Updated Odds of Winning Out: 68.1%

[sjt18]

This deserves a separate discussion.

Could you argue that 90% was an inaccurate projection? Sure.

This is very important as well. Where did those "probabilities" come from? I personally believe UT is far more likely to beat MU than 80%.

Can you argue that the laws of mathematics are suspended and magically conform to a rule you completely made up based on ignorance of math? You can, but you'll be wrong.

I am not arguing that the "laws" of math should be suspended. I am arguing that you are incorrectly applying methods that apply to random events to largely non-random events. You can be as snooty as you like... but that's what you're doing.

 

[Special Ed]

If our defense doesn't stop "targeting", we may lose all four because of player suspensions.

 

[sjt18]

you obviously do not understand

No. I understand what he's trying to do. I simply disagree. Football is NOT a sequence of purely random events predictable by the statistics you all are trying to use... without even saying where the probabilities come from to start with.

Interesting that some arguing your side of things think I am negative.

I believe that UT and Jones will win the last 4. Last week when some of you were fretting over UK... I said this: "I don't think UK will score more points than Bama did or than UGA did on O" in response to someone predicting that UT would win 41-20.

You people are trying WAY too hard to unrealistically lowball UT's likelihood of winning out.

UT is a better team than any of the remaining opponents and better coached than any except Mizzou.

 

[DiderotsGhost]

Sjt, I enjoy most of your posts. I'm not trying to be "snooty." I am not an expert on probability, but I've taken about 6 advanced courses on data analysis, probability, and statistics, so I'm pretty familiar with how it works. From your statements, it's obvious to anyone who is familiar with the field that you do not understand what you are critiquing.

Let me put this in really basic terms. PREDICTING SPORTS OUTCOMES IS HARD! There are several ways to do it. One of the most accurate ways to do it is to build models that incorporate a wide variety of data analysis. ESPN's statisticians have built a data model that analyzes thousands of individual plays, statistics, outcomes, etc, from college football games. Based on that data, they built a model that projects the likelihood of one team winning versus another.

You have proposed an alternative model that says "the odds are X because I say so."

Is it possible that ESPN is underestimating the odds of UT winning these individual games? Yes. Is it possible that ESPN is overestimating the odds of UT winning individual games? Also yes. However, while ESPN's model may potentially have flaws, your model, which is nothing more than repeating "the odds are higher because I say so!" is likely much more flawed. Maybe the real odds of us winning out are 74%. Maybe they are 61%. I don't know and you don't know, but I trust ESPN's model which is based on a ton of data analysis, sound mathematical reasoning, and has been refined over time way more than I trust your "it's so because I say it's so" model.

While you may believe that probability can't be applied to anything other than flipping coins and cards, as VFL has already pointed out, it's used in virtually every field to help predict outcomes. One good example is that you could use it to predict the frequency of traffic accidents. Once again, this is not like flipping a coin, and it's not "random" in the sense that picking a card might be. However, if you analyze traffic accidents on certain roadways over the course of 1,000 days, you probably have a good general sense of how frequently they occur.

Let's say a traffic model says you will make it to work "on time" under 25 minutes 83 days out of 100. You drive to work 5 days and make it to work "on time" every single day. Based on this, you propose that you will make it to work "on time" 100 days out of 100. Over the course of the next 100 days, you are "on time" 85 days. The data model slightly underestimated the real outcome, but was fairly close. Your "alternative model", which was based on poor logic and an extremely limited data sample, greatly overestimated the number of days you arrived "on time."

This basically sums up your argument.

ESPN may be either underestimating or overestimated the Vols' chances of winning out. However, I'd be willing to bet my life savings that over the course of 1,000 games, that ESPN's projection model is much more accurate than your "it's so because I say it's so" model.

 

[tvolsfan]

Again, every team left on the schedule (and go ahead and throw in Kentucky) is horrible and none beat Tennessee unless Dobbs and Hurd both get injured. Tennessee is going 8-4.

 

[Boca Vol]

Sjt, not to rag on you, but I don't think you understand how probability models work. I play poker tournaments occasionally and this stuff is pretty common in poker.

