Spring Ball 2024…

[Josh]

Would average distance each year from the yearly actual do the trick? (Closest to 0 distance per year being the most accurate?)

For example: 15-21 in the last 3 seasons.

Profcat predictions: 12-24
Josh predictions: 18-18
We're both 3 off,

However:
Profcat (0+0+3)/3=1
Josh (2+2+1)/3=1.67

Demonstrates that Profcat has been quite a bit more accurate over the 3 years.

 

[2centsworth]

Would average distance each year from the yearly actual do the trick? (Closest to 0 distance per year being the most accurate?)

For example: 15-21 in the last 3 seasons.

Profcat predictions: 12-24
Josh predictions: 18-18
We're both 3 off,

However:
Profcat (0+0+3)/3=1
Josh (2+2+1)/3=1.67

Demonstrates that Profcat has been quite a bit more accurate over the 3 years.

Generally, I dislike averages to measure anything.

However, if we used averages, I believe the data would look like:

ProfCat: (0+0-3)/3 = 1
Josh: (2+2-1)/3 = 1

Otherwise, predicting the correct record of 15-21 (5,5,5) would be considered less accurate than ProfCat's.

If we use standard deviation, you and ProfCat were equally accurate (4,4,4 = 6,6,6). In fact, I would argue someone who guessed 4,4,6 wins would be considered more accurate than someone who guessed 5,5,5.

A math whiz can definately come up with a better formula.

 

[Bobcat0612]

Math wizards, please correct me if I'm wrong, but we would use standard deviation to help us gauge accuracy, correct? For the sake of example, if someone guessed 1, 2, 12 for a total of 15 wins in 3 years, they would not be considered more accurate than ProfCat even though they guessed the correct total number of wins.

The standard deviation of the actual win totals is 1.73 over 3 years, so the closer the standard deviation of our guesses is to 1.73, the more accurate our predictions would be considered, correct?

This is my SWT education talking, so please correct me if I'm wrong.

No, but close. As an example, if our wins were 5, 8, 10 in that order, then guesses of (5, 8, 10) and (10, 5, 8) would have the same st. dev. but one is clearly more accurate than the other. You would probably want to do a paired analysis and look at the average difference and the standard deviation of the differences. But I just woke and would have to think about it a little more.

 

[Bobcat0612]

Lets see here, I am not going to worry about any assumptions we have to make about the data but I think this would be what we would use:

Paired T-Test

Paired sample t-test is a statistical technique that is used to compare two population means in the case of two samples that are correlated.

www.statisticssolutions.com

.

The Null Hypothesis would be that there is no difference between a guess and the actual record. The Alternative would be that they are different, so a two-tailed test. Since a guess could be larger or smaller we would not want to do a one-tailed test. Then, when comparing two guess sequences, whichever guess sequence has the smaller p-value is the better guess.

On the surface it seems to be a little strange to do that but thats because I don’t normally do this with this type of data. So I feel like there might be better way but off the top of my head it looks good. The other thing that makes it feels weird is the small sample size. Normalcy might be problem too but like I said I was not going to worry about assumptions.

 

[Simbo0722]

The idea was likely to not show the fact that we have zero depth at most positions, hence can’t afford injuries, hence can’t afford to play an actual game. This team is being built to win 10 games in a season as fast as possible and then bolt, while at the same time hedging that bet in case we end up in a prolonged 6-8 win “rut”. Thats why we have a lot of flashy cards at the top of the depth , but nothing below that. Other than RB, we are thin everywhere else at the moment.

I will die on the hill that Spav actually recruited well, and his last season’s team should have won 7-10 games with Rogers or Evans at QB. We actually had legit FBS level two deep that last season. Right now, we don’t. We are 100% relying on McCloud and Mahdi to outscore all our future opponents.

I think you are selling th WR room short. Tons of talent there. The Edge rushers are better this year. QB room tons better. #1's advantage McCloud oner TJ. #2 Hatter over Hornsby. #3 Jackson over last years Hatter or Rogers. Don't forget the portal to add to this years room.

 

[LTK5H]

QB room tons better. #1's advantage McCloud oner TJ. #2 Hatter over Hornsby. #3 Jackson over last years Hatter or Rogers.

With all the kool aid you're chugging, your name has to be Jim Jones.

 

[2centsworth]

Lets see here, I am not going to worry about any assumptions we have to make about the data but I think this would be what we would use:

Paired T-Test

Paired sample t-test is a statistical technique that is used to compare two population means in the case of two samples that are correlated.

www.statisticssolutions.com

.

The Null Hypothesis would be that there is no difference between a guess and the actual record. The Alternative would be that they are different, so a two-tailed test. Since a guess could be larger or smaller we would not want to do a one-tailed test. Then, when comparing two guess sequences, whichever guess sequence has the smaller p-value is the better guess.

