Football Home Game
Sat, August 31, 2024
Sat, August 31, 2024
Basketball Home Game
Fri, November 1, 2024
Fri, November 1, 2024
NET ranking unchanged
NET ranking unchanged
The Aggies are still at 30, while the Aztecs are still at 18. New Mexico moved up one spot without playing.
I have come to understand that winning doesn’t really move the needle, while USU is likely to move up to about 26 over the next week while not playing.
https://bracketologists.com/
I have come to understand that winning doesn’t really move the needle, while USU is likely to move up to about 26 over the next week while not playing.
https://bracketologists.com/
Last edited by dogie on February 21st, 2024, 7:04 am, edited 3 times in total.
-
- Posts: 2772
- Joined: November 17th, 2010, 6:59 pm
- Location: Pullman, WA
- Has thanked: 320 times
- Been thanked: 657 times
- Contact:
Re: NET ranking unchanged
I am beginning too think there is some sort of decaying exponential factor included in some weights as the season goes on. There can be no other explanation as to why there was zero movement for us and SDSU
My side projects:
Internet-connected Aggie A's: www.sports-iot.com
Physics and the Pinewood Derby: www.pinewoodphysics.com
Internet-connected Aggie A's: www.sports-iot.com
Physics and the Pinewood Derby: www.pinewoodphysics.com
-
- Aggie Insider, Pick'em Champ - '18 Kickoff, '19 Weekly
- Posts: 19759
- Joined: November 3rd, 2010, 8:17 pm
- Location: Smithfield, Utah
- Has thanked: 23838 times
- Been thanked: 16464 times
- Contact:
Re: NET ranking unchanged
In the eyes of the NET the game was a draw?
- These users thanked the author aggies22 for the post (total 2):
- flying_scotsman2.0 • SaintAggie
-
- Posts: 3090
- Joined: July 4th, 2013, 12:04 pm
- Has thanked: 1836 times
- Been thanked: 2504 times
Re: NET ranking unchanged
I am not that suprised. We were favored by single digits and won by single digits. The outcome of the game was almost exactly what net predicted.
- AggieFBObsession
- Posts: 3200
- Joined: January 25th, 2011, 12:15 pm
- Has thanked: 6806 times
- Been thanked: 1237 times
-
- Posts: 14412
- Joined: December 15th, 2010, 6:29 pm
- Has thanked: 4524 times
- Been thanked: 4169 times
Re: NET ranking unchanged
Boise jumped 7 spots after their big home win vs sjsu.
But the key there is they are almost to 30 now which would give us another q1
But the key there is they are almost to 30 now which would give us another q1
Re: NET ranking unchanged
That’s right. Boise State’s Q4 victory over SJSU moved them from 42 to 35.
- These users thanked the author dogie for the post (total 2):
- flying_scotsman2.0 • treesap32
- Roy McAvoy
- Posts: 7617
- Joined: November 2nd, 2011, 1:30 pm
- Has thanked: 1238 times
- Been thanked: 3061 times
Re: NET ranking unchanged
I do think the NET seems to do a pretty good job overall, but I have to think it’s wrong on that had we beat Fresno by 30 tonight, I bet we would’ve moved up a spot or two, but we beat SDSU by 5 and had no change.
If you look at Kenpom, we’re #40. But that’s because it’s not meant to be a rank, it’s meant to be a predictor. It gives no weight to actually winning. Since we’ve had so many close wins we’re only 40. The thing is this team (maybe mostly Darius Brown) is better when the game is close and on the line.
That aspect isn’t factored into Kenpom.
If you look at Kenpom, we’re #40. But that’s because it’s not meant to be a rank, it’s meant to be a predictor. It gives no weight to actually winning. Since we’ve had so many close wins we’re only 40. The thing is this team (maybe mostly Darius Brown) is better when the game is close and on the line.
That aspect isn’t factored into Kenpom.
- hipsterdoofus21
- Mr. Buttface
- Posts: 18342
- Joined: November 3rd, 2010, 9:39 pm
- Has thanked: 3452 times
- Been thanked: 3371 times
Re: NET ranking unchanged
I’m not going to worry about NET. It’ll likely improve for us through the week. I’m focused on the regular season title. I’ve liked the NET system but this year it’s become crystal clear that it can be easily games by RUTS against Q3-4s. And while we were picked to win by 2-3 pts and won by 5 pts, we still deserved to move up.
