Make me Laugh. Try to explain SOS - Strength of Schedule

[4down20]

I said I'm not arguing in favor of best to worst record. However, facts regarding wins , losses, and score are all that matter to me.

Ok, and exactly how are you going to apply those things?

 

[CJH9972]

Ok, and exactly how are you going to apply those things?

I know you will love this idea based on previous posts of yours. My own point system concept bases the value of wins and losses on an opponent's record. That said, I'm open to other ideas.

 

[4down20]

I know you will love this idea based on previous posts of yours. My own point system concept bases the value of wins and losses on an opponent's record. That said, I'm open to other ideas.

I'll just cut to the chase - No math poll should never be used to make such decisions. They aren't able to take into the account the same things humans do, and to ignore loads of data because of few stats that you call "facts" is just dumb.

 

[LawDawg]

I know you will love this idea based on previous posts of yours. My own point system concept bases the value of wins and losses on an opponent's record. That said, I'm open to other ideas.

Any pure mathematical method is biased based on the creators idea as to what the best "facts" are. That's why in the BCS they consider a number of computer polls ... each of them used different inputs they thought resulted in determining the best order of the teams. Many articles were written how the algorithms they used affected the results. As you point out, yours is just your subjective opinion reduced to math, but it's still opinion not pure fact.

You also say, "I'm not arguing in favor of best to worst record." Well, that is what we are all discussing isn't it. How do we best determine who are the best teams that should be in the playoff. What good are your "facts" if not to determine who is best?

 

[UCFhonors]

I know you will love this idea based on previous posts of yours. My own point system concept bases the value of wins and losses on an opponent's record. That said, I'm open to other ideas.

I too believe he’ll love your system (albeit probably doesn’t have an open mind like we do, or won’t admit to liking it or both)

It just dawned on me that your system is an objective way of determining quality wins and losses.

What FBS wouldn’t like that!!?

I know I do, my UCFriend.

#UCFacts

 

[ellupo]

55-24

Even a broken clock is right twice a day. Show us yours.

 

[Kevin12773]

#UCFail (as usual)

 

[CJH9972]

Any pure mathematical method is biased based on the creators idea as to what the best "facts" are. That's why in the BCS they consider a number of computer polls ... each of them used different inputs they thought resulted in determining the best order of the teams. Many articles were written how the algorithms they used affected the results. As you point out, yours is just your subjective opinion reduced to math, but it's still opinion not pure fact.

You also say, "I'm not arguing in favor of best to worst record." Well, that is what we are all discussing isn't it. How do we best determine who are the best teams that should be in the playoff. What good are your "facts" if not to determine who is best?

Rules are not biased if they apply equally to all teams without regard for identity and perception. Your argument makes all rules everywhere biased simply because someone is needed to invent them.

That said, I'm not attempting to determine the best team at all. My purpose is to determine placement, advancement, and ultimately a winner in a competition based on which teams best accomplish a shared objective. What rules to play by is simply a matter of preference.

 

[CJH9972]

I'll just cut to the chase - No math poll should never be used to make such decisions. They aren't able to take into the account the same things humans do, and to ignore loads of data because of few stats that you call "facts" is just dumb.

What things do humans take in account? How do we know they do this? How do we know what value is given to these things from one moment to the next so that teams may maximize their value regarding these things? Why is it that despite the loads of data they consider, the committee essentially ranks P5 teams by fewest losses 97% of the time with exceptions that don't really matter and slot G5 teams anywhere below several P5 teams with more losses by invoking an undefined SOS?

Boil this down to two teams. We can have A and B compete against each other according to a known objective (do X better than the other team) or we can have them play their schedule and hope when it is a close call, that a committee of experts cherry picks a rational that favors their team.

 

[Olyduck]

What "facts" do you think are so important?

I think you meant UCFacts

 

[UCFhonors]

What things do humans take in account? How do we know they do this? How do we know what value is given to these things from one moment to the next so that teams may maximize their value regarding these things? Why is it that despite the loads of data they consider, the committee essentially ranks P5 teams by fewest losses 97% of the time with exceptions that don't really matter and slot G5 teams anywhere below several P5 teams with more losses by invoking an undefined SOS?

Boil this down to two teams. We can have A and B compete against each other according to a known objective (do X better than the other team) or we can have them play their schedule and hope when it is a close call, that a committee of experts cherry picks a rational that favors their team.

