Analysis (Long): Pitt Football Attendance, 2001-2017

[ProvoSpain]

Hi everyone. I've been reading for years, but I'm quite new to the board as a contributor. Thanks for welcoming me to the discussions up to now.

I wanted to share some (amateur) analysis that I had done on a often-discussed topic around here: Pitt football attendance. I don't mean to dwell on it, and I'm kind of annoyed that this is the first thread that I've started, as I risk being called out as a troll. But I think that I have done some work on it that you might find to be different than the same tired arguments. Specifically, I did a statistical analysis of all 112 Pitt home football games from the opening of Heinz Field through the end of the 2017 season. The sample is small, and I need to revisit the model with data for the last couple years, but I was surprised to find so many factors that are statistically significant. I looked at all available data points that I could think of: team performance, time of day, time of year, opponent, weather, and several other factors. And here is what I found:

• Start with 25,400
• Add 2,000 if playing a G5 school.
• Add 4,000 if playing a conference game.
• Add 6,000 if playing a non-conference P5 school that is not ND, PSU, WVU
• Add 16,100 if playing Penn State, Notre Dame, or West Virginia
• Add 3,500 if the Steelers are at home the next day
• Add 4,500 if first home game
• Add 2,800 if Homecoming
• Add 2,600 if it’s a dry day (not raining)
• Add 2,700 is the opponent is ranked but Pitt is not
• Add 5,400 if Pitt is ranked but the opponent is not
• Add 8,100 if both teams are ranked
• Count how many wins the team has at the end of the season. Multiply that number by 1,600 and add it to the total (this doesn't help look forward, but it really increases accuracy).
• If not first home game, count how many idle/away weeks the team has had since it’s last home game. Multiply that number by 900 and add it to the total.
• Count how many home games Pitt has already had so far that season. Multiply that number by 1,100 and subtract it from the total.

There are some other things that correlate with attendance (they get 2,000 less when the Penguins have a home game the same day), but it doesn’t make the model any more accurate (it’s just correlation, not causation), so I left it out.

If this is found to be interesting, I can share more findings. If not, well it'll just get buried below the other interesting topics here, no harm done. Thanks for reading, if you've made it this far.

 

[UPitt '89]

Hi everyone. I've been reading for years, but I'm quite new to the board as a contributor. Thanks for welcoming me to the discussions up to now.

I wanted to share some (amateur) analysis that I had done on a often-discussed topic around here: Pitt football attendance. I don't mean to dwell on it, and I'm kind of annoyed that this is the first thread that I've started, as I risk being called out as a troll. But I think that I have done some work on it that you might find to be different than the same tired arguments. Specifically, I did a statistical analysis of all 112 Pitt home football games from the opening of Heinz Field through the end of the 2017 season. The sample is small, and I need to revisit the model with data for the last couple years, but I was surprised to find so many factors that are statistically significant. I looked at all available data points that I could think of: team performance, time of day, time of year, opponent, weather, and several other factors. And here is what I found:

• Start with 25,400
• Add 2,000 if playing a G5 school.
• Add 4,000 if playing a conference game.
• Add 6,000 if playing a non-conference P5 school that is not ND, PSU, WVU
• Add 16,100 if playing Penn State, Notre Dame, or West Virginia
• Add 3,500 if the Steelers are at home the next day
• Add 4,500 if first home game
• Add 2,800 if Homecoming
• Add 2,600 if it’s a dry day (not raining)
• Add 2,700 is the opponent is ranked but Pitt is not
• Add 5,400 if Pitt is ranked but the opponent is not
• Add 8,100 if both teams are ranked
• Count how many wins the team has at the end of the season. Multiply that number by 1,600 and add it to the total (this doesn't help look forward, but it really increases accuracy).
• If not first home game, count how many idle/away weeks the team has had since it’s last home game. Multiply that number by 900 and add it to the total.
• Count how many home games Pitt has already had so far that season. Multiply that number by 1,100 and subtract it from the total.

There are some other things that correlate with attendance (they get 2,000 less when the Penguins have a home game the same day), but it doesn’t make the model any more accurate (it’s just correlation, not causation), so I left it out.

If this is found to be interesting, I can share more findings. If not, well it'll just get buried below the other interesting topics here, no harm done. Thanks for reading, if you've made it this far.

 

[recruitsreadtheseboards]

I would think time, game start, would be interesting.