- These users thanked the author hipsterdoofus21 for the post (total 2):
- Bullnamed_gus • flying_scotsman2.0
-
- Posts: 2068
- Joined: October 31st, 2022, 12:25 pm
- Has thanked: 280 times
- Been thanked: 1290 times
Re: NET ranking unchanged
Bad day for the Stanford Aggie Agenda.
Oh well, NET is just a sorting tool. As long as we stay top 35/40 and no bad losses we will be fine.
Our seed floor is an 8, with a NM win and a few wins in the tournament, we could probably get to a 5.
-
- Posts: 385
- Joined: November 6th, 2010, 2:35 pm
- Location: Tuscaloosa, AL
- Has thanked: 94 times
- Been thanked: 130 times
Re: NET ranking unchanged
I expected our NET ranking to go up after the win, as well. However, one thing we don't know is how our overall numerical score that rankings are based on may have changed, and what it is relative to the other teams around us. It's possible that our score did increase, but that it was just barely above the #31 ranked team yesterday and is now just below the #29 team today.
- flying_scotsman2.0
- Posts: 3597
- Joined: January 23rd, 2018, 12:29 pm
- Location: The Mighty City-State of Roy, Utah
- Has thanked: 6064 times
- Been thanked: 2297 times
Re: NET ranking unchanged
You guys might stress Stanford out enough to kill him. You just want to poke him to see how much stats you can get out of him.Bullnamed_gus wrote: ↑February 21st, 2024, 7:22 amBad day for the Stanford Aggie Agenda.
Oh well, NET is just a sorting tool. As long as we stay top 35/40 and no bad losses we will be fine.
Our seed floor is an 8, with a NM win and a few wins in the tournament, we could probably get to a 5.
- These users thanked the author flying_scotsman2.0 for the post:
- Bullnamed_gus
-
- Posts: 2068
- Joined: October 31st, 2022, 12:25 pm
- Has thanked: 280 times
- Been thanked: 1290 times
Re: NET ranking unchanged
I respect SA a lot. Hes obviously very smart. I just enjoy poking the bearflying_scotsman2.0 wrote: ↑February 21st, 2024, 8:17 amYou guys might stress Stanford out enough to kill him. You just want to poke him to see how much stats you can get out of him.Bullnamed_gus wrote: ↑February 21st, 2024, 7:22 amBad day for the Stanford Aggie Agenda.
Oh well, NET is just a sorting tool. As long as we stay top 35/40 and no bad losses we will be fine.
Our seed floor is an 8, with a NM win and a few wins in the tournament, we could probably get to a 5.
- These users thanked the author Bullnamed_gus for the post (total 2):
- flying_scotsman2.0 • StanfordAggie
-
- SJSU Ultimate Loser Award Winner - Given to someone that should probably give up but won't.
- Posts: 23541
- Joined: November 3rd, 2010, 8:09 am
- Location: Where the sagebrush grows!
- Has thanked: 1428 times
- Been thanked: 3282 times
Re: NET ranking unchanged
We have to admit though, he was right about the B12 being a good conference and BYU not being the worst in the conference.
- These users thanked the author NVAggie for the post:
- sam tingey
- Roy McAvoy
- Posts: 7617
- Joined: November 2nd, 2011, 1:30 pm
- Has thanked: 1238 times
- Been thanked: 3061 times
Re: NET ranking unchanged
How so? The way I see it is last night's results supported his argument.Bullnamed_gus wrote: ↑February 21st, 2024, 7:22 amBad day for the Stanford Aggie Agenda.
Oh well, NET is just a sorting tool. As long as we stay top 35/40 and no bad losses we will be fine.
Our seed floor is an 8, with a NM win and a few wins in the tournament, we could probably get to a 5.
- These users thanked the author Roy McAvoy for the post (total 2):
- AGinNEIowa • QuackAttackAggie
-
- Posts: 7923
- Joined: November 3rd, 2010, 12:07 pm
- Has thanked: 413 times
- Been thanked: 5002 times
Re: NET ranking unchanged
The result of the game last night is what should have happened when the NET 18 team plays the NET 30 team on the NET 30 team’s home floor. I don’t take issue with the lack of movement.
What is a little silly is a team moving up a ton for blasting a bad team.
What is a little silly is a team moving up a ton for blasting a bad team.