The whole post is A MIC DROP.


#UCFacts

 

[UCFhonors]

I'll just cut to the chase - No math poll should never be used to make such decisions. They aren't able to take into the account the same things humans do, and to ignore loads of data because of few stats that you call "facts" is just dumb.

You have amazing hubris to think that you can account for more things than a computer can....

Really what you're doing is confirmation bias validation of cherry picking stats. To prove that: You would have to watch 14.2 of awake hour days every week of of 7 day weeks to see all the FBS teams play. Nobody does that. (The math: 130 teams /2 for 65 games. 65 * 3.5 hours per game = 227.5. 227.5 / 16 awake hours per day = 14.2 days every week for 12 weeks.)

I'm not calling for computers or humans to choose Playoff teams. I want on field results to choose teams, just like all other sports do.

#UCFacts

 

[4down20]

You have amazing hubris to think that you can account for more things than a computer can....

Really what you're doing is confirmation bias validation of cherry picking stats. To prove that: You would have to watch 14.2 of awake hour days every week of of 7 day weeks to see all the FBS teams play. Nobody does that. (The math: 130 teams /2 for 65 games. 65 * 3.5 hours per game = 227.5. 227.5 / 16 awake hours per day = 14.2 days every week for 12 weeks.)

I'm not calling for computers or humans to choose Playoff teams. I want on field results to choose teams, just like all other sports do.

#UCFacts

I have amazing hubris? I'm sorry, I thought I was the person who has done ranking systems in the past as well as also being the person who spends a large part of their days working on machine learning and AI. But hey, maybe you're the guy who has an extra computer that runs 24 hours a day doing nothing but training computers to learn instead of me.

If you could make an AI that even came close to being able to account for the things humans do, you could make billions. Google and many companies are already spending millions of dollars every year trying to develop what is called "artificial general intelligence". All you are doing here is showing me you don't understand the differences between human intelligence and computer intelligence.

That doesn't mean computers are useless. They are a great tool for humans to use. Because they can take the stats from every single game and apply the math to all teams without bias. It's extremely valuable and is something a human is incapable of. However in the end a human is needed to interpret the results and account for all the things a computer is unable to account for. Weather, injuries, data bubbles, getting better as the year goes on, etc.

But FYI, the computers are generally more harsh against teams with lower SoS's, not nicer. So I'm not sure why someone such as yourself would be big on computers, other than the formulas vary enough that maybe some crappy one out of the 100+ will pick your team as the champion because it doesn't put enough weight into SoS. Which is basically like looking around town for the dumbest person to crown you champion.

 

[4down20]

What things do humans take in account? How do we know they do this? How do we know what value is given to these things from one moment to the next so that teams may maximize their value regarding these things?

Humans are able to take into account things like weather, if a team had injuries for a game. If a team had a QB change midseason. Take Alabama vs Florida St 2 years ago. FSU comes into that game ranked in the top5. They lose their returning starting QB and have to put in a freshman for the rest of the year and they got 6-7 or some shit.

How does a computer adjust for this? The answer: It doesn't. It has no idea that Alabama played a better FSU team than others. But that is the kind of quick adjustment that hard rules/computers aren't capable of that humans are. And there are tons of little things like this that happen constantly over the year.

So what smart people do is they look at the data and adjust. So you say - oh well Alabama had a SoS ranked 33rd while counting FSU as a 6-7 team, and you can personally adjust for it. Now you can't do these kinds of things across all teams as humans, but when you get to the end of the year you just need to look at a few teams.

This is all especially if you ignorantly limit the data to only scores. I wouldn't even waste my time with such a shitty metric.

Why is it that despite the loads of data they consider, the committee essentially ranks P5 teams by fewest losses 97% of the time with exceptions that don't really matter and slot G5 teams anywhere below several P5 teams with more losses by invoking an undefined SOS?

#1 where do you get "undefined SoS" As far as the SoS they are allowed to look at...wait for this - they only have access to exactly the formula you citing wanting earlier, win%.

#2 Your premise is fake. Show me the example of this happening. Why is it that people such as yourself never provide examples, just "what-ifs"? And it's really not just "what-ifs", because you provide your own conclusions and refuse to accept any other possible conclusion all for something that hasn't even happened.

