Year 1 of the NET

#51
#51
Bruin, you have to have the index to determine quadrant wins. Therefore, the NET is not meaningless. You have to somehow rank the teams in order to determine what quadrant the wins are in.

You can't use polls to determine quadrant wins because they are biased by humans, and teams fall out of the top 25 during the year. You have to have a metric first to establish quadrants.


Yea that's true. From what we have seen here the RPI is much better to use for that ranking. No way in hell a Clemson at 1-10 should be in the top 40. Their RPI is much more accurate imo
 
#52
#52
I can't speak to how important. I just know that they wanted to replace it with the RPI, which was also flawed. They are never going to just go right down the rankings to include teams. There is always more to it than that.

I just read that, unlike KenPom, the NET does not adjust its efficiency ratings for the caliber of the team. Therefore, as I stated above, that's a reason why NC State got a boost. That's a problem, but this committee went deeper to evaluate NC State and realized that they didn't beat anybody even with their 30+ NET ranking.

This year was as about as fair as it is going to get. I have criticized past committees, but there this year had a light bubble. There aren't teams that were left out that truly deserved to be in.
It’s clearly not important. Which is fine. My biggest problem is they are never consistent in their logic. Even within the same year. Seems odd that they can’t be.
 
#53
#53
Yea that's true. From what we have seen here the RPI is much better to use for that ranking. No way in hell a Clemson at 1-10 should be in the top 40. Their RPI is much more accurate imo

Well, the 1-10 record doesn't reflect on their NET ranking. Just like NC State, it appears Clemson's NET was boosted by beating bad teams by 10+ points and having a better net efficiency. They probably played a tougher schedule than NC State.
 
#54
#54
Well, the 1-10 record doesn't reflect on their NET ranking. Just like NC State, it appears Clemson's NET was boosted by beating bad teams by 10+ points and having a better net efficiency. They probably played a tougher schedule than NC State.

So if that's the key to these higher rankings than are appropriate are there any plans to tinker with that factor?
 
#55
#55
So if that's the key to these higher rankings than are appropriate are there any plans to tinker with that factor?

No idea. They should use the net efficiencies like KenPom and adjust it based on the caliber of team you play.

I am sure UT got a boost for winning 20+ games by double digit. Our OOC SOS ended up being very mediocre because of GTech, WF, and WV.
 
#56
#56
One additional point on SOS - this is overly simplistic, but if Team A plays 1, 2 and 300, it has an average SOS of 101. If Team B plays 99, 100 and 101, it has an average SOS of 100. Of course, clearly Team A has played the harder schedule. RPI doesn't really distinguish, but the more advance metrics can.
 
#57
#57
One additional point on SOS - this is overly simplistic, but if Team A plays 1, 2 and 300, it has an average SOS of 101. If Team B plays 99, 100 and 101, it has an average SOS of 100. Of course, clearly Team A has played the harder schedule. RPI doesn't really distinguish, but the more advance metrics can.


How does SOS figure in location of the game?
 
#58
#58
One additional point on SOS - this is overly simplistic, but if Team A plays 1, 2 and 300, it has an average SOS of 101. If Team B plays 99, 100 and 101, it has an average SOS of 100. Of course, clearly Team A has played the harder schedule. RPI doesn't really distinguish, but the more advance metrics can.

Well, in the RPI itself, it actually uses your opponents' records. So, a team that is ranked 300 most likely has very few wins, which effects your metric.

That's why P5 teams shouldn't play teams that are 250+.
 
#60
#60
It doesn't. The RPI and the NET do in their calculations. You are awarded more points for a road win versus a home win.


Ok thanks.


Obviously all these rankings have flaws and I am sure that can't be remedied. So we need to use the one that is the lesser evil.


For me RPI is much more accurate than What we have seen from the NET this year
 
#61
#61
Well, in the RPI itself, it actually uses your opponents' records. So, a team that is ranked 300 most likely has very few wins, which effects your metric.

That's why P5 teams shouldn't play teams that are 250+.


Fair enough, but that is sort of my point. RPI puts too much focus on SOS and not enough focus on winning and other factors. In the RPI, a loss against great team is more valuable than a win against a bad team, regardless of margin of victory and other factors. So, NC State plays a bad team, but beats them by 60. That is worse than if they had played a good team but lost.
 