- These users thanked the author ineptimusprime for the post (total 4):
- AGinNEIowa • Roy McAvoy • treesap32 • SweepDance
- 3rdGenAggie
- Pick'em Champ - '16 Kickoff
- Posts: 12475
- Joined: November 3rd, 2010, 11:53 pm
- Location: The City of the Salty Lake
- Has thanked: 4131 times
- Been thanked: 2414 times
Re: NET ranking unchanged
I'm betting next year a lot more teams will try the Big XII approach of scheduling an appalling OOC schedule in order to destroy bad and terrible teams to artificially inflate their NET.
It'll be interesting to see what happens overall as more teams do that.
It'll be interesting to see what happens overall as more teams do that.
"I have no idea what I'm doing, but I know I'm doing it really, really well." -Andy Dwyer
- QuackAttackAggie
- Pick'em Champ - '12, '22 Bowl; '15, '17 Weekly; '18 BB Predict the Score; '22 Kickoff
- Posts: 18313
- Joined: November 3rd, 2010, 12:08 pm
- Location: Sonora, MX
- Has thanked: 323 times
- Been thanked: 3060 times
Re: NET ranking unchanged
We actually would have moved to 29, but Wake jumped us (and like 18 other teams) after blowing out Pitt
-
- Posts: 2343
- Joined: January 16th, 2011, 8:11 am
- Has thanked: 7 times
- Been thanked: 432 times
Re: NET ranking unchanged
The 3 pointer at the end probably cost us a few NET points as well.
The debate with SA is dumb. The argument is apples and oranges. His argument is that playing better opponents on road or neutral sites is best way to improve your net standings which is probably true. But that is all done in scheduling prior to the season. Can't do anything about it now.
The rest argue that RUTSing bumps up your NET rating, also true. That's something teams can do now. Just look at Wake Forest win. If that is not proof then I don't know what is. Just run up the score, no matter who you are playing and you will have a better NET ranking.
The debate with SA is dumb. The argument is apples and oranges. His argument is that playing better opponents on road or neutral sites is best way to improve your net standings which is probably true. But that is all done in scheduling prior to the season. Can't do anything about it now.
The rest argue that RUTSing bumps up your NET rating, also true. That's something teams can do now. Just look at Wake Forest win. If that is not proof then I don't know what is. Just run up the score, no matter who you are playing and you will have a better NET ranking.
-
- Posts: 1337
- Joined: November 8th, 2010, 7:57 pm
- Has thanked: 239 times
- Been thanked: 427 times
Re: NET ranking unchanged
Just noticed Wake Forest jumped from 40 to 26
They beat Pitt a Q2 team at home by 34
That’s quite the jump in one day
They beat Pitt a Q2 team at home by 34
That’s quite the jump in one day
-
- Posts: 2480
- Joined: November 15th, 2010, 6:36 pm
- Has thanked: 637 times
- Been thanked: 637 times
Re: NET ranking unchanged
No, I'm pretty sure it's just a simple average. (Specifically, it's the average number of points/allowed per possession after adjusting for the strength of your opponents.) And averages don't move as much once you have a larger number of data points. Hence you see a lot more big jumps in NET early in the season that you do later in the season.
-
- Posts: 2068
- Joined: October 31st, 2022, 12:25 pm
- Has thanked: 280 times
- Been thanked: 1290 times
Re: NET ranking unchanged
His argument is that beating better teams helps your net more than beating worse teams by a bunch. His theory would mean a win at home over SDSU, even if closer, helps more than a blowout win agaisnt a bad team. Boise jumped what 7 spots, Utah State didnt move.Roy McAvoy wrote: ↑February 21st, 2024, 8:46 amHow so? The way I see it is last night's results supported his argument.Bullnamed_gus wrote: ↑February 21st, 2024, 7:22 amBad day for the Stanford Aggie Agenda.
Oh well, NET is just a sorting tool. As long as we stay top 35/40 and no bad losses we will be fine.
Our seed floor is an 8, with a NM win and a few wins in the tournament, we could probably get to a 5.
-
- Posts: 1127
- Joined: November 13th, 2010, 7:34 pm
- Has thanked: 634 times
- Been thanked: 380 times
Re: NET ranking unchanged
The NET uses a consistent formula with multiple factors. I find it funny every time someone mentions a single factor and then says the impact to NET doesn’t make sense based on that single factor.
How did all of Boise’s prior opponents do?
How close was Boise in the calculated NET to those teams they jumped?
