Statistics tell the whole story. Independently verified?

#51
#51
GrowVol said:
Hungry and determined D lineman don't make a difference in seasonal outcomes??? are you joking???
Click to expand...

I didn't say they can't make a difference. Just that when looking for seasonal outcomes, you can ignore those sort of things and still come up with very strongly correlated results (talent=success). Without counting for who is hungry one day, or who is upset that their girlfriend broke up with them the next day, you can get to a prediction of roughly 70%.

If you aren't satisfied until you get a prediction rate of 100% then come up with a way to quantify "passion" and all of the other intangibles that you think are important, test it, and publish your results. I would love to read about them, and I am not being sarcastic.
 
Last edited:
#53
#53
Okay, let me say it this way, for the last time. For those of you who think that there are other things that contribute to the outcome of games, you are correct. At best, however, you are trying to alleviate the 1/3 of the outcomes that are not already predicted by simple talent averages. So, if you want to invest an almost infinite amount of time trying to predict the effect of the moon cycle, tides, presidential elections, when a butterfly flaps its wings in China, or what color underwear the back up quarterback wears at half-time, be my guest. At best you will take an evaluation that already works the vast majority of the time, and make it more predictive. At worst, you will find that you are more likely describing superstitions than actual determiners of game outcomes.
 
  • Like
Reactions: 1 person
#55
#55
daj2576 said:
I didn't say they can't make a difference. Just that when looking for seasonal outcomes, you can ignore those sort of things and still come up with very strongly correlated results (talent=success). Without counting for who is hungry one day, or who is upset that their girlfriend broke up with them the next day, you can get to a prediction of roughly 70%.

If you aren't satisfied until you get a prediction rate of 100% then come up with a way to quantify "passion" and all of the other intangibles that you think are important, test it, and publish your results. I would love to read about them, and I am not being sarcastic.
Click to expand...

Thanx man, I was thinking of intangibles, which i guess can't be put into a number ratio.
 
#56
#56
GrowVol said:
Not trying to be a jerk or sunshine pumper, But in any formulated analysis, are Seniors that are fed up with losing taken into account? I believe on a shear fighting spirit, and a spirit of "not again", it changes everything. Can that be figured into systems modeled for predictions?
Click to expand...

Good point but in our case.....it seem not to be a factor.
 
#58
#58
Interesting read, but predicting 7-5 or 8-4 for UT isn't that hard:

Moderate to significant talent advantage that will equal wins:

APSU
WKU
S. Alabama
UK

Roughly Equal Talent:

Missouri
Vandy
Auburn

Slight Disadvantage in talent:

USC
UGA

Considerable Disadvantage:

Oregon
UF
Bama

So take the 1st 4 teams mentioned wins and the last 3 as losses. That means UT will need to win 3 or 4 among:

UGA- Home
USC- Home
Missouri- Away
Auburn- Home
Vandy- Home

Seems doable to me.
 
  • Like
Reactions: 1 person
#59
#59
Tired of predictions. Tired of myself making predictions and justifying to myself or others. I always get baited in and fall in line, yet when I bypass the what if's altogether, my stomach loosens up. Maybe, mods can combine all the prediction threads into one, I don't know. Guess we just 3 got more weeks.
 
#60
#60
I like the simplicity of cfmatrix system of prediction. Often when building a predictive data model people will inadvertently create an issue wherein the same piece of data is considered twice.

For instance, let's say I want to use recruiting ranking as a predictive stat because there is a ton of easily gathered data out there and there is a proven correlation between aggregate recruiting rankings and success. Let's say I get a 5-star composite ranking recruit and he becomes an All-American. Do I give my data a positive nudge? No. That is already inherent in the recruiting rankings. That player merely fulfilled potential.

Next, let's imagine that player never got off the practice team. Shouldn't we then downgrade his value in our model? No. All this should even out over a 5-year recruiting cycle--some players will over-perform and some will under-perform their composite recruiting ranking. There is no real need to try and refine the data bit by bit if you are merely trying to predict wins and losses.

I would also account for attrition, coaching, and home-field advantage, but all that data could not be said to be part of the recruiting rankings. I'd say include injuries as well, but you can't do that if you are predicting the season in advance.

A simple system like this is going to accurately predict a majority of games.

For instance, Tennessee will beat Vanderbilt this year. Why? We objectively have more talented players. We are playing in Neyland. Our coaches will not be a detriment to our game plan and execution. It really is that simple. (In fact, I expect Coach Jones to have a incredible game plan for this one and smother the spark that is happening in Nashville before it becomes a flame).

There will always be games that are decided by an odd bounce or a bad call (Memphis 1996 for instance) but no statistical system will help you there. That is where the so-called "intangibles" come into play on the player-side, but what Coach Jones calls "mental conditioning" is a part of that as well so that leads us back to coaching and recruiting.