#62
#62
Fair enough, but that is sort of my point. RPI puts too much focus on SOS and not enough focus on winning and other factors. In the RPI, a loss against great team is more valuable than a win against a bad team, regardless of margin of victory and other factors. So, NC State plays a bad team, but beats them by 60. That is worse than if they had played a good team but lost.


And for that reason we see Kansas at 2 RPI which is obviously a poor ranking
 
#63
#63
I am sure they all have flaws, but all in all the RPI is much worse than the advanced metrics. Going back to NC State - let's say they beat some team that had 2 wins. Is that really that much different than beating a team with say 10 wins? Both are bad teams and NC State should (and did) beat both. If you think there should be a big difference, then you probably like RPI. If you think they shouldn't get credit for beating a 2 win team, but they also shouldn't be penalized, then you will like NET (or KenPom, etc.).
 
#65
#65
I am sure they all have flaws, but all in all the RPI is much worse than the advanced metrics. Going back to NC State - let's say they beat some team that had 2 wins. Is that really that much different than beating a team with say 10 wins? Both are bad teams and NC State should (and did) beat both. If you think there should be a big difference, then you probably like RPI. If you think they shouldn't get credit for beating a 2 win team, but they also shouldn't be penalized, then you will like NET (or KenPom, etc.).


Hell I am not sure how they end up with the rankings but I think it's clear that NC state Clemson are much more properly ranked by the RPI than they are by the NET
 
#66
#66
Many of NC States opponents ended up being a lot worse than they were expected to be. It’s not like the purposefully sought out the easiest OOC schedule, but they were penalized for it. The NET controls for that somewhat by factoring in how they played in those games.
 
#67
#67
Hell I am not sure how they end up with the rankings but I think it's clear that NC state Clemson are much more properly ranked by the RPI than they are by the NET

Why is that clear? Both KenPom and Sagarin have them similarly ranked.
 
#68
#68
In the RPI, any team that plays a difficult SOS will almost by default have a strong RPI, regardless of its record. Obviously, that doesn't make sense.


I would have agreed with that until looking at Florida.

Top 25 schedule

30 NET but has an RPI of 58

Imo the 58 is closer to what they Deserve with a 4-12 Q1 record
 
#70
#70
Many of NC States opponents ended up being a lot worse than they were expected to be. It’s not like the purposefully sought out the easiest OOC schedule, but they were penalized for it. The NET controls for that somewhat by factoring in how they played in those games.

This is true. (Playing Vandy is a good example of this)
 
#71
#71
I would have agreed with that until looking at Florida.

Top 25 schedule

30 NET but has an RPI of 58

Imo the 58 is closer to what they Deserve

"In your opinion . . ." I think you just answered your own question. You already have an opinion and it just happens to match the RPI, so you prefer that. Others have different opinions. The computer ratings are designed to provide an objective rating, not simply confirm what you already think.
 
#72
#72
"In your opinion . . ." I think you just answered your own question. You already have an opinion and it just happens to match the RPI, so you prefer that. Others have different opinions. The computer ratings are designed to provide an objective rating, not simply confirm what you already think.


I had zero opinion before looking into this close today.

For me florida being 4-12 in no way matches the 30 ranking.
Same for Clemson and NC state. I can't find a logical answer for why they are ranked that high
 
#73
#73
The biggest thing I have learned today is the NET uses efficiency stats that can be inflated by playing really poor teams without penalty
 
#74
#74
The biggest thing I have learned today is the NET uses efficiency stats that can be inflated by playing really poor teams without penalty
It’s only a component. There’s another component which does take opponent into consideration. It’s basically like RPI with efficiency numbers layered on top. Ken Pom is entirely efficiency based and factors in strength of opponents and he has Nc state ranked 33
 
#75
#75
It’s only a component. There’s another component which does take opponent into consideration. It’s basically like RPI with efficiency numbers layered on top. Ken Pom is entirely efficiency based and factors in strength of opponents and he has Nc state ranked 33


And ken pom doesn't use w/L?
 

VN Store



Back
Top