How did those teams that were jumped do? How did the prior opponents of those who were jumped do?
How did all of Boise’s prior opponents do?
How close was Boise in the calculated NET to those teams they jumped?
How did those teams that were jumped do? How did the prior opponents of those who were jumped do?
Last edited by Pacobag on February 21st, 2024, 10:15 pm, edited 2 times in total.
-
- Posts: 656
- Joined: October 26th, 2019, 10:00 pm
- Has thanked: 417 times
- Been thanked: 429 times
Re: NET ranking unchanged
How so? Winning does nothing for your net? That doesn’t make sense to me at all. I can see the argument that SDSU shouldn’t fall much given how close the game was. But USU not gaining a few points for beating a great SDSU team is a complete head scratcher to me. What win is more impressive, Wake over Pitt or USU over SDSU. The NET says the Wake win was better……definitely some flaws in this algorithm.ineptimusprime wrote: ↑February 21st, 2024, 8:48 amThe result of the game last night is what should have happened when the NET 18 team plays the NET 30 team on the NET 30 team’s home floor. I don’t take issue with the lack of movement.
What is a little silly is a team moving up a ton for blasting a bad team.
-
- Posts: 656
- Joined: October 26th, 2019, 10:00 pm
- Has thanked: 417 times
- Been thanked: 429 times
Re: NET ranking unchanged
I get what you are saying but not all of Boises opponents played that same night just like not all of USUs opponents played last night. It’s obvious that some factors are weighted much heavier than others. Otherwise teams like Wake wouldn’t jump that far in a single night.Pacobag wrote: ↑February 21st, 2024, 10:12 pmThe NET uses a consistent formula with multiple factors. I find it funny every time some quotes a single factor and then says the impact to NET doesn’t make sense based on that single factor.
How did all of Boise’s opponents do?
How close was Boise in the calculated NET to those teams they jumped?
How did d those teams that were jumped do? How did the opponents of those who were jumped do?
-
- Posts: 1127
- Joined: November 13th, 2010, 7:34 pm
- Has thanked: 634 times
- Been thanked: 380 times
Re: NET ranking unchanged
Hypothetically (and I’m completely making up fictitious numbers), let’s say the 30th NET team had a NET calculation that was 0.030 behind the 29th NET team. An improvement of 0.029 could still mean 30th in the NET.Slim80 wrote: ↑February 21st, 2024, 10:17 pmI get what you are saying but not all of Boises opponents played that same night just like not all of USUs opponents played last night. It’s obvious that some factors are weighted much heavier than others. Otherwise teams like Wake wouldn’t jump that far in a single night.Pacobag wrote: ↑February 21st, 2024, 10:12 pmThe NET uses a consistent formula with multiple factors. I find it funny every time some quotes a single factor and then says the impact to NET doesn’t make sense based on that single factor.
How did all of Boise’s opponents do?
How close was Boise in the calculated NET to those teams they jumped?
How did d those teams that were jumped do? How did the opponents of those who were jumped do?
Let’s say Boise State and the 7 teams ahead of them in the NET were only separated by 0.010 in the NET calculation. If BSU improved by 0.011, they would jump 7 spots while the larger 0.029 change wouldn’t result in any movement for the 30th NET team.
Last edited by Pacobag on February 22nd, 2024, 12:03 am, edited 2 times in total.
-
- Posts: 656
- Joined: October 26th, 2019, 10:00 pm
- Has thanked: 417 times
- Been thanked: 429 times
Re: NET ranking unchanged
I understand what you are saying Paco, I really do. But at some point you have to remove yourself from the inner workings of the algorithms and look at things holistically. It’s flat out asinine that a team like Wake jumped us in the net in a single night of play given the games that were played.
-
- Posts: 1127
- Joined: November 13th, 2010, 7:34 pm
- Has thanked: 634 times
- Been thanked: 380 times
Re: NET ranking unchanged
Fair callout, I too am surprised that they were 40th and jumped us. Wish I had the exact formula and before and after numerical factors that caused the change.Slim80 wrote: ↑February 21st, 2024, 10:40 pmI understand what you are saying Paco, I really do. But at some point you have to remove yourself from the inner workings of the algorithms and look at things holistically. It’s flat out asinine that a team like Wake jumped us in the net in a single night of play given the games that were played.