 

[Titos Panther]

TIMMY PATERNO

 

[9nationalchampionships]

E = mc2

 

[ProvoSpain]

Time of day was not found to be statistically significant, but here are the raw data, grouped by early, mid, late.

Time2​	Avg Attendance	# of Games
After 7pm​	49,938​	25​
Before 2pm​	44,448​	66​
2-7pm​	43,250​	21​
Grand Total​	45,449​	112​

 

[PittMan2003]

Hi everyone. I've been reading for years, but I'm quite new to the board as a contributor. Thanks for welcoming me to the discussions up to now.

I wanted to share some (amateur) analysis that I had done on a often-discussed topic around here: Pitt football attendance. I don't mean to dwell on it, and I'm kind of annoyed that this is the first thread that I've started, as I risk being called out as a troll. But I think that I have done some work on it that you might find to be different than the same tired arguments. Specifically, I did a statistical analysis of all 112 Pitt home football games from the opening of Heinz Field through the end of the 2017 season. The sample is small, and I need to revisit the model with data for the last couple years, but I was surprised to find so many factors that are statistically significant. I looked at all available data points that I could think of: team performance, time of day, time of year, opponent, weather, and several other factors. And here is what I found:

• Start with 25,400
• Add 2,000 if playing a G5 school.
• Add 4,000 if playing a conference game.
• Add 6,000 if playing a non-conference P5 school that is not ND, PSU, WVU
• Add 16,100 if playing Penn State, Notre Dame, or West Virginia
• Add 3,500 if the Steelers are at home the next day
• Add 4,500 if first home game
• Add 2,800 if Homecoming
• Add 2,600 if it’s a dry day (not raining)
• Add 2,700 is the opponent is ranked but Pitt is not
• Add 5,400 if Pitt is ranked but the opponent is not
• Add 8,100 if both teams are ranked
• Count how many wins the team has at the end of the season. Multiply that number by 1,600 and add it to the total (this doesn't help look forward, but it really increases accuracy).
• If not first home game, count how many idle/away weeks the team has had since it’s last home game. Multiply that number by 900 and add it to the total.
• Count how many home games Pitt has already had so far that season. Multiply that number by 1,100 and subtract it from the total.

There are some other things that correlate with attendance (they get 2,000 less when the Penguins have a home game the same day), but it doesn’t make the model any more accurate (it’s just correlation, not causation), so I left it out.

If this is found to be interesting, I can share more findings. If not, well it'll just get buried below the other interesting topics here, no harm done. Thanks for reading, if you've made it this far.

Very interesting. Thanks for sharing

 

[HailToPitt725]

Hi everyone. I've been reading for years, but I'm quite new to the board as a contributor. Thanks for welcoming me to the discussions up to now.

I wanted to share some (amateur) analysis that I had done on a often-discussed topic around here: Pitt football attendance. I don't mean to dwell on it, and I'm kind of annoyed that this is the first thread that I've started, as I risk being called out as a troll. But I think that I have done some work on it that you might find to be different than the same tired arguments. Specifically, I did a statistical analysis of all 112 Pitt home football games from the opening of Heinz Field through the end of the 2017 season. The sample is small, and I need to revisit the model with data for the last couple years, but I was surprised to find so many factors that are statistically significant. I looked at all available data points that I could think of: team performance, time of day, time of year, opponent, weather, and several other factors. And here is what I found:

• Start with 25,400
• Add 2,000 if playing a G5 school.
• Add 4,000 if playing a conference game.
• Add 6,000 if playing a non-conference P5 school that is not ND, PSU, WVU
• Add 16,100 if playing Penn State, Notre Dame, or West Virginia
• Add 3,500 if the Steelers are at home the next day
• Add 4,500 if first home game
• Add 2,800 if Homecoming
• Add 2,600 if it’s a dry day (not raining)
• Add 2,700 is the opponent is ranked but Pitt is not
• Add 5,400 if Pitt is ranked but the opponent is not
• Add 8,100 if both teams are ranked
• Count how many wins the team has at the end of the season. Multiply that number by 1,600 and add it to the total (this doesn't help look forward, but it really increases accuracy).
• If not first home game, count how many idle/away weeks the team has had since it’s last home game. Multiply that number by 900 and add it to the total.
• Count how many home games Pitt has already had so far that season. Multiply that number by 1,100 and subtract it from the total.

There are some other things that correlate with attendance (they get 2,000 less when the Penguins have a home game the same day), but it doesn’t make the model any more accurate (it’s just correlation, not causation), so I left it out.

