Latest ESPN FPI has Vols favorites in nine games

[BeardedVol]

We were underdogs to A&M and Bama.

Not according to ESPN's 2016 FPI

 

[volfan102455]

The Football Power Index (FPI) is a measure of team strength that is meant to be the best predictor of a team's performance going forward for the rest of the season. FPI represents how many points above or below average a team is. Projected results are based on 10,000 simulations of the rest of the season using FPI, results to date, and the remaining schedule. Ratings and projections update daily.

The FPI and predictions will change with each game played. It is not static - aka like my free throw example where probability of success for next free throw increases or decreases based on result from last free throw. If Tennessee were to be undefeated after the UGA game, you can bet the projection for South Carolina would be higher than 75% - but according to all of you it would be 42% no matter what happens in the first 5 games.

 

[volfan102455]

Not according to ESPN's 2016 FPI

We were underdogs when we played those games. The FPI changes week to week based on results obtained.

I would bet if you looked at the number post Appy State some of those may have went to a predicted loss.

 

[BeardedVol]

The Football Power Index (FPI) is a measure of team strength that is meant to be the best predictor of a team's performance going forward for the rest of the season. FPI represents how many points above or below average a team is. Projected results are based on 10,000 simulations of the rest of the season using FPI, results to date, and the remaining schedule. Ratings and projections update daily.

The FPI and predictions will change with each game played. It is not static - aka like my free throw example where probability of success for next free throw increases or decreases based on result from last free throw. If Tennessee were to be undefeated after the UGA game, you can bet the projection for South Carolina would be higher than 75% - but according to all of you it would be 42% no matter what happens in the first 5 games.

Even if the power index rating for a team changes, and the projected win percentage changes, that still will not change the equation P(A and B) = P(A)*P(B).

Plugging the updated percentages into the equation still doesn't change the equation, just the probability.

If UT had a 50/50 shot of winning every game, and you wanted to know the likelihood that they win the first three games, it's still determined by using the same equation, .5*.5*.5= 12.5%. With a 50/50 probability to win each game, UT would have a 12.5% probability of winning all three.

Basic probability of independent events.

 

[sameolvol]

Well Mr. Engineer, sir, while it appears you have attended a stats 201 class, you didn't read my post. That method skews the numbers by not fully representing the talent differential and the plethora of other variables. So no, there isn't a valid way to calculate that probability given only those numbers, but your probability makes the likelihood of use winning those games look much smaller than it is in reality. Which was my point, not looking to get in to a measuring contest.

:rock:This! Plus, let's replace UT as the primary integer with bammer. Is that going to garner the same expostulation? Nope.
Injuries killed our guys not only last season, but it did same to a lesser extent for team 119 and very much so for team 118. How does that formula adjust to the injury coefficient...or to home versus away...or to matchups... or to (fill in the blanks)? IMHO, team 121 will fare, for the most part, as well as our players can avoid that ugly injury bug. Coach Rock may well be the best addition to a boat load of truly impressive new coaching additions. I say, if our guys can manage to stay notably healthy, then we've got 10 or 11 Ws. If not, well, maybe all the negas will cause Butch's brick building to tumble. GBO!

 

[81Orange]

Typically when you use probability like that it's for x where x is done in a series and expected to yield Y%. Applying that to this doesn't transition well because they are unrelated events. I see why your doing but your method skews the probability a good bit.

Seems correct to me. Games are independent events just like each toss if a coin is independent. Football may be less independent because injuries and mental aspects can carry over. But who knows what direction the skew may be in. Winning and loosing are contagious.

 

[81Orange]

While mathematically you are correct (multiplying probability of each individual events happening and getting probability for the all three events), I would argue these events would need a better predictive tool to guess probability of winning all 3 of GT, GA and USCe and/or Kentucky and Vanderbilt.
Saying that there is a less than 50% chance we beat Kentucky and Vandy both is not a good predicting tool. I would take those odds and bet against it any day.

This is how betters go broke. "Didn't seem right" is correct. What the fan forgets is that the proability for a series changes drastically once each event completes. Suppose you beat KY then calculate the probiouty if beating both. The probability of beating both is then simply the proability of beating Vandy. Looks more like you expect then.

 

[volfan102455]

Absolutely none of what you said changes the fact that the probability of independent events is computed as P(A and B)=P(A)*P(B). Whether it's the probability of flipping a coin and it landing on heads 3 times in a row, or rolling dice and coming up 6 100 times in a row, or the liklehood of making 20 free throws in a row, it's all computed using the same equation.

You can argue that the FPI projections are wrong, but you can't argue the way they are used when determining the probability of winning multiple games based off of the FPI projections.

These are not independent events. ESPN's analysis of how they calculate this is clear on that. As the season progresses the result of a game(s) will influence the results which follow - thus they are conditional.

 

[bigdaddy]

Thoughts?