On the surface it seems to be a little strange to do that but thats because I don’t normally do this with this type of data. So I feel like there might be better way but off the top of my head it looks good. The other thing that makes it feels weird is the small sample size. Normalcy might be problem too but like I said I was not going to worry about assumptions.

Would a simple formula like this work in this case? 5,5,5 would be considered more accurate than 1,2,12, and 4,4,6 would be more accurate than 5,5,5.

|Wa – Wp| + (Sa - Sp) = A

Wa = Acutual Wins
Wp =Predicted Wins

Sa = Std dev. of wins
Sp= Std dev. of predicted wins

Here's a simple std dev. calculator

 

[2centsworth]

I asked AI for help and this is what I got:

My Question to AI:
As a statistician, how would I determine which prognosticator is more accurate than the others? For example, there are three prognosticators trying to predict how many wins a football team will win each year over the next three years. Three years later, we see that the team won 4 games in year one, 4 games in year two, and 7 games in year three for a total of 15 games. Prognosticator #1 guessed 1 game in year one, 2 games in year two, and 12 games in year three for a total of 15 games. Prognosticator #2 guessed 5 games won in each year for a total of 15 games. Prognosticator #3 guessed 4 games, 3 games, and 8 games for a total of 15 games. Who was more accurate?


AI's Answer:
puts on statistician hat

To determine which prognosticator is more accurate in predicting the number of wins for this football team over 3 years, we need to look beyond just the total predicted wins. The distribution and variability of their predictions across the 3 years matter.
One metric I would use is the standard deviation of the prediction errors. This measures how much, on average, the predictions deviated from the actual outcomes each year.

Let's calculate the prediction errors for each prognosticator per year:

Year 1 (Actual: 4 wins)
Prognosticator 1 Error: 4 - 1 = 3
Prognosticator 2 Error: 4 - 5 = -1
Prognosticator 3 Error: 4 - 4 = 0

Year 2 (Actual: 4 wins)
Prognosticator 1 Error: 4 - 2 = 2
Prognosticator 2 Error: 4 - 5 = -1
Prognosticator 3 Error: 4 - 3 = 1

Year 3 (Actual: 7 wins)
Prognosticator 1 Error: 7 - 12 = -5
Prognosticator 2 Error: 7 - 5 = 2
Prognosticator 3 Error: 7 - 8 = -1

Now to find the standard deviation of these errors for each:
Prognosticator 1 Std Dev = sqrt((3^2 + 2^2 + (-5)^2)/3) = 3.51
Prognosticator 2 Std Dev = sqrt((-1)^2 + (-1)^2 + 2^2)/3) = 1.63
Prognosticator 3 Std Dev = sqrt(0^2 + 1^2 + (-1)^2)/3) = 0.82

The lower the standard deviation, the more accurate and consistent the predictions were overall.
So based on the standard deviation of prediction errors, Prognosticator 3 with a std dev of 0.82 was the most accurate, followed by Prognosticator 2 (1.63), then Prognosticator 1 (3.51). Even though they all predicted the same total wins, the distribution of Prognosticator 3's yearly predictions more closely matched the actual outcomes.

 

[Bobcat0612]

I asked AI for help and this is what I got:

My Question to AI:
As a statistician, how would I determine which prognosticator is more accurate than the others? For example, there are three prognosticators trying to predict how many wins a football team will win each year over the next three years. Three years later, we see that the team won 4 games in year one, 4 games in year two, and 7 games in year three for a total of 15 games. Prognosticator #1 guessed 1 game in year one, 2 games in year two, and 12 games in year three for a total of 15 games. Prognosticator #2 guessed 5 games won in each year for a total of 15 games. Prognosticator #3 guessed 4 games, 3 games, and 8 games for a total of 15 games. Who was more accurate?


AI's Answer:
puts on statistician hat

To determine which prognosticator is more accurate in predicting the number of wins for this football team over 3 years, we need to look beyond just the total predicted wins. The distribution and variability of their predictions across the 3 years matter.
One metric I would use is the standard deviation of the prediction errors. This measures how much, on average, the predictions deviated from the actual outcomes each year.

Let's calculate the prediction errors for each prognosticator per year:

Year 1 (Actual: 4 wins)
Prognosticator 1 Error: 4 - 1 = 3
Prognosticator 2 Error: 4 - 5 = -1
Prognosticator 3 Error: 4 - 4 = 0

Year 2 (Actual: 4 wins)
Prognosticator 1 Error: 4 - 2 = 2
Prognosticator 2 Error: 4 - 5 = -1
Prognosticator 3 Error: 4 - 3 = 1

Year 3 (Actual: 7 wins)
Prognosticator 1 Error: 7 - 12 = -5
Prognosticator 2 Error: 7 - 5 = 2
Prognosticator 3 Error: 7 - 8 = -1

Now to find the standard deviation of these errors for each:
Prognosticator 1 Std Dev = sqrt((3^2 + 2^2 + (-5)^2)/3) = 3.51
Prognosticator 2 Std Dev = sqrt((-1)^2 + (-1)^2 + 2^2)/3) = 1.63
Prognosticator 3 Std Dev = sqrt(0^2 + 1^2 + (-1)^2)/3) = 0.82

The lower the standard deviation, the more accurate and consistent the predictions were overall.
So based on the standard deviation of prediction errors, Prognosticator 3 with a std dev of 0.82 was the most accurate, followed by Prognosticator 2 (1.63), then Prognosticator 1 (3.51). Even though they all predicted the same total wins, the distribution of Prognosticator 3's yearly predictions more closely matched the actual outcomes.

