2018 Pre-Season Probability Poll Discussion Thread

18in32

Petard Hoister
Joined
May 23, 2010
Messages
27,979
In reality, as some have pointed out, a voter's confidence in our likelihood of winning or losing a game is somewhat independent of the scoring outcome. One could be equally highly confident in us beating Bowling Green and UVA, yet you might have different expectations regarding the score differential.
I don't think it's possible to come up with a system that accomplishes what we're trying to accomplish, until we're clear on what we're trying to accomplish. And this sentence suggests to me that we aren't.

Someone may indeed be equally confident that we'll beat Bowling Green and UVA, but if they have different expectations regarding the score differential, then their confidence is likely not persuasive to someone with different intuitions. What we're trying to measure isn't whose irrational confidence happens to be correct across a tiny data set, but whose knowledge of football leads them to predict most accurately what the future holds.

So I would say that if you believe GT will beat UVA by a tiny margin, your confidence in the victory cannot be high. (I would also say you're a very young Tech fan if you're confident of any victory over UVA, but that's a different thread.)
 

coit

Y’all got any more of that D Fence?
Joined
Nov 29, 2007
Messages
87,938
I don't think it's possible to come up with a system that accomplishes what we're trying to accomplish, until we're clear on what we're trying to accomplish. And this sentence suggests to me that we aren't.

Someone may indeed be equally confident that we'll beat Bowling Green and UVA, but if they have different expectations regarding the score differential, then their confidence is likely not persuasive to someone with different intuitions. What we're trying to measure isn't whose irrational confidence happens to be correct across a tiny data set, but whose knowledge of football leads them to predict most accurately what the future holds.

So I would say that if you believe GT will beat UVA by a tiny margin, your confidence in the victory cannot be high. (I would also say you're a very young Tech fan if you're confident of any victory over UVA, but that's a different thread.)
I was just throwing out UVA as an example. If we played Syracuse or WF I'd say them I guess. Point being that we expect to blow out BG but we don't typically expect to blow out any ACC opponent although most of us would have high confidence that we would win against the weaker ACC foes.

As for the system, what we need is a weighting function that peaks when you pick a game a confidence level that matches the score values we've selected, and falls off on either side. a square root of the square of the inverse of the differential would work but I'm not sure how to deal with the inverse of zero that is produced when you are spot on. The weighting factor for that case could be 1 with all other weighting factors lower but I just need to figure out how to implement that.
 

GT76

Damn Good Rat
Joined
Jul 13, 2016
Messages
1,054
I was just throwing out UVA as an example. If we played Syracuse or WF I'd say them I guess. Point being that we expect to blow out BG but we don't typically expect to blow out any ACC opponent although most of us would have high confidence that we would win against the weaker ACC foes.

As for the system, what we need is a weighting function that peaks when you pick a game a confidence level that matches the score values we've selected, and falls off on either side. a square root of the square of the inverse of the differential would work but I'm not sure how to deal with the inverse of zero that is produced when you are spot on. The weighting factor for that case could be 1 with all other weighting factors lower but I just need to figure out how to implement that.
Ok, so the game is being changed from probability to point spread. That’s probably a good idea. The problem with probability as used here is it was undefined; there was no standard. It is apparent people have different ideas about what you need to have a 90%+ probability of a win. One person might think thats 24 points. Another might think it’s 15. Others more or less; everyone has their own standard. A defined point spread sets a standard so picking a range and using that for “probability” is a good idea.
As far as scoring it, you’ve got all season to figure that part out!
 

GT76

Damn Good Rat
Joined
Jul 13, 2016
Messages
1,054
What if I'm absolutely certain we are going to win by one point?
That was why I said probability and point spread are two different things but some feel different. And that is why the point spread thing at least makes a standard so we’re all playing on a level field. Pick your .5 or .6 probability or whatever and be a winner
 

smokey_wasp

Dodd-Like
Joined
Nov 25, 2013
Messages
11,021
I was mostly joking. I think I get it for the most part...for a sidewalker humanities graduate anyway.
 

18in32

Petard Hoister
Joined
May 23, 2010
Messages
27,979
Because some people are absolutely certain we are going to win by one point?
A kind of certainty which has no explanation is not relevant to what I thought we were measuring.

I start every year certain this will be 1990 redux... my reasons? 'Because why not.'
 

18in32

Petard Hoister
Joined
May 23, 2010
Messages
27,979
I was mostly joking. I think I get it for the most part...for a sidewalker humanities graduate anyway.
The last person you want to trust with an evaluation of "certainty" is an engineer. Kinda like the last person you want to trust to measure happiness is a social scientist.
 

coit

Y’all got any more of that D Fence?
Joined
Nov 29, 2007
Messages
87,938
I’ll be showing results all 3 different ways as shown on the spreadsheet so we can continue to debate and argue it as we go through the season. I’ve been reading some statistical tools that we could possibly use but I don’t see anything that’s directly applicable.
 

GSUsTALON

Flats Noob
Joined
Sep 9, 2014
Messages
589
I’ll be showing results all 3 different ways as shown on the spreadsheet so we can continue to debate and argue it as we go through the season. I’ve been reading some statistical tools that we could possibly use but I don’t see anything that’s directly applicable.

Good man!
 

GoldZ

Dodd-Like
Joined
Oct 18, 2002
Messages
6,253
OK I've updated the OP with the scoring rules.

To be successful in the contest you will need to have voted on each game. If you have missed a vote and want to add votes in, just put them in this thread and I'll included them.
Thanks, gimme a .9+ for winning the Alcorn St. game.
 

coit

Y’all got any more of that D Fence?
Joined
Nov 29, 2007
Messages
87,938
Reminder, if you want to give me scores for any games you missed, please let me know by Thursday evening. Here's a list of folks who are missing just one or two games.

@00Burdell - USF
@BigDanT - BG
@BLEEDGOLD - USF
@cbee - Louis
@CHRjacket - Pitt
@GT_LUVER - Clem
@GT1981 - BG
@GT98 - Alcorn
@GTCrew4b - Alcorn
@JDfan - Alcorn
@mookie43 - BG
@Rayburncd5@bellsouth.net - Pitt
@Technician - VT
@vapspwi - Louis
@Akinji07 - Louis UNC
@Athletic Supporter - Pitt VT
@BrentwoodJacket - Pitt VT
@cyclejacket - Alcorn VT
@FatJacket - Pitt BG
@FHstinger - Clem Louis
@GT65_UGA89 - BG UNC
@gtrower - Pitt Clem
@helluvaparamedic - BG UNC
@RamblinWreck09 - VT UNC
@texstinger - Alcorn Clem
@wesleyd21 - Pitt Louis
@WracerX - BG Louis
@wreckless - Duke VT
 
Top