Georgia Tech (W) 68.5%
Indiana State (W) 98.9%
Florida (L) 37.9%
Massachusetts (W) 97.4%
Georgia (W) 51.3%
South Carolina (W) 75.7%
Alabama (L) 13.5%
Kentucky (W) 60.0%
Southern Mississippi (W) 96.1%
Missouri (W) 64.3%
LSU (L) 40.5%
Vanderbilt (W) 81.1%

I believe going into the season last year we were favorites in every game. I put zero stock in FPI

 

[BeardedVol]

These are not independent events. ESPN's analysis of how they calculate this is clear on that. As the season progresses the result of a game(s) will influence the results which follow - thus they are conditional.

Lol...you really should educate yourself on independent and dependent events in probability before spouting such nonsense.

Dependent events in probability affect the probability of other events and it's expressed in the equation P(A and B) = P(A)*P(B|A). If you have a jar with 12 marbles in it, 4 orange, 4 white, and 4 grey, the probability of you drawing an orange white and grey marble in order is 4/12*4/11*4/10= 64/1320=8/165=4.8% chance of drawing orange, white, and grey marbles out in succession. Dependent event, as the odds are modified with each successive drawing of a marble.

FPI projections may change for individual games as they are recomputed throughout the season, but the sheer act of playing games does not increase or decrease the chances of winning future games, as the games themselves are independent events. The same equation is used, regardless of the numbers that you plug into it; P (A and B)= P(A)*P(B)

Thanks for making me feel like I'm back at my student work-study in the math lab.:good!:

 

[82_VOL_83]

The probability of both events occurring is the product of the 2 probabilities.

Just think about it like shooting free throws. You could be a 75% shooter from the free throw line, which is good. however, if you shoot 3 free throws, you are likely to miss 1 free throw, even though you are heavily favored in each individual shot.
0.75*0.75*0.75 = 0.42, or 42% to make 3 free throws in a row.

The 75℅ FT shooter example assumes a finite set of practiced skills and the same conditions each time. There are variables but by definition, the basket is always in the same place. Applying that same statistics law to two football games doesn't work nearly as well. The same as giving odds at this time of year for the games.

 

[BeardedVol]

The 75℅ FT shooter example assumes a finite set of practiced skills and the same conditions each time. There are variables but by definition, the basket is always in the same place. Applying that same statistics law to two football games doesn't work nearly as well. The same as giving odds at this time of year for the games.

It works fine. It doesn't matter if you use Phil Steele's predictions, ESPN FPI, Oddshark's, any of them will work, and you still use the same equation to determine the probability of multiple independent events occurring, even if the numbers are different because you are using Phil Steele's numbers instead of FPI.

 

[BigOrangeTrain]

How can anyone really expect us to be favorites over South Carolina and Vandy?

 

[jakez4ut]

they re calculate the FPI DURING the games, as they're played.

what i see is 3 losses and 2 toss ups. which is about how anyone would/should look at it in April. making 8-4/7-5 about where it's predicted.

 

[BeardedVol]

How can anyone really expect us to be favorites over South Carolina and Vandy?

On paper we are. Last year, preseason FPI had us favored in every game, even Alabama.

 

[Iam4utalways]

they re calculate the FPI DURING the games, as they're played.

what i see is 3 losses and 2 toss ups. which is about how anyone would/should look at it in April. making 8-4/7-5 about where it's predicted.

So you're counting SC and Vandy as gimmies? Me too.

 

[jakez4ut]

So you're counting SC and Vandy as gimmies? Me too.

gimmies? ok. your words. let's just say, despite last year's results, i don't go in to this year expecting history to repeat itself.

GA, Bama and LSU are games that i think you can legitimately make a case as losses on the schedule.

i think GT and FL are games you can look at and say we can win those games based on talent levels between the two teams.

everybody else we play we "should" win.

and i wouldn't be shocked if we, say, beat LSU, but lost to FL. or beat GA or lost to GT.

there's a path to 9-3 or better. but it's going to depend on some things happening, some questions getting positive answers, and staying away from the devastating injuries.

in the end, i still think this team will flash at times, and flounder at times. and we'll wind up in that 8-4 ball park.

 

[HoptownVol]

I may have to take a drive down to Gainesville this fall.

 

[volfan102455]

Lol...you really should educate yourself on independent and dependent events in probability before spouting such nonsense.

Dependent events in probability affect the probability of other events and it's expressed in the equation P(A and B) = P(A)*P(B|A). If you have a jar with 12 marbles in it, 4 orange, 4 white, and 4 grey, the probability of you drawing an orange white and grey marble in order is 4/12*4/11*4/10= 64/1320=8/165=4.8% chance of drawing orange, white, and grey marbles out in succession. Dependent event, as the odds are modified with each successive drawing of a marble.

