SOS - What it does and doesn't tell us

[molly jabali]

For some weird reason, my brain flashed back to Sean Connery SNL skits upon reading your post.

"Trebek"..... love it....

[Jamar Howard 4 President]

Were they all home games? Why have team A and team B each only played 2 games while the rest of the country has played 21 games? Seems like they both have terrible SOS and should play a lot more games.

Thanks for the comedy. Let me know if you care to join my discussion.

[ShockerExpress]

I mean it just depends... No one can tell anything from two games. If this was the first two games of the season... I guess Team A. But barely, and with no confidence at all. No offense, but I don't know where you are going with this, and you sound preachy. It is literally impossible to come to any consensus based on two games, IMO.

[ShockBand]

I know you want to keep it simple, but just two games is too small of a sample size, leaving too much potential variance. But based on that limited information, I would go with Team B because you at least have TWO consistent results. For Team A, the beating of a top 10 is a huge plus, but with the second result being a very weak team, the data is unreliable at that point.

[Bob Sacamano]

Were they all home games? Why have team A and team B each only played 2 games while the rest of the country has played 21 games? Seems like they both have terrible SOS and should play a lot more games.

I'll fight you for the rights to that joke.

http://shockernet.net/forum/showthread.php?19049-2014-15-Games-of-Interest&p=538401&viewfull=1#post538401

[ShockBand]

Now if Team A were in a so-called "power" conference, and Team B was a Valley team, I know what the talking heads would say.

A! A! A!

[jdshock]

This conversation was started in the games of interest thread, but I felt that it would be better to move it into its own thread instead.

Original question:

Team A is 2-0. Beat a team ranked in the top 10 and a terrible 300+ team.
Team B is 2-0. Beat Vermont and Louisiana Lafayette, who are a couple of mediocre teams each ranked in the mid 100s and each with 11-10 records.

You know nothing else about either team. Based on admittedly very limited information, which team do you rank higher?

I've already seen some complaints that the premise isn't valid. I think the premise is very valid. If I take the next step in my logical progression and you feel my logic doesn't hold, then call me out on it. However, for now, I think this is a very valid starting point.

I'm waiting with bated breath for your next logical progression. I think all else equal, you obviously pick Team A. I don't think it's probably a fair question, though. If the teams were actually 2-0, that means it's super early in the season. Pre-season rankings are based solely on talent evaluations and hype about teams. If Team B started out ranked higher than Team A, then I just don't think it justifies dropping Team B below Team A because Team B did win the games they were supposed to do. If rankings were based on this type of analysis only, you'd have teams sky rocket up in the rankings until you had a much larger sample size.

And really, I was being slightly sarcastic about your next step.... but I would like to know where you are going with this. I think people think you're going to try to pull out the rug and say "GOTCHA" so people are hesitant.

[Jamar Howard 4 President]

Let's simplify the question even further.

Team A is 1-0, beat a top 10 team.
Team B is 2-0, both against teams around 150.

Information is clearly limited, but this would seem to me to be an obvious scenario to rank Team A higher assuming the information above is ALL that you know.

Do those of you who said Team B previously still say Team B in this scenario?

[Steeleshocker]

For someone who extols the value of Ken Pomeroy and how as the season progresses along more information is a good thing, this seems oddly simplistic.

If knowing the names on the jersey allows us to introduce more information, that is a good thing. Not something to avoid.

[ShockerExpress]

Let's simplify the question even further.

Team A is 1-0, beat a top 10 team.
Team B is 2-0, both against teams around 150.

Information is clearly limited, but this would seem to me to be an obvious scenario to rank Team A higher assuming the information above is ALL that you know.

Do those of you who said Team B previously still say Team B in this scenario?

This doesn't solve the problem that everyone has, and has already mentioned. There is literally no way to tell. Sure, A is better, based off one game. KU beats Duke, WSU beats Western Kentucky and Memphis. A week into the season, and we know nothing.

[im4wsu]

This conversation was started in the games of interest thread, but I felt that it would be better to move it into its own thread instead.

Original question:

Team A is 2-0. Beat a team ranked in the top 10 and a terrible 300+ team.
Team B is 2-0. Beat Vermont and Louisiana Lafayette, who are a couple of mediocre teams each ranked in the mid 100s and each with 11-10 records.

You know nothing else about either team. Based on admittedly very limited information, which team do you rank higher?

I've already seen some complaints that the premise isn't valid. I think the premise is very valid. If I take the next step in my logical progression and you feel my logic doesn't hold, then call me out on it. However, for now, I think this is a very valid starting point.

Question on validity. Why is Team B playing only two games but their opponents have played 21 games each? How many games have the opponents of Team A played?

[Jamar Howard 4 President]

Pre-season rankings are based solely on talent evaluations and hype...

Thanks for answering the question (you selected Team A), but the part of your answer that I just quoted here is the type of thing that is blowing my mind. Pre-season rankings. Whether or not we are talking about early in the season. Etc. NONE of that has any impact on my hypothetical. I've purposefully excluded all of it to make things simple.

If new information comes into play, you are all free to explain why you think the new information changes things. However, for now, everything outside my simple scenario is left outside for a reason.

[The Mad Hatter]

In spite of your injunction about premise, the only reasonable answer is to question the premise. There is simply not enough information to say. This is why RPI is an inferior tool to other statistical models (although even RPI looks at home/away which is information we aren't given here), just looking at wins and losses with no other contextual information gives mediocre ranking results. Any good evaluative system will look at also at a minimum at margin of victory in some form and preferably have an intelligent approach to more sophisticated metrics. The real key, however, is that the model and it's weighing system of these factors be set and followed precisely so that we don't do what we are doing here, which is arbitrarily picking games and instead of trusting the ranking the model gives us (with all the data that goes into it), trying to make an opinion out of a small sample size.

So my question is this: if I believe in a ranking system enough to believe that team A played a top 10 team and a 300+ team and that team B played 2 mid 100 teams, why would I doubt what the ranking system said about teams A & B? I can't reasonably claim that the ranking system knew what it was talking about in regard to the opponents, but not the teams in question. I could say the ranking system is flawed, but then I couldn't use it as a gauge of SOS.

[Jatrain]

Let's simplify the question even further.

Team A is 1-0, beat a top 10 team.
Team B is 2-0, both against teams around 150.

Information is clearly limited, but this would seem to me to be an obvious scenario to rank Team A higher assuming the information above is ALL that you know.

Do those of you who said Team B previously still say Team B in this scenario?

This isn't really working out how you wanted it to so how about just moving on to whatever point you are trying to make?

[Cdizzle]

The fact that you are introducing RPI, how to "play it", scheduling philosophy, ect. tells me that you are off on some tangent that goes way outside of the basic question I initially proposed.

I'm trying to start out simple so we can begin from a point of agreement.

You post completely sidetracks that process.

You introduced RPI, or some facsimile in your original post.