Let's say you go all-in on 5 different hands and get called, each time with an 80% chance of winning. What that means is that the moment you lose, you're out of the tournament. Here are you odds of advancing to the 5th hand and winning.

Win 1st hand: 80%
Win 1st and 2nd hand: 64% (.8 X .8 = .64)
Win 1st, 2nd, and 3rd hand: 51.2% (.8 X .8 X .8 = .512)
Win 1st, 2nd, 3rd, and 4th hand: 41.0%
Win all 5 hands: 32.8%

Even though you are favored to win every hand, the odds of actually winning all 5 hands is relatively low at 32.8%.

DG,

Thanks for using poker as a comparison. I, too, enjoy NL poker but I've never understood the odds because the odds assume all "outs" are available. Using your "all-in" scenario from above do the odds change if you're holding pocket AA or KK in each hand? Just curious.

 

[jpope9]

I don't want to overestimate Missou, either. They are very bad. We need to not let them even feel they belong on the same field on 11/21.

they've beat us every year since they have been in the SEC, I highly doubt the team will overlook them. Our defense will not let them score a TD. Butch wont let that happen with a potential 8-4 record on the line.

 

[DiderotsGhost]

DG,

Thanks for using poker as a comparison. I, too, enjoy NL poker but I've never understood the odds because the odds assume all "outs" are available. Using your "all-in" scenario from above do the odds change if you're holding pocket AA or KK in each hand? Just curious.

Boca, there are actually a few good odds calculators online.

CardPlayer Odds Calculator

You'd have to have a reasonable idea of your opponents' hands to accurately calculate odds. Preflop odds are more difficult and speculative. Pocket Aces is the only hand where you always know you're the favorite preflop and you are normally about an 80-20 favorite with that hand. With any other hand, it's very dependent upon your opponents holdings. Even with KK, your odds would be very different if your opponent holds AQs as opposed to 64o.

There's a simple rule of thumb for calculating odds post-flop. If you have a good estimate of how many outs you have, then on the flop, multiple that by 4. On the turn, you would multiple it by 2. For instance, a flush draw, on the flop would have 9 outs. On the flop, that's roughly a 36% chance of hitting the flush, and on the turn, there's roughly an 18% chance. You don't really have to be concerned with the cards you don't know. It's possible that you have a diamond flush draw and 7 other diamonds were folded, leaving only 2 in the deck. It's also possible that 0 diamonds were folded, thereby, effectively increasing your odds. However, you have access to none of that information, so you don't really consider it.

Poker tournaments can be very frustrating because even if you play "perfectly", you're still going to lose most of the time. Eventually, you end up in a pot where you're an 80% favorite, and your opponent ends up hitting the exact cards they need. The more confrontations like that you have, the more likely it becomes that you lose.

 

[99gator]

Im not a math major.

Would this mean that vegas would offer 1 to 3 odds for tennessee to win out.

You put up 3 to win 1

 

[Boca Vol]

Boca, there are actually a few good odds calculators online.

CardPlayer Odds Calculator

You'd have to have a reasonable idea of your opponents' hands to accurately calculate odds. Preflop odds are more difficult and speculative. Pocket Aces is the only hand where you always know you're the favorite preflop and you are normally about an 80-20 favorite with that hand. With any other hand, it's very dependent upon your opponents holdings. Even with KK, your odds would be very different if your opponent holds AQs as opposed to 64o.

There's a simple rule of thumb for calculating odds post-flop. If you have a good estimate of how many outs you have, then on the flop, multiple that by 4. On the turn, you would multiple it by 2. For instance, a flush draw, on the flop would have 9 outs. On the flop, that's roughly a 36% chance of hitting the flush, and on the turn, there's roughly an 18% chance. You don't really have to be concerned with the cards you don't know. It's possible that you have a diamond flush draw and 7 other diamonds were folded, leaving only 2 in the deck. It's also possible that 0 diamonds were folded, thereby, effectively increasing your odds. However, you have access to none of that information, so you don't really consider it.