FPI projections may change for individual games as they are recomputed throughout the season, but the sheer act of playing games does not increase or decrease the chances of winning future games, as the games themselves are independent events. The same equation is used, regardless of the numbers that you plug into it; P (A and B)= P(A)*P(B)

Thanks for making me feel like I'm back at my student work-study in the math lab.:good!:

I said conditional - conditional is when the outcome of A impacts B based. ESPN makes it clear that the FPI is recalculated weekly and that the results from the prior week impact the FPI as the season progresses. So they are in effect saying that the outcome of games played prior to a game impact the projected result of that game.

I suspect when they run the process that creates the percentages they are using a algorithm that takes into account what the other teams are projected to do week by week - thus giving them that first basis to compare.

When the South Carolina game is played, if you are using the FPI as a predictor the percentage the day before the game will be the chance that Tennessee wins the game. It won't be the percentage you get right now by multiplying the others together.

You are trying to use a simple example to prove your point but the FPI calculation is not a simple roll the dice, make a free throw statistic.

 

[volfan102455]

they re calculate the FPI DURING the games, as they're played.

what i see is 3 losses and 2 toss ups. which is about how anyone would/should look at it in April. making 8-4/7-5 about where it's predicted.

They do - I remember before the Hail Mary the chance of Tennessee winning did not exist and was as close to 0% as you could get.

 

[temptn]

The probability of both events occurring is the product of the 2 probabilities.

Just think about it like shooting free throws. You could be a 75% shooter from the free throw line, which is good. however, if you shoot 3 free throws, you are likely to miss 1 free throw, even though you are heavily favored in each individual shot.
0.75*0.75*0.75 = 0.42, or 42% to make 3 free throws in a row.

I think your reasoning is flawed. They are two independent events. Not like shooting free throws, more like flipping a coin and rolling dice and thinking that rolling the dice influences whether you get a heads or tails.

 

[jakez4ut]

I think your reasoning is flawed. They are two independent events. Not like shooting free throws, more like flipping a coin and rolling dice and thinking that rolling the dice influences whether you get a heads or tails.

the difference is the series of events.

we're favored in 9 individual games according to this index.

what's the probability that we win all 9?

or in your example, instead of the outcomes affecting the next event, rather, it's can they both occur, and what's the probability of that happening?

50% on flipping heads
20% on rolling snake eyes

what's the probability of doing both?

combined as a series of events, it's less probable than each individual outcome's probability.

 

[BeardedVol]

I said conditional - conditional is when the outcome of A impacts B based. ESPN makes it clear that the FPI is recalculated weekly and that the results from the prior week impact the FPI as the season progresses. So they are in effect saying that the outcome of games played prior to a game impact the projected result of that game.

I suspect when they run the process that creates the percentages they are using a algorithm that takes into account what the other teams are projected to do week by week - thus giving them that first basis to compare.

When the South Carolina game is played, if you are using the FPI as a predictor the percentage the day before the game will be the chance that Tennessee wins the game. It won't be the percentage you get right now by multiplying the others together.

You are trying to use a simple example to prove your point but the FPI calculation is not a simple roll the dice, make a free throw statistic.

Again, it does not matter if you use FPI, or Phil Steele, or Oddshark, or any other set of predictions. It doesn't matter if FPI is a rolling prediction that changes throughout the season. Computing the probability of independent events occurring together, which in terms of probability football games are whether you choose to acknowledge it or not, is still determined using the equationP(A and B)= P(A)*P(B).

Feel free to ask one of the professors in the math program you claim to be in, or just google it.:zeitung_lesen:

 

[VFL-82-JP]

I said conditional - conditional is when the outcome of A impacts B based. ESPN makes it clear that the FPI is recalculated weekly and that the results from the prior week impact the FPI as the season progresses. So they are in effect saying that the outcome of games played prior to a game impact the projected result of that game.

...

You are trying to use a simple example to prove your point but the FPI calculation is not a simple roll the dice, make a free throw statistic.

Dude, you're totally missing the fact that any prediction of future events is based on known and assumed conditions at the time of the prediction. As conditions change, the prediction can change with them.

At any snap shot in time, whether 5 minutes before the game, the day before, a week before, or six months before the season starts, you can make or update a prediction.

Not only about one event (one game), but about a series of events (games).

And just because the known conditions and assumptions today aren't going to be the same as those a month from now, that doesn't mean your current predictions aren't as good as they can currently be.

Bottom line is, Bearded is laying out the facts of calculating probabilities of a series of events for you really well; you should believe him.

 

[Hairy Vols]

The probability of both events occurring is the product of the 2 probabilities.

Just think about it like shooting free throws. You could be a 75% shooter from the free throw line, which is good. however, if you shoot 3 free throws, you are likely to miss 1 free throw, even though you are heavily favored in each individual shot.
0.75*0.75*0.75 = 0.42, or 42% to make 3 free throws in a row.

If we look at it like that tho. What is Vandy's probability to beat both us and Kentucky? And what is Kentucky's probably to beat both us and Vanderbilt?

You can literally play this both ways.