Obviously everything can be quantified. If Stephen Hawking can figure out that there must be black holes in in the Universe by doing math, then we should surely be able to figure out the chances of Tennessee getting to a bowl without too much work. I think we'll be 8-5 when the dust clears.
 
  • Like
Reactions: 2 people
#61
#61
Boca Vol said:
Interesting read, but predicting 7-5 or 8-4 for UT isn't that hard:

Moderate to significant talent advantage that will equal wins:

APSU
WKU
S. Alabama
UK

Roughly Equal Talent:

Missouri
Vandy
Auburn

Slight Disadvantage in talent:

USC
UGA

Considerable Disadvantage:

Oregon
UF
Bama

So take the 1st 4 teams mentioned wins and the last 3 as losses. That means UT will need to win 3 or 4 among:

UGA- Home
USC- Home
Missouri- Away
Auburn- Home
Vandy- Home

Seems doable to me.
Click to expand...

I agree with your conclusion, that a 7-5 or 8-4 season is predictable. But, the way you get there is baffling to anyone who has seen these numbers or their correlation to outcome. Even this Bartoo guy (who is an admitted Oregon fan) says that the UT v. Oregon game is going to be much closer than most would expect.

Other than that you are on the right track. The numbers basically show the following:

UT has significantly less talent than Bama.

Slightly less talent than Auburn, Florida and Georgia.

Roughly equivalent talent to Oregon.

Slightly more talent than SCAR.

Significantly more talent than Missouri, Kentucky, Vanderbilt, Western Kentucky, South Alabama and Austin Peay.

So, when I say 7-5 I am predicting losses to Georgia, Florida, Bama, Oregon and Auburn.

But, you can mix and match however you want to get to the 7-5/8-4 range and have roughly the same outcome. Jones has the potential to over-perform and beat a team who is truly superior, I am just not counting that in my evaluation.
 
#62
#62
Rifleman said:
I like the simplicity of cfmatrix system of prediction. Often when building a predictive data model people will inadvertently create an issue wherein the same piece of data is considered twice.

For instance, let's say I want to use recruiting ranking as a predictive stat because there is a ton of easily gathered data out there and there is a proven correlation between aggregate recruiting rankings and success. Let's say I get a 5-star composite ranking recruit and he becomes an All-American. Do I give my data a positive nudge? No. That is already inherent in the recruiting rankings. That player merely fulfilled potential.

Next, let's imagine that player never got off the practice team. Shouldn't we then downgrade his value in our model? No. All this should even out over a 5-year recruiting cycle--some players will over-perform and some will under-perform their composite recruiting ranking. There is no real need to try and refine the data bit by bit if you are merely trying to predict wins and losses.

I would also account for attrition, coaching, and home-field advantage, but all that data could not be said to be part of the recruiting rankings. I'd say include injuries as well, but you can't do that if you are predicting the season in advance.

A simple system like this is going to accurately predict a majority of games.

For instance, Tennessee will beat Vanderbilt this year. Why? We objectively have more talented players. We are playing in Neyland. Our coaches will not be a detriment to our game plan and execution. It really is that simple. (In fact, I expect Coach Jones to have a incredible game plan for this one and smother the spark that is happening in Nashville before it becomes a flame).

There will always be games that are decided by an odd bounce or a bad call (Memphis 1996 for instance) but no statistical system will help you there. That is where the so-called "intangibles" come into play on the player-side, but what Coach Jones calls "mental conditioning" is a part of that as well so that leads us back to coaching and recruiting.

Obviously everything can be quantified. If Stephen Hawking can figure out that there must be black holes in in the Universe by doing math, then we should surely be able to figure out the chances of Tennessee getting to a bowl without too much work. I think we'll be 8-5 when the dust clears.
Click to expand...

An absolute pleasure to read. Thank you, and those like you.

Sometimes I feel like this sort of evaluation is placed in the "Axis of Evil" because "It involves math, and math is very much a part of the Axis of Evil". SNL - Will Ferrel - George W Bush and the Axis of Evil Video - paulkaoke - MyVideo
 
#63
#63
GrowVol said:
Tired of predictions. Tired of myself making predictions and justifying to myself or others. I always get baited in and fall in line, yet when I bypass the what if's altogether, my stomach loosens up. Maybe, mods can combine all the prediction threads into one, I don't know. Guess we just 3 got more weeks.
Click to expand...

This isn't intended to be a prediction thread, necessarily. This is a thread about predictive systems, which is a much larger issue than just trying to pull numbers out of the air and justify them.
 
#64
#64
daj2576 said:
This isn't intended to be a prediction thread, necessarily. This is a thread about predictive systems, which is a much larger issue than just trying to pull numbers out of the air and justify them.
Click to expand...