Poker tournaments can be very frustrating because even if you play "perfectly", you're still going to lose most of the time. Eventually, you end up in a pot where you're an 80% favorite, and your opponent ends up hitting the exact cards they need. The more confrontations like that you have, the more likely it becomes that you lose.

Thanks DG! :good!:

I play in a bar league in FL and it's a lot of fun. I've played in a $1, $2 cash game at a casino 1 time and realized that's not for me. :crazy:

Tournament play is more my speed. I wish I had the knowledge of "odds" like you perfectly explained. I just don't. I tend to play my hand vs implying a hand. I never get away with bluffing.

 

[sjt18]

Sjt, I enjoy most of your posts. I'm not trying to be "snooty." I am not an expert on probability, but I've taken about 6 advanced courses on data analysis, probability, and statistics, so I'm pretty familiar with how it works. From your statements, it's obvious to anyone who is familiar with the field that you do not understand what you are critiquing.

I understand just fine.

"Vegas" would use a number like this to induce betting. But they would never JUST use UT. If you take 128 teams or at least a significant number then not only do you introduce more meaningful "randomness"... you balance the equation. As many teams will fail as succeed so the ultimate objective is achieved- the house wins.

This is MUCH like recruiting rankings. They are "generally" but not specifically accurate. Most 5* recruits are successful. A relatively high % of them are drafted. But you can NEVER say that one player or another will succeed based on being rated 5*. So while 60% or whatever of 5* players will eventually be drafted.... you cannot say that Chris Donald had a 60% chance of being drafted.

Let me put this in really basic terms. PREDICTING SPORTS OUTCOMES IS HARD! There are several ways to do it. One of the most accurate ways to do it is to build models that incorporate a wide variety of data analysis. ESPN's statisticians have built a data model that analyzes thousands of individual plays, statistics, outcomes, etc, from college football games. Based on that data, they built a model that projects the likelihood of one team winning versus another.

You have proposed an alternative model that says "the odds are X because I say so."

No, I am not. And you aren't telling me anything I do not know.

You are still determined to apply a method that applies to things that are random to something that in the "local" and limited circumstance... aren't random.

Is it possible that ESPN is underestimating the odds of UT winning these individual games? Yes. Is it possible that ESPN is overestimating the odds of UT winning individual games? Also yes. However, while ESPN's model may potentially have flaws, your model, which is nothing more than repeating "the odds are higher because I say so!" is likely much more flawed.

If I had done that then you still wouldn't have a point. I have said that the limit of UT's "odds" is set by the least likely win.

Maybe the real odds of us winning out are 74%. Maybe they are 61%. I don't know and you don't know, but I trust ESPN's model which is based on a ton of data analysis, sound mathematical reasoning, and has been refined over time way more than I trust your "it's so because I say it's so" model.

I am still waiting on a reference that establishes that supports the 80% win possibility against MU.

While you may believe that probability can't be applied to anything other than flipping coins and cards, as VFL has already pointed out, it's used in virtually every field to help predict outcomes. One good example is that you could use it to predict the frequency of traffic accidents. Once again, this is not like flipping a coin, and it's not "random" in the sense that picking a card might be. However, if you analyze traffic accidents on certain roadways over the course of 1,000 days, you probably have a good general sense of how frequently they occur.

You CAN predict the frequency of traffic accidents. You CANNOT use that method to predict the likelihood of any individual driver of having an accident.

Let's say a traffic model says you will make it to work "on time" under 25 minutes 83 days out of 100. You drive to work 5 days and make it to work "on time" every single day. Based on this, you propose that you will make it to work "on time" 100 days out of 100. Over the course of the next 100 days, you are "on time" 85 days. The data model slightly underestimated the real outcome, but was fairly close. Your "alternative model", which was based on poor logic and an extremely limited data sample, greatly overestimated the number of days you arrived "on time."

No. You either do not understand my argument or are engaging in sophistry.

