearly on different levels (#17 and #24).

However, as an educated guess based on how right I was earlier (289%!), I'll wager WSU gets screwed on the S-Curve even in comparison to the screwed they got in the RPI and KenPom.

[ShockdaWorld]

WSU 2015 should be 20%, not 2%. Not that it makes a difference in the grand scheme of things. I just see numbers and my mind kicks in. Stupid math major.

[Jamar Howard 4 President]

WSU 2015 should be 20%, not 2%. Not that it makes a difference in the grand scheme of things. I just see numbers and my mind kicks in. Stupid math major.

Fixed. Thanks for catching that.