- Aglicious
- Site Admin
- Posts: 7209
- Joined: January 14th, 2004, 12:00 am
- Location: Vega$
- Has thanked: 978 times
- Been thanked: 2544 times
Re: NET ranking unchanged
Well, Pitt was NET 47 and Wake was NET 40 going into that game. Wake was favored by 6pts at home but ended up winning by 33! Beating the expectations by more than 5x vaulted WF.Slim80 wrote: ↑February 21st, 2024, 10:13 pmHow so? Winning does nothing for your net? That doesn’t make sense to me at all. I can see the argument that SDSU shouldn’t fall much given how close the game was. But USU not gaining a few points for beating a great SDSU team is a complete head scratcher to me. What win is more impressive, Wake over Pitt or USU over SDSU. The NET says the Wake win was better……definitely some flaws in this algorithm.ineptimusprime wrote: ↑February 21st, 2024, 8:48 amThe result of the game last night is what should have happened when the NET 18 team plays the NET 30 team on the NET 30 team’s home floor. I don’t take issue with the lack of movement.
What is a little silly is a team moving up a ton for blasting a bad team.
USU was favored by 2.5pts at home and won by 5.
Beating the expectations by large margins (RUTS) and thus affecting efficiency numbers is what moves the NET needle. While our win was great and gave us another Q1 win (maybe more valuable in the eyes of committee members), the results were about what was expected and therefore didn't affect the ratings of either team involved. Had we won by 12-15 points we may have moved a spot or two.
- USU78
- Pick'em Champ - '16 Weekly
- Posts: 15486
- Joined: November 3rd, 2010, 6:43 am
- Location: Sandy
- Has thanked: 7228 times
- Been thanked: 2111 times
Re: NET ranking unchanged
GIGO.Pacobag wrote: ↑February 21st, 2024, 10:47 pmFair callout, I too am surprised that they were 40th and jumped us. Wish I had the exact formula and before and after numerical factors that caused the change.Slim80 wrote: ↑February 21st, 2024, 10:40 pmI understand what you are saying Paco, I really do. But at some point you have to remove yourself from the inner workings of the algorithms and look at things holistically. It’s flat out asinine that a team like Wake jumped us in the net in a single night of play given the games that were played.
Or ...
Desire determines outcomes, not principles, mathematical or otherwise.
You keep using that word. I do not think it means what you think it means.
-
- Posts: 1127
- Joined: November 13th, 2010, 7:34 pm
- Has thanked: 634 times
- Been thanked: 380 times
Re: NET ranking unchanged
What determines the outcome of a 20 trillion ton asteroid crashing into the earth at 50,000 mph?USU78 wrote: ↑February 21st, 2024, 11:06 pmGIGO.Pacobag wrote: ↑February 21st, 2024, 10:47 pmFair callout, I too am surprised that they were 40th and jumped us. Wish I had the exact formula and before and after numerical factors that caused the change.Slim80 wrote: ↑February 21st, 2024, 10:40 pmI understand what you are saying Paco, I really do. But at some point you have to remove yourself from the inner workings of the algorithms and look at things holistically. It’s flat out asinine that a team like Wake jumped us in the net in a single night of play given the games that were played.
Or ...
Desire determines outcomes, not principles, mathematical or otherwise.
Who had more desire, Rudy Ruettiger or the guys that started ahead of him. Or maybe closer to home, Sam Merrill or LeBron James?
-
- Posts: 2480
- Joined: November 15th, 2010, 6:36 pm
- Has thanked: 637 times
- Been thanked: 637 times
Re: NET ranking unchanged
No, that's not my position. I agree that one can increase one's NET by RUTS against a bad team. My claim is that beating a bad team by enough points to increase one's NET is not easy to do, and it's just as likely to hurt you as help you. Remember when USU's NET dropped because we beat Fresno at home by only 20? I don't really want to rehash this debate, but I can point to numerous examples of teams who hurt their NET (not to mention their at-large chances) by having an off night against a bad team. Nebraska dropped 11 spots in the NET when they beat #235 North Dakota by only 8. Or Mississippi State managed to lose to #203 Southern at home. (It was too early in the season to see what it did to their NET.) Or some team named Utah State dropped 24 spots in the NET in 2022 when they lost at home to #187 Weber StateBullnamed_gus wrote: ↑February 21st, 2024, 8:17 pmHis argument is that beating better teams helps your net more than beating worse teams by a bunch. His theory would mean a win at home over SDSU, even if closer, helps more than a blowout win agaisnt a bad team. Boise jumped what 7 spots, Utah State didnt move.Roy McAvoy wrote: ↑February 21st, 2024, 8:46 amHow so? The way I see it is last night's results supported his argument.Bullnamed_gus wrote: ↑February 21st, 2024, 7:22 amBad day for the Stanford Aggie Agenda.