If this is found to be interesting, I can share more findings. If not, well it'll just get buried below the other interesting topics here, no harm done. Thanks for reading, if you've made it this far.

Love stuff like this, thanks for sharing.

 

[Macchiato]

How does Thurs evening game change attendance

 

[King Of All Message Boards]

What impact does the time the parking lots open have?

 

[ProvoSpain]

How does Thurs evening game change attendance

Not statistically significant but here is the data!

Day of Week​	Avg Attendance	# of Games
Monday​	65,500​	1​
Wednesday​	30,103​	1​
Thursday​	46,435​	10​
Friday​	47,938​	5​
Saturday​	45,164​	95​
Grand Total​	45,449​	112​

 

[Joe the Panther Fan]

How does the fact that the numbers that Pitt has announced frequently had little to do with reality play into this? For instance everyone who was there knows that that 30,103 for the Wednesday game is so ridiculously overinflated that that number is meaningless.

 

[pitt1976]

All of this shows me that a 50,000-seat new on campus stadium will satisfy our needs.

 

[ProvoSpain]

How does the fact that the numbers that Pitt has announced frequently had little to do with reality play into this? For instance everyone who was there knows that that 30,103 for the Wednesday game is so ridiculously overinflated that that number is meaningless.

I agree that “announced” attendance is very different than actual. And it definitely takes away from the value of the analysis.

I reached out to the athletic department at the time to ask them about their numbers. They were careful not to reveal too much, but they assured me that there is a consistency to how they report it. It gave me confidence to move forward, even if actual “butts in seats” is more interesting or actionable data.

 

[Sean Miller Fan]

Hi everyone. I've been reading for years, but I'm quite new to the board as a contributor. Thanks for welcoming me to the discussions up to now.

I wanted to share some (amateur) analysis that I had done on a often-discussed topic around here: Pitt football attendance. I don't mean to dwell on it, and I'm kind of annoyed that this is the first thread that I've started, as I risk being called out as a troll. But I think that I have done some work on it that you might find to be different than the same tired arguments. Specifically, I did a statistical analysis of all 112 Pitt home football games from the opening of Heinz Field through the end of the 2017 season. The sample is small, and I need to revisit the model with data for the last couple years, but I was surprised to find so many factors that are statistically significant. I looked at all available data points that I could think of: team performance, time of day, time of year, opponent, weather, and several other factors. And here is what I found:

• Start with 25,400
• Add 2,000 if playing a G5 school.
• Add 4,000 if playing a conference game.
• Add 6,000 if playing a non-conference P5 school that is not ND, PSU, WVU
• Add 16,100 if playing Penn State, Notre Dame, or West Virginia
• Add 3,500 if the Steelers are at home the next day
• Add 4,500 if first home game
• Add 2,800 if Homecoming
• Add 2,600 if it’s a dry day (not raining)
• Add 2,700 is the opponent is ranked but Pitt is not
• Add 5,400 if Pitt is ranked but the opponent is not
• Add 8,100 if both teams are ranked
• Count how many wins the team has at the end of the season. Multiply that number by 1,600 and add it to the total (this doesn't help look forward, but it really increases accuracy).
• If not first home game, count how many idle/away weeks the team has had since it’s last home game. Multiply that number by 900 and add it to the total.
• Count how many home games Pitt has already had so far that season. Multiply that number by 1,100 and subtract it from the total.

There are some other things that correlate with attendance (they get 2,000 less when the Penguins have a home game the same day), but it doesn’t make the model any more accurate (it’s just correlation, not causation), so I left it out.

If this is found to be interesting, I can share more findings. If not, well it'll just get buried below the other interesting topics here, no harm done. Thanks for reading, if you've made it this far.

Pitt makes up attendance numbers though. This year, they seemed to be pretty accurate.

 

[Sean Miller Fan]

How does the fact that the numbers that Pitt has announced frequently had little to do with reality play into this? For instance everyone who was there knows that that 30,103 for the Wednesday game is so ridiculously overinflated that that number is meaningless.

That was the game against Navy and actually, I thought that was accurate. However, they do make up a lot of others.

 

[ProvoSpain]

Pitt makes up attendance numbers though. This year, they seemed to be pretty accurate.

I’m not a statistician by trade, but my analysis attempts to PROVE that the numbers AREN’T made up. The listed factors explain the variance with statistical significance. The numbers might be “inflated” relative to actual attendance for non-sellouts, but they are not random.