And there are G5 teams who play tougher schedules, but they don't win them very often. The same year UCF went undefeated on the easiest schedule in football, Fresno St went all year with like 2 losses in the regular season on a schedule ranked in the top30. 1 of it's losses was to Alabama and the other loss was to Washington IIRC. They beat Boise St in the regular season(a better win than UCF had), but lost to Boise St in the conference champion.

Is it just a coincidence that the G5 teams that play the tougher schedules lose the bigger games while the G5 teams that play the easiest possible schedules sometimes go undefeated? Fuck no it's not.

So maybe come up with a real example instead of making up shit?

Boil this down to two teams. We can have A and B compete against each other according to a known objective (do X better than the other team) or we can have them play their schedule and hope when it is a close call, that a committee of experts cherry picks a rational that favors their team.

"known objective".

Ok, so how the fuck are you going to measure "do X better than the other team"....

 

[LawDawg]

Rules are not biased if they apply equally to all teams without regard for identity and perception. Your argument makes all rules everywhere biased simply because someone is needed to invent them.

That said, I'm not attempting to determine the best team at all. My purpose is to determine placement, advancement, and ultimately a winner in a competition based on which teams best accomplish a shared objective. What rules to play by is simply a matter of preference.

You realize this makes no sense? Are you denying that bias can be included in the choice of the algorithm you choose to use? That is not a debatable point. You've picked certain metrics that you think determine placement properly. It contains your bias as to what that means. If all teams can't play each other, then your system can't be exact, and any way you work the math contains your bias. I am not saying your bias is a bad thing ... you may the one person who finally figured it all out and your bias is perfect. But to state that your mathematical system is not biased is simply wrong.

 

[CJH9972]

You realize this makes no sense? Are you denying that bias can be included in the choice of the algorithm you choose to use? That is not a debatable point. You've picked certain metrics that you think determine placement properly. It contains your bias as to what that means. If all teams can't play each other, then your system can't be exact, and any way you work the math contains your bias. I am not saying your bias is a bad thing ... you may the one person who finally figured it all out and your bias is perfect. But to state that your mathematical system is not biased is simply wrong.

Preference for different criteria and/or different weights for the same criteria is not bias in favor of specific teams. The rules I favor apply equally to all teams without regard to identity and perception. The kind of bias that matters is the kind that requires the committee adopt a recusal policy so that voters cannot favor teams with whom they have allegiance. My rules have no allegiances.

 

[CJH9972]

Humans are able to take into account things like weather, if a team had injuries for a game. If a team had a QB change midseason. Take Alabama vs Florida St 2 years ago. FSU comes into that game ranked in the top5. They lose their returning starting QB and have to put in a freshman for the rest of the year and they got 6-7 or some shit.

How does a computer adjust for this? The answer: It doesn't. It has no idea that Alabama played a better FSU team than others. But that is the kind of quick adjustment that hard rules/computers aren't capable of that humans are. And there are tons of little things like this that happen constantly over the year.

So what smart people do is they look at the data and adjust. So you say - oh well Alabama had a SoS ranked 33rd while counting FSU as a 6-7 team, and you can personally adjust for it. Now you can't do these kinds of things across all teams as humans, but when you get to the end of the year you just need to look at a few teams.

This is all especially if you ignorantly limit the data to only scores. I wouldn't even waste my time with such a shitty metric.



#1 where do you get "undefined SoS" As far as the SoS they are allowed to look at...wait for this - they only have access to exactly the formula you citing wanting earlier, win%.

#2 Your premise is fake. Show me the example of this happening. Why is it that people such as yourself never provide examples, just "what-ifs"? And it's really not just "what-ifs", because you provide your own conclusions and refuse to accept any other possible conclusion all for something that hasn't even happened.

And there are G5 teams who play tougher schedules, but they don't win them very often. The same year UCF went undefeated on the easiest schedule in football, Fresno St went all year with like 2 losses in the regular season on a schedule ranked in the top30. 1 of it's losses was to Alabama and the other loss was to Washington IIRC. They beat Boise St in the regular season(a better win than UCF had), but lost to Boise St in the conference champion.

Is it just a coincidence that the G5 teams that play the tougher schedules lose the bigger games while the G5 teams that play the easiest possible schedules sometimes go undefeated? Fuck no it's not.

So maybe come up with a real example instead of making up shit?



"known objective".

Ok, so how the fuck are you going to measure "do X better than the other team"....

