Attention National Media on GM

[Jamar Howard 4 President]

So long as they lean on the RPI as a reliable tool, we will get screwed. The RPI needs to be dissolved.

2016 RPIs
South Dakota St - 29
Akron - 34
St Mary's - 38
Princeton - 39
San Diego St - 41
Yale - 43
UNC Wilmington - 44
Arkansas Little Rock - 45
WSU - 47 (which is very fair pre adjustments for injuries, eye test, etc.)

I am not a proponent of giving the RPI more weight than it has today, but clearly, leaning harder on the RPI would help mid-majors, not hurt them.

The RPI is not the problem.

[ShockdaWorld]

The problem is the obvious, oblivious P5 bias. It doesn't matter what metric you use, as long as it's left up to a bunch of honks that are forced to have absolutely zero transparency and accountability, it will remain that way.

[shoxlax]

2016 RPIs
South Dakota St - 29
Akron - 34
St Mary's - 38
Princeton - 39
San Diego St - 41
Yale - 43
UNC Wilmington - 44
Arkansas Little Rock - 45
WSU - 47 (which is very fair pre adjustments for injuries, eye test, etc.)

I am not a proponent of giving the RPI more weight than it has today, but clearly, leaning harder on the RPI would help mid-majors, not hurt them.

The RPI is not the problem.

Top 50 wins?

[Cdizzle]

Top 50 wins?

Don't make me go on a rant about this again.

[ShockerFever]

Top 50 wins?

I lol'ed.

The hypocrisy of the selection committee and P5-toting media honks is pure comedy.

[AZ Shocker]

If CBS were smart they would bring Gregg in for some studio analysis

Sure would enhance recruiting. P-5 schools and CBS probably wouldn't allow it though. Be awesome though!

[CBB_Fan]

Just for clarity's sake, I want to dispel the notion that "RPI" alone is the metric that hurts mid-majors. No, RPI is basically not a metric to the selection committee. Instead, the only real number they looked at was wins vs the top 100. This is provable. I lined each team up according to the S-Curve and compared it to a variety of metrics. The single best metric was Wins vs the top 100, which gave an average seedline difference of 1.15 from the actual bracket. RPI was 1.65, SOS 2.4, wins vs Top50 1.79. Winning percentage versus the top 100 had a deviation of 2.25.

Essentially, this means that a mid-major could be a quality OOC opponent to a major conference team without being a threat to be an at-large, as they'd go back to a conference with no or very few chances to get wins. It, more than any other metric, essentially grades teams based on their conference composition. It becomes very important to have 3, 6, 10 teams in conference above 100 RPI. And believe it or not, most of the seedline discrepancy is actually higher in the brackets.

BTW, if we had been gifted the route of Syracuse we'd be in the Sweet 16 with really good odds to make the Final Four. If they had our route, I think they'd have had a 50/50 shot of losing in the first round. This isn't really pertinent, but it shows seeding matters a LOT.

[Cdizzle]

Wins vs. the Top 100 of what?

[CBB_Fan]

RPI.

[Cdizzle]

Your honor, I rest my case.

[Kel Varnsen]

[Jamar Howard 4 President]

The single best metric was Wins vs the top 100.

Wins vs. the Top 100 of what?

RPI.

Your honor, I rest my case.

This is not to pick on you @Cdizzle, because I hear that argument ALL THE TIME, but the truth is that the difference in metrics, when used purely for grouping purposes, is extremely minimal.

For one single team, like WSU, there can be large differences (11 KenPom vs 47 RPI, if I remember right). However, once you start grouping 30+ games, the categories end up looking almost identical regardless of which metric you use. For every opponent that is overrated by one metric, you are likely to find another opponent on the schedule that was underrated. Give me a second to put together a follow-up post with some data and I'll show you what I mean.

[Jamar Howard 4 President]

Consider the following as of selection Sunday looking at D1 opponents only:

2016 WSU RPI – Ranked #47
1-2 vs 1-50
3-5 vs 51-100
8-1 vs 101-200
11-0 vs 200+

2016 WSU KenPom – Ranked #11
1-2 vs 1-50
3-5 vs 51-100
11-1 vs 101-200
8-0 vs 201+

Despite WSU being #11 in KenPom, #47 in RPI in their own rank, the ranking groups look nearly identical. Apparently Loyola, who the Shox beat 3 times, was top 200 in KenPom, 200+ in RPI, but that was the only difference. Sure, in some cases teams would see a difference across more important thresholds, like top 50 in one metric, but not in the other, but more often than not, those things average out. Let’s not forget that WSU was an outlier individually, and most teams aren’t that far apart between metrics. Yes, the best possible metric should be used, and yes, the RPI is less than perfect, but if you move past an individual team’s rank, which the committee says it does, and merely use the RPI for grouping purposes, then at that point switching out the RPI for KenPom wouldn’t make much difference. A small difference, yes, but not nearly as much as most people think based on seeing WSU, as an individual team, ranked so differently.

Here’s one more example

2016 Kansas RPI – Ranked #1
15-3 vs 1-50
5-0 vs 51-100
6-1 vs 101-200
3-0 vs 200+

2016 Kansas KenPom – Ranked #2
13-3 vs 1-50
8-0 vs 51-100
5-1 vs 101-200
3-0 vs 200+

Different? Yes. Wildly different? No. RPI shows 2 more top 50 wins (15 vs 13), but 1 less top 100 win in total (20 vs 21). Losses are grouped identically.

The problem is not using RPI to group teams. Using KenPom would make little difference. The problem is a failure to properly evaluate the differences between "easy" and "hard" schedules. Is 3-9 vs the top 50 better than 11-1 vs 101-200? Regardless of the metric you use, if grouping teams, how you answer those questions is the big issue. Right now, there is too much favoritism for winning 20-30% of games vs elite teams, and not enough favoritism for winning 90% of games vs the 101-200 range by 19 points per night. Until that is fixed, no change in metric is going to make any significant difference.

[Jamar Howard 4 President]

One more post with stats and then I'll wait to hear everyone else's thoughts.

Big 12, teams ranked top 50, 51-100, 101-200, 201+, as of selection sunday
RPI - 7, 1, 2, 0
KenPom - 7, 1, 2, 0

Same breakdown for MVC
RPI - 1, 2, 3, 4
KenPom - 1, 2, 4, 3

When looking at groupings, the metric used rarely matters.

[Cdizzle]

The point still stands. If ALL you care about is performance against a subset of a given metric, just use the friggin metric.