RPI is no longer to be used for NCAAT selection

[ESPNjacket]

Excerpt: The new metric, which is called the NCAA Evaluation Tool (or NET), will rely on “game results, strength of schedule, game location, scoring margin, net offensive and defensive efficiency, and the quality of wins and losses.” (It’s worth noting here that while scoring margin is factored in, it’s impact will be capped at 10 points to prevent running up the score.)

Link:
https://collegebasketball.nbcsports...etric-eliminates-rpi-as-primary-sorting-tool/

 

[Deleted member 2897]

Excerpt: The new metric, which is called the NCAA Evaluation Tool (or NET), will rely on “game results, strength of schedule, game location, scoring margin, net offensive and defensive efficiency, and the quality of wins and losses.” (It’s worth noting here that while scoring margin is factored in, it’s impact will be capped at 10 points to prevent running up the score.)

Link:
https://collegebasketball.nbcsports...etric-eliminates-rpi-as-primary-sorting-tool/

They had pretty much stopped using the RPI in previous years anyway.

Here are some teams who didn't make the NCAA Tournament last year and their RPI. (As a reminder, there are 64 teams, and due to conference championships there are typically only about 50 invites out there for 'best available'.)
Middle Tennessee State: 33
Southern California: 34
UCLA: 36 (got to a play-in game, but lost)
Louisville: 38
Western Kentucky: 39
St. Mary's: 40
Texas: 51

 

[slugboy]

Not RPI, but a formula based on the same basic building blocks as ROI.
OK, I'm sure there won't be any issues with this.

 

[Deleted member 2897]

Not RPI, but a formula based on the same basic building blocks as ROI.
OK, I'm sure there won't be any issues with this.

Exactly. RPI, Ken Pom, S&P, FEI. They need to just fess up and say they use a combination of these metrics to get to a draft list, and then they use subjective discussions amongst the committee to whittle it down. But they probably don't want to admit they are subjective.

 

[AUFC]

I know there have been gripes with RPI in recent years but with my Georgia Tech fan face on, is this good for us or bad for us? Are we usually underrated by RPI, overrated by RPI, or neither?

 

[CuseJacket]

If there is one concerning element about this new metric, it’s that the NCAA is as of yet undecided on whether or not to peel back the current and show us how the sausage is made, so to speak. Gavitt told CBS Sports that the algorithm will be powered by artificial intelligence and that it will not be “readable”, and that should be mildly concerning.

Mildly?

Even if they get this right and share the formula, the fact that the NCAA is even considering putting a curtain over the formula demonstrates that there has been no forward progress in their own ineptitude.

 

[ESPNjacket]

I know there have been gripes with RPI in recent years but with my Georgia Tech fan face on, is this good for us or bad for us? Are we usually underrated by RPI, overrated by RPI, or neither?

RPI is simply a bad metric. They are replacing it with a better one. GT doesn't have a side in it other than a school like GT should always support better math.

 

[Peacone36]

This metric is far better from what I have been able to read about it. I like the win differential being a factor. Not sure the cap shouldnt be 15 instead of 10 though. Also, I hope this doesnt lead to something such as the BCS. I love teh reveal show of the NCAAT.

 

[jeffgt14]

Seems similar to the RPI to me. Maybe a little better though.

 

[orientalnc]

Years ago I tried building a metric based ranking system. I had all the mainframe power I needed in our Bell Labs center. My bosses were even OK'd the project because I was using the new UNIX based DB called Ingres. But, no matter how much we tried, we could not remove the subjective aspects of the algorithm completely. Every ranking system has to have a starting point and it takes a while for those decisions to weed themselves out of the model. Inevitably, you have to accept that someone(s) will be looking at team A and saying they are better than team B.

This is very similar to a football ref following a play where the ball carrier is tackled 25 yards away. He races across the field and puts his toe where he thinks the runner first touched the ground. Then the chains are brought in to determine and the spot can be a half inch short or beyond the stick. That is a very precise metric for a very imprecise event.

If we beat Duke by 8 points at home is that better than UNC beating Duke by 1 in Durham? I am sure the model will say it knows, but is that really correct?

 

[CuseJacket]

RPI is simply a bad metric. They are replacing it with a better one. GT doesn't have a side in it other than a school like GT should always support better math.

This metric is far better from what I have been able to read about it. I like the win differential being a factor. Not sure the cap shouldnt be 15 instead of 10 though. Also, I hope this doesnt lead to something such as the BCS. I love teh reveal show of the NCAAT.

How do we know it's better math, other than they say it is?

I understand RPI was flawed, but if this formula is kept behind a curtain, they can call it "world peace" and it doesn't mean anything has improved.

No one has a clue how anything is weighted, as far as I've seen.

 

[CuseJacket]

Common opponent
Team A: Maintains a 10 point lead throughout the game. Up by 9 with 1 minute to go. Opponent hits a 3 at the buzzer. Team A wins by 6.
Team B: 30 lead changes. Up by 1 with 25 seconds to go. Foul, foul, foul. foul. Team B wins by 9.

Conclusion: Team B is better based on margin of victory......... sorry, the last 2 minutes of basketball looks nothing like football.

Give me the formula before it gets anointed as better. I want to believe it is, but I don't have blind faith in the uNCAA to meddle in something and get it right.