I like that. It makes more sense and is easier to calculate. I have been teaching t-tests the last week so that was on my brain.

 

[Goldievirtuoso]

3-year total was 15-21. I predicted 17-19. Anyone do better?

My predictions came out to 16-20. I was off by one each year (+1 the last two seasons for Spav and -1 for Kinne's first season). Not too bad.

 

[2centsworth]

My predictions came out to 16-20. I was off by one each year (+1 the last two seasons for Spav and -1 for Kinne's first season). Not too bad.

Using my formula, that would give you a score of 2.16. vs Pro and Josh at 4.73. Lower the score the more accurate your prediction. My score is 2.77.

You're the most accurate so far.

 

[Bobcat1]

I think you are selling th WR room short. Tons of talent there. The Edge rushers are better this year. QB room tons better. #1's advantage McCloud oner TJ. #2 Hatter over Hornsby. #3 Jackson over last years Hatter or Rogers. Don't forget the portal to add to this years room.

Until we see what Donley and Sparks can do, we don’t have a strong WR corps other than Holbert & Wilson. QB is definitely a step below last seasons…if we use standard deviation for that too.

 

[Bobcat1]

Yet hatcher went on the play for another fbs school after txst and rogers got no fbs looks and evans hung it up. You seem to be the only one who saw their hidden talents.

Saw it many times live…that’s why they closed practices. When the second and third stringers are schooling the starter, you don’t want prying eyes to see the truth when you already had a predetermined course of action.

Evans was better than Rogers, and Rogers was better than Hatcher. One the players didn’t like, the other the coaches didn’t like, and the third one nobody liked. I’ll keep it to myself which one was which…but on the field, I saw what I saw in person.

 

[Simbo0722]

Vitt was better than all of them.

How bout we just go undefeated and not worry about deviation, prognostication and formulas

 

[franslasttwinkie]

I’m curious which bobcat QB has the most career rushes for 10+ yards.

I’m curious if as a team and QB room we will break the single season record this fall for carries by a QB of 10 yards or more. For as slow as Vitt was he was one of the better downhill QB runners we’ve had. For as nearly equally slow as T. Jones was he might have been one of the most effective downhill QBs we’ve ever had.

Of course if we rush for over 800 yards as a QB room in the regular season it may be an indication of way too many broken plays in Kinne’s scheme or QBs having difficulty adjusting to the speed of his system.

Either way I’ll be interested if Hatter will be used more frequently in packages than Hornsby was. Hatter may not be as quick from point A to B or have MH’s build but he seems to have some more wiggle and ability to shake defenders that might bode better with more PA formations.

 

[Bobcat2013]

I’m curious which bobcat QB has the most career rushes for 10+ yards.

I’m curious if as a team and QB room we will break the single season record this fall for carries by a QB of 10 yards or more. For as slow as Vitt was he was one of the better downhill QB runners we’ve had. For as nearly equally slow as T. Jones was he might have been one of the most effective downhill QBs we’ve ever had.

Of course if we rush for over 800 yards as a QB room in the regular season it may be an indication of way too many broken plays in Kinne’s scheme or QBs having difficulty adjusting to the speed of his system.

Either way I’ll be interested if Hatter will be used more frequently in packages than Hornsby was. Hatter may not be as quick from point A to B or have MH’s build but he seems to have some more wiggle and ability to shake defenders that might bode better with more PA formations.

Idk if I'd call Jones slow. Dude had 718 yds rushing in 2014 and 767 in 2015. That doesn't happen with a slow qb. Vitt's numbers never came close to Tyler Jones.

 

[franslasttwinkie]

Fran’s offense also involved more designed QB runs but there’s no doubt about it that TJ was also great at making something happen out of nothing by tucking it and taking off. He had great vision when it came to eyeing the box and when nobody stayed home on D.

Vitt also played behind trash OLs with both Withers and Spav though and Fran’s was at least mostly serviceable throughout his time. Vitt also split a significant amount of snaps throughout his career too whereas Tyler Jones was never in any real jeopardy of splitting time or losing the job from the Wyoming game forward. Vitt lowered his shoulder when a defender came near and TJ had a tendency to slide.