VN is the predictive system.:thumbsup:
 
#67
#67
daj2576 said:
Everyone: I am also having trouble pulling the website up on mobile. I have no trouble accessing it on a computer, any computer. Are those of you who are saying that you have trouble viewing the link doing so on a mobile device?
Click to expand...

Yeah I haven't been able to view any links on my phone for almost 2 months.
 
#68
#68
daj2576 said:
Roughly equivalent talent to Oregon.
Click to expand...

Maybe if you simply go by what a subjective recruiting service ranked these kids when they were 17 years old. Those recruiting rankings don't account for attrition, coaching, systems, improvement, or digression.
 
#70
#70
Heres the thing that makes life so interesting. The theory of evolution claims that only the strong shall survive. Maybe so...maybe so...But the theory of competition says just because their the strong doesn't mean they cant get their ***'s kicked. Thats right. See what every long shot, come from behind underdog will tell you is this. The other guy may in fact be the favorite, the odds maybe stacked against you, fair enough. But what the odds don't know is this isn't a math test. This is a completely different kind of test. One where passion has a funny way of trumping logic. So before you step up to the starting line, before the whistle blows, and the clock starts ticking. Just remember out here the results don't always add up. No matter what the stats may say, and the experts may think, and the commentators may have predicted, when the race is on all bets are off. Don't be surprised if someone decides to flip the script and take a pass on yelling uncle. And then suddenly as the old saying goes,we got ourselves a game.
 
#71
#71
Boca Vol said:
UGA- Home
USC- Home
Missouri- Away
Auburn- Home
Vandy- Home
Click to expand...

While I disagree that UGA and SC are only slightly more talented than us(they're in a different zip code than us talent and depth-wise), I think you've hit on a key advantage with the schedule.
 
#72
#72
Boca Vol said:
Maybe if you simply go by what a subjective recruiting service ranked these kids when they were 17 years old. Those recruiting rankings don't account for attrition, coaching, systems, improvement, or digression.
Click to expand...

I really don't know how else to say it.

None of that matters. You can get to a prediction rate of around 70% without having to factor ANY of that in. This CFB matrix guy actually factors in attrition to his numbers and gets a higher correlation than I do, which is expected.

What it also proves is that coaching, systems and everything else fall within the 30% or so of the games that aren't predicted by this very simple numerical analysis.

It isn't hard to understand, it is just hard to let go of years of bad information given to you by coaches, ESPN, your own heart and head.
 
Last edited:
  • Like
Reactions: 1 person
#73
#73
JuggernaUT_VOL said:
Just wandering if the player analysis is by star rating or by overall postion rating?
Click to expand...

The way I do it is by the way they are ranked. I don't look at star averages or any individual positions (not for this evaluation anyway). While those things are taken into consideration in the rankings, they aren't specifically reviewed for my analysis. It is incredibly simple, and that is probably why people have such a hart time understanding the high performance of it all.

So if Bama is ranked one, I don't care how many points or average stars they have. I rank them one. If Florida is two, same thing. I do that for all the teams four years including this year, then rank them by those averages.

This CFB matrix guy does that for each recruiting service, then averages those and also accounts for attrition. His correlation is about 10% higher than mine, but I figure almost 70% prediction rate is good enough (he is closer to 80%).
 
#74
#74
GAVol said:
While I disagree that UGA and SC are only slightly more talented than us(they're in a different zip code than us talent and depth-wise), I think you've hit on a key advantage with the schedule.
Click to expand...

I can be in a different zip code than you and live next door. :)
 
  • Like
Reactions: 1 person
#75
#75
Vol4life25 said:
Heres the thing that makes life so interesting. The theory of evolution claims that only the strong shall survive. Maybe so...maybe so...But the theory of competition says just because their the strong doesn't mean they cant get their ***'s kicked. Thats right. See what every long shot, come from behind underdog will tell you is this. The other guy may in fact be the favorite, the odds maybe stacked against you, fair enough. But what the odds don't know is this isn't a math test. This is a completely different kind of test. One where passion has a funny way of trumping logic. So before you step up to the starting line, before the whistle blows, and the clock starts ticking. Just remember out here the results don't always add up. No matter what the stats may say, and the experts may think, and the commentators may have predicted, when the race is on all bets are off. Don't be surprised if someone decides to flip the script and take a pass on yelling uncle. And then suddenly as the old saying goes,we got ourselves a game.
Click to expand...

You are right, upsets happen. Just not nearly as often as you would be lead to believe. Seriously, using this system, you can get to the correct answer for almost 70% of the games. One of the other problems is what makes for an upset. Frequently you will find teams that are ranked higher than others who are vastly inferior in talent, but their schedule is more favorable. In other words, when you play a team that is ranked higher than you but they actually have less talent than your team, it isn't an upset when you win. That effect is precisely why Oregon is usually ranked very high, but that is another subject.
 
New posts
You must log in or register to reply here.

VN Store



Back
Top