Things like weather and injuries have no value in a competition's standings. These things like so many others are simply things where teams must play the hand they are dealt. With injuries, if Alabama finished 5th and out of the playoffs, should I bump their score to account for how much better FSU presumably would have fared if their QB had not been hurt? After doing that, do I go through the 4th place team's opponents results and adjust their value for every game won and lost due to injury. Where do you draw the line and how do you decide when an injury is significant enough or not?

As for SOS, I don't know what SOS the committee looks at and even if I did, that doesn't tell when and if any SOS differences make up for additional losses. That said, I don't know what fake premise you are referring to. You will have to point it out to me.

Finally, the object of the game is to win. Not sure why we need to consider much more than wins, losses, and score especially when, as mentioned before, record is highly favored the committee.

 

[LawDawg]

Preference for different criteria and/or different weights for the same criteria is not bias in favor of specific teams. The rules I favor apply equally to all teams without regard to identity and perception. The kind of bias that matters is the kind that requires the committee adopt a recusal policy so that voters cannot favor teams with whom they have allegiance. My rules have no allegiances.

Your biases, which I stated I am not using as a pejorative, result in an outcome that represents your bias. For example, your math doesn't include an SOS calculation, which is also math, or "facts" as you like to put it. That your "math" is different than others' math makes my point ... you each have biases as to what math you think is best. Let's say other people's math includes weather, geographic location, conference strength mathematically determined, or other inputs. We might agree that some of those are dumb ... but they are facts, just ones we might disagree with.

You seem like a smart guy, certainly you understand this. Again, I am not saying it is good or bad, just that the math you choose represents your bias as to the outcome that the math arrives at.

 

[LawDawg]

Things like weather and injuries have no value in a competition's standings. These things like so many others are simply things where teams must play the hand they are dealt. With injuries, if Alabama finished 5th and out of the playoffs, should I bump their score to account for how much better FSU presumably would have fared if their QB had not been hurt? After doing that, do I go through the 4th place team's opponents results and adjust their value for every game won and lost due to injury. Where do you draw the line and how do you decide when an injury is significant enough or not?

As for SOS, I don't know what SOS the committee looks at and even if I did, that doesn't tell when and if any SOS differences make up for additional losses. That said, I don't know what fake premise you are referring to. You will have to point it out to me.

Finally, the object of the game is to win. Not sure why we need to consider much more than wins, losses, and score especially when, as mentioned before, record is highly favored the committee.

Your reference to the weather and injuries reveals a bias you have ... you simply dismiss it, yet it is facts or math as you call it. There is no subjectivity to it. You just don't like it.

As to the objective being to win, of course it is. But, if you think all wins are equal you've lost before you've begun. G5 teams play far inferior competition, as an example. If you don't factor that in, then your rankings simply aren't legitimate.

 

[4down20]

Things like weather and injuries have no value in a competition's standings. These things like so many others are simply things where teams must play the hand they are dealt. With injuries, if Alabama finished 5th and out of the playoffs, should I bump their score to account for how much better FSU presumably would have fared if their QB had not been hurt? After doing that, do I go through the 4th place team's opponents results and adjust their value for every game won and lost due to injury. Where do you draw the line and how do you decide when an injury is significant enough or not?

As for SOS, I don't know what SOS the committee looks at and even if I did, that doesn't tell when and if any SOS differences make up for additional losses. That said, I don't know what fake premise you are referring to. You will have to point it out to me.

Finally, the object of the game is to win. Not sure why we need to consider much more than wins, losses, and score especially when, as mentioned before, record is highly favored the committee.

Things like weather and injuries have tons of value when you are comparing what happened in games. If you don't understand the large affect these things can have on the game, then what the fuck are you even in this discussion for?

And then you go further and talk about you don't know what SoS the committee uses. It sounds to me like you don't know much about the topic you keep speaking on at all. But once again, it's based on win%.

Not sure why we need to consider more stats? Because you think more data is bad, or because you are just incapable of understand things beyond that and must then automatically assume it's bad? Because I mean, understanding how good a rush defense is and if a team that has been putting up big numbers against easier teams and then doesn't much against that good defense - well that's just useless information right? More likely, you don't understand how to come to that conclusion mathematically and that's why you're "not sure why we need to consider" stuff that matters.

"Anything I don't understand is invalid".

Or are you going to tell me beating Clemson and UTEP are equal wins. If they aren't equal wins, then obviously you need to find out why and the math needs to support it.