Comparing WSU as a 1, and Kentucky as an 8

[RoyalShock]

I'm part way through it but it's an interesting read, so far. If you aren't a fan of computer models to evaluate teams, move along, nothing to see here.

https://www.numberfire.com/ncaab/news/1932/why-wichita-state-s-loss-was-historic

[Shockeriffic]

Interesting read but even for people that are a fan of computer models there's not much to see here. Here's the explanation they give for their ranking system:
nERD/nF Score (Team)
The team ranking is on a scale from 0-100, with 50 as the league average. This ranking is predictive of the team's ultimate winning percentage.

That tells me absolutely nothing about how it is calculated or even what the underlying methodology is. One reason I like using kenpom's numbers is that I at least know what his methodology is for his rankings. While kenpom has had a worse number 1 seed than us every single season for the last 5 years, numberfire has us as one of the worst ever. It's impossible to tell what the reason for this might be since their explanation for the ranking system is terrible.

[Flip1381]

Interesting read but even for people that are a fan of computer models there's not much to see here. Here's the explanation they give for their ranking system:
nERD/nF Score (Team)
The team ranking is on a scale from 0-100, with 50 as the league average. This ranking is predictive of the team's ultimate winning percentage.

That tells me absolutely nothing about how it is calculated or even what the underlying methodology is. One reason I like using kenpom's numbers is that I at least know what his methodology is for his rankings. While kenpom has had a worse number 1 seed than us every single season for the last 5 years, numberfire has us as one of the worst ever. It's impossible to tell what the reason for this might be since their explanation for the ranking system is terrible.

Exactly.

Does anyone understand why NumberFire says the nERD metric is 0-100 with 50 being the league average, but then says that UK was at 14.xx and WSU was at 15.xx? What am I missing?
I'm usually not this dense, but since I don't know how they calculate the nERD team rating to begin with, I'm not sure what the UK and WSU rating mean.

[RoyalShock]

I'm not positive, but I think the 0-100 is used to determine what an "average" team is, in order to determine the number that says how much a team would beat the average team by.

[CBB_Fan]

I have started my own stat system, mostly similar to Massey rating system. Massey uses only three variables to create his ratings: the score, the venue, and the date. He creates an algorithm that predicts the likelihood of one team winning a rematch with the other after they have played. This connects the two teams, in a way similar to how RPI connects one team to its opponents, and its opponents' opponents. However, Massey goes further and repeats the process many times (opponents' opponents' opponents' &c).

The key is that Massey doesn't just assign a numerical strength value to teams. He also finds the standard deviation of their results. My method does something similar, but its goal is not to determine pure team strength from scores but the impact they have on their opponents stats (ie, how Wichita State affects normal offensive rebounding percentage), with further analysis based on the stats themselves which is similar to the way KenPom assesses teams.

The point isn't to hype my method, which I don't use to compare teams on any forum and is purely for internal use. The point is that winning involves not only your overall level of play but also your consistency. A team with a very high standard deviation between individual results can knock off a good team but probably isn't going to win a championship, whereas sometimes a team with lower overall values is expected to have a better winning percentage simply because they are more consistent.

In both my model and Massey's we had the lowest standard deviation of any 1 seed in the time set I measured (2004-2014). Kentucky, on the other hand, had a VERY high standard deviation, and an overall strength that corresponded more to a 4 seed than an 8. Against a normal regional, my method gave us a relatively high chance of going to the Final Four, because even though a fair number of teams ranked higher in pure strength we had the least chance of "dropping" a game through poor performance. However, Kentucky was the 8-9 had the highest chance of upsetting a #1 seed in that time frame by a fair margin.

[wsushox1]

I have started my own stat system, mostly similar to Massey rating system. Massey uses only three variables to create his ratings: the score, the venue, and the date. He creates an algorithm that predicts the likelihood of one team winning a rematch with the other after they have played. This connects the two teams, in a way similar to how RPI connects one team to its opponents, and its opponents' opponents. However, Massey goes further and repeats the process many times (opponents' opponents' opponents' &c).

The key is that Massey doesn't just assign a numerical strength value to teams. He also finds the standard deviation of their results. My method does something similar, but its goal is not to determine pure team strength from scores but the impact they have on their opponents stats (ie, how Wichita State affects normal offensive rebounding percentage), with further analysis based on the stats themselves which is similar to the way KenPom assesses teams.

The point isn't to hype my method, which I don't use to compare teams on any forum and is purely for internal use. The point is that winning involves not only your overall level of play but also your consistency. A team with a very high standard deviation between individual results can knock off a good team but probably isn't going to win a championship, whereas sometimes a team with lower overall values is expected to have a better winning percentage simply because they are more consistent.

In both my model and Massey's we had the lowest standard deviation of any 1 seed in the time set I measured (2004-2014). Kentucky, on the other hand, had a VERY high standard deviation, and an overall strength that corresponded more to a 4 seed than an 8. Against a normal regional, my method gave us a relatively high chance of going to the Final Four, because even though a fair number of teams ranked higher in pure strength we had the least chance of "dropping" a game through poor performance. However, Kentucky was the 8-9 had the highest chance of upsetting a #1 seed in that time frame by a fair margin.

What are you using to model it?

[wsushox1]

I should love big data more than I do considering my future (potential) career but right now I'm just too busy to deal with it.

[Rosewood]

I should love big data more than I do considering my future (potential) career but right now I'm just too busy to deal with it.

Big data is like teenage sex: everyone talks about it, nobody really knows how to do it, everyone thinks everyone else is doing it, so everyone claims they are doing it...

[rialaigh]

Big data is like teenage sex: everyone talks about it, nobody really knows how to do it, everyone thinks everyone else is doing it, so everyone claims they are doing it...

My data is bigger than your data! No really...want to compare data?

[ShockBand]

My data is bigger than your data! No really...want to compare data?

[wsushox1]

Big data is like teenage sex: everyone talks about it, nobody really knows how to do it, everyone thinks everyone else is doing it, so everyone claims they are doing it...

So effing true.