 

[ESPNjacket]

How do we know it's better math, other than they say it is?

I understand RPI was flawed, but if this formula is kept behind a curtain, they can call it "world peace" and it doesn't mean anything has improved.

No one has a clue how anything is weighted, as far as I've seen.

It uses offensive and defensive efficiency, used input from Ken Pomeroy and the BPI guys at ESPN, and was tested during last season. Those things make it superior to RPI, which is a terrible model based upon nothing but guesswork.

 

[CuseJacket]

It uses offensive and defensive efficiency, used input from Ken Pomeroy and the BPI guys at ESPN, and was tested during last season. Those things make it superior to RPI, which is a terrible model based upon nothing but guesswork.

I'm not trying to be obtuse. I acknowledge all of those things sound good. But no one knows how much efficiency matters. No one knows how much margin of victory matters. At least the RPI formula was transparent and coaches knew how the things they did affected their resume and tournament chances.

Since the NCAA has not shared the new formula, have they shared results of what it looked like when modeling last season? Would be interesting to see how teams stacked up as a point of reference.

 

[orientalnc]

I'm not trying to be obtuse. I acknowledge all of those things sound good. But no one knows how much efficiency matters. No one knows how much margin of victory matters. At least the RPI formula was transparent and coaches knew how the things they did affected their resume and tournament chances.

Since the NCAA has not shared the new formula, have they shared results of what it looked like when modeling last season? Would be interesting to see how teams stacked up as a point of reference.

The problem with modeling last season as a test case is that EVERY ranking system has to begin with a set of assumptions about the teams rank order before games are played. I would advance that it is impossible to do that for 2017-18 without introducing the bias known as the 2018 March Madness results (or the RPI at season's end). You could, I guess, accept the assumptions someone else used for the RPI to rank teams prior to the first 2017 games. Either way you are going to make a choice between UCLA and Ga Tech and one of those teams will be rewarded and the other penalized based on how they performed vs. the model's expectation.

 

[ESPNjacket]

I'm not trying to be obtuse. I acknowledge all of those things sound good. But no one knows how much efficiency matters. No one knows how much margin of victory matters. At least the RPI formula was transparent and coaches knew how the things they did affected their resume and tournament chances.

Since the NCAA has not shared the new formula, have they shared results of what it looked like when modeling last season? Would be interesting to see how teams stacked up as a point of reference.

I don't think trying to work the system is a good thing. We have a fundamental disagreement.

 

[ESPNjacket]

This link has a more complete explanation than the previous one:

http://www.sportingnews.com/ca/ncaa...enpom-march-madness/1wk0ufr4jnljvzwm8dnrfq266

 

[CuseJacket]

I don't think trying to work the system is a good thing. We have a fundamental disagreement.

I don't think that's the disagreement. I agree coaches should not "work the system". I believe we'd both be comfortable defining "working the system" as skirting the spirit of the game, manipulating the calculation by taking shortcuts (e.g., scheduling easy road W's at Kennesaw or Wofford to increase RPI ) or frankly doing anything other than just making their own team better.

I think the disagreement is around whether performance measures (this calculation) and incentives (NCAAT berths) are aligned appropriately. If those are aligned in the spirit of identifying the best team, then "working the system" ultimately boils down to coaching their teams better. So if this new calculation is so great, why not just share it?

Oversimplistic, I know, but it is crazy to not know how your team is being measured. For all we know they're counting margin of victory as 95% of the weight. Or 1%. So should a team sit on the ball up 5 with 45 sec to go or start chucking 3's? That other team on the bubble just beat that opponent by 7...

 

[ESPNjacket]

I don't think that's the disagreement. I agree coaches should not "work the system". I believe we'd both be comfortable defining "working the system" as skirting the spirit of the game, manipulating the calculation by taking shortcuts (e.g., scheduling easy road W's at Kennesaw or Wofford to increase RPI ) or frankly doing anything other than just making their own team better.

I think the disagreement is around whether performance measures (this calculation) and incentives (NCAAT berths) are aligned appropriately. If those are aligned in the spirit of identifying the best team, then "working the system" ultimately boils down to coaching their teams better. So if this new calculation is so great, why not just share it?

Oversimplistic, I know, but it is crazy to not know how your team is being measured. For all we know they're counting margin of victory as 95% of the weight. Or 1%. So should a team sit on the ball up 5 with 45 sec to go or start chucking 3's? That other team on the bubble just beat that opponent by 7...

I'm sorry. You are just being dramatic. They are replacing RPI with a better metric. So the quadrant information on the sheet will better represent what they are meant to do (the top X are more accurately represented so the corresponding quadrant information is inherently more accurate). This will not lead to odd decision making at the end of games. It is a false premise.

 

[CuseJacket]

I'm sorry. You are just being dramatic. They are replacing RPI with a better metric. So the quadrant information on the sheet will better represent what they are meant to do (the top X are more accurately represented so the corresponding quadrant information is inherently more accurate). This will not lead to odd decision making at the end of games. It is a false premise.

Yea, I almost deleted that example because I thought it might be cherrypicked to refute the broader point. I don't think asking for the formula is dramatic. In fact, I think failing to disclose the formula is inherently a backwards step. It is difficult to convince me that the NCAA got this right, because any argument about "better" is based on faith the NCAA got something right. Maybe that is the other fundamental disagreement.