For one, I am in control of when I leave. I am in control of the route I take. I am in control of the method of travel. I have some measure of control of how far I must travel. I am in control of my speed at least at times. The time of day I travel is variable and controllable. This is only a few of many CONTROLLED choices that apply.

ONLY if you set up the parameters that remove my influence does the model even come close to being accurate.

That is the SAME THING with football. In every week, there are 100's if not 1000's of decisions made that are not random but that DO heavily influence the outcome for particular teams.

You are STILL trying to use math that applies to random events to a very specific situation that is NOT random.

 

[FLVOL_79]

68% chance we win out every time.

 

[GreyhoundMan]

You guys obviously know a lot more math than I do. All I know is that I was sitting in the stands (the closed part of the horseshoe as we called it back then) when North Texas State came to town and left with a win many moons ago.
Funny things can happen in football games and I hope our guys are focused and ready every week - I feel confident they will be. if the Vols come out ready to play their best, you won't need a calculator to figure the outcome. (But seeing the probability stats is interesting, gotta say.)

 

[dieseldaddy]

cain't figure,% against all 4 teams are above 80%,yet to win out only 68%.

Math sucks

 

[thebjd]

I understand just fine.

"Vegas" would use a number like this to induce betting. But they would never JUST use UT. If you take 128 teams or at least a significant number then not only do you introduce more meaningful "randomness"... you balance the equation. As many teams will fail as succeed so the ultimate objective is achieved- the house wins.

This is MUCH like recruiting rankings. They are "generally" but not specifically accurate. Most 5* recruits are successful. A relatively high % of them are drafted. But you can NEVER say that one player or another will succeed based on being rated 5*. So while 60% or whatever of 5* players will eventually be drafted.... you cannot say that Chris Donald had a 60% chance of being drafted.

No, I am not. And you aren't telling me anything I do not know.

You are still determined to apply a method that applies to things that are random to something that in the "local" and limited circumstance... aren't random.

If I had done that then you still wouldn't have a point. I have said that the limit of UT's "odds" is set by the least likely win.

I am still waiting on a reference that establishes that supports the 80% win possibility against MU.

You CAN predict the frequency of traffic accidents. You CANNOT use that method to predict the likelihood of any individual driver of having an accident.



No. You either do not understand my argument or are engaging in sophistry.

For one, I am in control of when I leave. I am in control of the route I take. I am in control of the method of travel. I have some measure of control of how far I must travel. I am in control of my speed at least at times. The time of day I travel is variable and controllable. This is only a few of many CONTROLLED choices that apply.

ONLY if you set up the parameters that remove my influence does the model even come close to being accurate.

That is the SAME THING with football. In every week, there are 100's if not 1000's of decisions made that are not random but that DO heavily influence the outcome for particular teams.

You are STILL trying to use math that applies to random events to a very specific situation that is NOT random.

It is a predictive model. It is not supposed to be 100% precise nor is it intended to count how motivated a team is or how many push-ups they do in the week leading up to the game. As the games fall off the schedule the odds of winning out will change. I assume the individual game probabilities can still change as well depending on the variables ESPN uses. I understand your point and your point will matter each week when we play the games. It does not matter when trying to interpret data or creating a predictive model.

 

[daniel3417]

I get some of what SJT is saying. A team like Ohio State might have 90% chance of winning each individual game. Thus, their chances of going undefeated at the season's beginning is 28%. I certainly understand the math, but if Vegas offered 4/1 payouts on Ohio State going undefeated this year, they'd go bankrupt.

The probability of winning and winning out changes constantly within a game.

 

[volleyvol]

Here are our updated odds of winning out per ESPN's statistical projections.

vs South Carolina: 90.7%
vs North Texas: 99.6%
@ Missouri: 80.4%
vs Vanderbilt: 93.7%

Odds of Winning all 4 Games: 68.1%

Don't want to get too excited, because we all know that one of the three SEC games remaining will probably be closer than we expect. But certainly, if the early part of our schedule ended up being tougher than expected (with Oklahoma being better than last year and Florida being resurgent), then the latter half looks to be easier than expected preseason.

