The unofficial online home for fans of the Old Dominion University Monarchs. Part of Collegiate Sports Nation (CSN).

(04-05-2023 10:09 AM)757ODU Wrote:  

(04-05-2023 10:01 AM)Monarchblue Wrote:  

(04-05-2023 09:37 AM)757ODU Wrote:  

(04-05-2023 09:21 AM)Monarchblue Wrote:  

(04-05-2023 07:50 AM)757ODU Wrote:  People trying to discredit KenPom is hilarious to me. It is factual information being plugged into kenpom, and being judged over an entirety of a season. Good grief man...some of the takes on her are absolutely horrible.

It is pieces of factual information that are chosen by a person or group of people. I could pull some random ass stats together and say this is how it ranks teams, but that doesn't make my ranking a fact. It is probably better than any other computer rating system, and does a pretty decent job of ranking teams' resumes (less so how good they are at a given time), but what bothers me is the idea that "it is just facts and data, so you can't argue with it", when in reality it is one person's interpretation of data.

Points per possession is a fact, offensive rebounding 5 is a fact, turnover % is a fact, effective fg % is a fact. All of these statistics are facts.

I don't disagree that those are facts. I disagree that how the data is parsed and the datpoints that are chosen to be used and/or emphasized are selected by humans. I could take the same "factual" data and parse it differently and come up with a different set of rankings.

What would be your opinion on how accurate it has been over the years?

I think it is fairly accurate, and probably more accurate than any other model, but not accurate enough to be treated as gospel like it is by people who are data first in their evaluation. It is also not accurate enough for the Chair of the Selection Committee to laugh at the idea of using the eye test on any level "because we have data".

Data models should always be a tool to help people evaluate reality. We get in trouble, as we have seen recently with gov't response to certain things that I won't raise here in hopes of not opening a can of worms, when we hold data models up as beyond scrutiny, we are really setting ourselves up to make some decisions that may not be correct. At the end of the day, data that has been pulled together to form a picture is not unbiased, as it is often represented to be. It is ultimately one person's, or a group of people's, interpretation of the story that the raw data tells.