Oh well, NET is just a sorting tool. As long as we stay top 35/40 and no bad losses we will be fine.
Our seed floor is an 8, with a NM win and a few wins in the tournament, we could probably get to a 5.
Furthermore, the NET algorithm gives a bonus for quality wins, particularly away from home. It doesn't seem to be a huge bonus. However, as I noted in my other thread, if you look at the teams whose NET is much higher than their KenPom ranking, all of those teams had a very strong road/neutral record. If we assume that KenPom is an accurate measure of how good a team really is (and hence cannot be "gamed"), then the best way to maximize your NET is to win road/neutral games. Maybe there is a way to game KenPom (which is part of the NET formula), but it doesn't seem to be easy to do given that it usually tracks Vegas very closely.
-
- Posts: 2480
- Joined: November 15th, 2010, 6:36 pm
- Has thanked: 637 times
- Been thanked: 637 times
Re: NET ranking unchanged
The main component of the NET rankings seems to be the KenPom efficiency rankings, which are designed to predict the results of future games. As a result, if the result of a game is close to the predicted result, the rankings don't change much, because the model's predictions were accurate. But if the outcome of a game is very different from what the model predicted, then the model adjusts itself to account for the unexpected result. Hence the reason USU's ranking didn't change much when we beat SDSU but Wake had a huge jump after beating Pitt.Aglicious wrote: ↑February 21st, 2024, 10:59 pmWell, Pitt was NET 47 and Wake was NET 40 going into that game. Wake was favored by 6pts at home but ended up winning by 33! Beating the expectations by more than 5x vaulted WF.Slim80 wrote: ↑February 21st, 2024, 10:13 pmHow so? Winning does nothing for your net? That doesn’t make sense to me at all. I can see the argument that SDSU shouldn’t fall much given how close the game was. But USU not gaining a few points for beating a great SDSU team is a complete head scratcher to me. What win is more impressive, Wake over Pitt or USU over SDSU. The NET says the Wake win was better……definitely some flaws in this algorithm.ineptimusprime wrote: ↑February 21st, 2024, 8:48 amThe result of the game last night is what should have happened when the NET 18 team plays the NET 30 team on the NET 30 team’s home floor. I don’t take issue with the lack of movement.
What is a little silly is a team moving up a ton for blasting a bad team.
USU was favored by 2.5pts at home and won by 5.
Beating the expectations by large margins (RUTS) and thus affecting efficiency numbers is what moves the NET needle. While our win was great and gave us another Q1 win (maybe more valuable in the eyes of committee members), the results were about what was expected and therefore didn't affect the ratings of either team involved. Had we won by 12-15 points we may have moved a spot or two.
Overall, I would say that this is a feature, not a bug. The fairest way to rank teams is to give a higher ranking to teams that are likely to win future games. And rankings that consider margin of victory do the best job of predicting future results. Having said that, if I were designing these rankings, I would tweak a few things. The NCAA says that teams receive a bonus in the NET rankings for beating quality opponents, particularly away from home. And, as I have noted elsewhere, teams that have won a lot of road games seem to punch above their weight in the NET. But I would still make this bonus even bigger to give teams a stronger incentive to play these games.
Also, I would change the loss function on the KenPom efficiency model. (Warning: math geek talk in this paragraph.) I don't think the KenPom efficiency formulas are public, but I am virtually certain they are based on some type of linear regression model using squared error loss. In other words, the model tries to minimize the sum of the squared errors between the predicted outcome and the actual outcome. Squared error loss is standard for these types of models for various mathematical and technical reasons. But the result is that the effect of big outliers is greatly magnified, since you are squaring them. And I think that's probably a big part of the reason why you see such huge jumps in the NET after a blowout (or an unexpected result generally). These days it is not that hard to fit a model using absolute error loss or some sort of hybrid loss function that doesn't give such huge weight to outliers. I think that would give a more accurate model and also mostly eliminate the incentive to try to inflate one's net by RUTS.
- These users thanked the author StanfordAggie for the post:
- Aggies1888