I also like that Mizzou is after that North Texas game. North Texas is atrociously bad this year. Sagarin has them rated below our I-AA opponent Western Carolina. We might be able to give some of our starters some rest vs North Texas, helping keep us in better shape for Mizzou and their tough D.

Allow me to be a wise guy for a minute!

Your computation assumes the outcomes of the games are independent. I would argue that these events are dependent, i.e., the odds of winning the 2nd game will increase if we win the 1st, and the odds of winning the 3rd will increase if we win the first two, etc. Unfortunately, there's no way a priori to compute these conditional probabilities, so the best we can say is that the odds of winning out are at least 68.1%.

Great post, regardless! :thumbsup:

Go Vols!

 

[sjt18]

It is a predictive model. It is not supposed to be 100% precise nor is it intended to count how motivated a team is or how many push-ups they do in the week leading up to the game. As the games fall off the schedule the odds of winning out will change. I assume the individual game probabilities can still change as well depending on the variables ESPN uses. I understand your point and your point will matter each week when we play the games. It does not matter when trying to interpret data or creating a predictive model.

And THAT is why it is not useful in the context it has been presented here... unless a bunch of folks are planning on betting for/against UT winning out. It is only accurate when you total the population... not when you look at a very small subset.

 

[maotai]

This pretty much sums up this thread.

 

[Volosaurus rex]

If you want some additional data to extrapolate Tennessee's probability of success down the stretch, here are a few nuggets to chew on that aren't the product of ESPN's analysts:

"Tennessee is 314-125-21 all-time in November, a .705 winning percentage. The Vols are 82-20 (.804) in November since 1990. In the 90's, UT was 37-3 (.925) in November, including 21 consecutive wins from 1990-96. . . . Tennessee is 14-3 when playing on Nov. 7, which is the date the calendar will ready when the Gamecocks come to visit on Saturday. UT has won 13 straight games on the date, the last loss coming to Vanderbilt in 1908.

The Volunteers are 15-2 against the Gamecocks in Knoxville and enter this meeting on two-game winning streak over USC after a pair of dramatic wins each of the last two seasons. UT drove down the field for a game-winning field goal as time expired to defeat the 11th-ranked Gamecocks 23-21 in 2013" (November Begins With South Carolina - University of Tennessee Official Athletic Site).

 

[Remy]

Prob win each of the games given as:
vs South Carolina: 90.7%
vs North Texas: 99.6%
@ Missouri: 80.4%
vs Vanderbilt: 93.7%

Prob win ALL 4 is .907 * .996 * .804 * .937 = .681 :hi:

Lol, the lowest is .804 and the average somehow is .681?!?!, turn in your 3rd grade math award. The mean should be somewhere between the lowest/.804 and highest/.996.

 

[OrngeChkrboardVFL]

I don't want to overestimate Missou, either. They are very bad. We need to not let them even feel they belong on the same field on 11/21.

Remember their fans coming on VN last season and gloating about how they backed into the SEC championship game. How not only did they deserve to be in the SEC but they are pretty much dominating. I hope we hang 40 on them.

 

[thebjd]

Lol, the lowest is .804 and the average somehow is .681?!?!, turn in your 3rd grade math award. The mean should be somewhere between the lowest/.804 and highest/.996.

 

[ptcarter]

I believe this is a misapplication of statistics.

If they were going to make this legit then they would have to go back to the original "probabilities" which they had absolutely no accurate means of determining and then predict the number of games UT would win. Most likely... it would have been 8 or 9.

I believe the best application of these probabilities which are suspect in and of themselves is to say the probability of winning out is equal to the lowest win probability... which would be 80.4%.

It is difficult though to determine how they derived at that number for MU when they've already been beaten by UK and VU... and UT's comparative strength to UK's is pretty fresh in our minds, right?

I read this and thought of the logic that made this woman a witch.

https://youtu.be/zrzMhU_4m-g

 

[DiderotsGhost]

Lol, the lowest is .804 and the average somehow is .681?!?!, turn in your 3rd grade math award. The mean should be somewhere between the lowest/.804 and highest/.996.