 

[HailToPitt725]

All of this shows me that a 50,000-seat new on campus stadium will satisfy our needs.

No better time than now, Heather. Do it!

 

[Chairman Moe]

Is that where the parking lot and parking garage are now?

 

[Sean Miller Fan]

I’m not a statistician by trade, but my analysis attempts to PROVE that the numbers AREN’T made up. The listed factors explain the variance with statistical significance. The numbers might be “inflated” relative to actual attendance for non-sellouts, but they are not random.

I went to the Pitt/UMBC basketball game and counted the fans. I counted 978 in total. Announced attendance was 7400. My count may have bee off by 100-200 either way but it certainly wasnt off by 6000+. You could have fit everyone in the arena in the Zoo section. I know this is basketball we're talking about but if they're wildly inflating basketball numbers, you'd be sure they're doing the same thing in football. However, this isn't just Pitt that does this. Most colleges lie about attendance.

 

[HailToPitt725]

Is that where the parking lot and parking garage are now?

Yup, currently slated to be the new home of Pitt women’s lacrosse. With that being said, I do have my theories on that site and Victory Heights as a whole…

 

[bwh05]

Yup, currently slated to be the new home of Pitt women’s lacrosse. With that being said, I do have my theories on that site and Victory Heights as a whole…

Lax isn’t going to be there now. There are no plans for Victory Heights facilities on the OC Lot.

 

[bwh05]

How does the fact that the numbers that Pitt has announced frequently had little to do with reality play into this? For instance everyone who was there knows that that 30,103 for the Wednesday game is so ridiculously overinflated that that number is meaningless.

This is definitely true. Pederson was the worst regime for padding the numbers.

 

[Chairman Moe]

Yup, currently slated to be the new home of Pitt women’s lacrosse. With that being said, I do have my theories on that site and Victory Heights as a whole…

I lived in Cincinnati 11 years ago, and UC has done a great makeover with Nippert Stadium. Like Pitt, it’s a “landlocked” city campus with limited access and parking. They somehow made it work. Would like to see Pitt do likewise but I’m not holding my breath.

Regarding lacrosse, couldn’t it be played on a football field? I’ve never played it so am unfamiliar with the field dimensions …

 

[bwh05]

I lived in Cincinnati 11 years ago, and UC has done a great makeover with Nippert Stadium. Like Pitt, it’s a “landlocked” city campus with limited access and parking. They somehow made it work. Would like to see Pitt do likewise but I’m not holding my breath.

Regarding lacrosse, couldn’t it be played on a football field? I’ve never played it so am unfamiliar with the field dimensions …

Lax is a few yards wider but it would be possible.

 

[Upg bobcat]

Pitt makes up attendance numbers though. This year, they seemed to be pretty accurate.

They very well could use this exact formula to make up those numbers -- Provo, do you work at Pitt and is this inside info?

 

[recruitsreadtheseboards]

Well it does say one thing. Like 80% of college football programs, you need no more than a 50K stadium.

 

[TFBaum]

I lived in Cincinnati 11 years ago, and UC has done a great makeover with Nippert Stadium. Like Pitt, it’s a “landlocked” city campus with limited access and parking. They somehow made it work. Would like to see Pitt do likewise but I’m not holding my breath.

Regarding lacrosse, couldn’t it be played on a football field? I’ve never played it so am unfamiliar with the field dimensions …

Yes we play lacrosse on football fields. A small seating capacity lacrosse field in affordable

 

[ProvoSpain]

They very well could use this exact formula to make up those numbers -- Provo, do you work at Pitt and is this inside info?

I have no affiliation with Pitt. I’m just a fan of their sports teams who used publicly available information to perform this analysis. I wanted to figure out which factors “move the needle” and which do not.

For example, it has been argued that 3:30 starts draw better than Noon starts. This is actually not true. And it’s also not true that Thursday night draws worse than Saturday.

Again, I fully concede that announced attendance does not equal actual attendance. But the two move together. And i’m confident that we can draw conclusions from the data.

For example, one can ask whether the extra 4K in attendance for a P5 OOC home game (as opposed to G5) is worth the increased risk of losing, given that winning and being ranked are also proven to impact perception (and attendance) in a positive way

Is it better to play a 50/50 game with Iowa? Or a 90/10 game with Kent State? Reasonable people have different opinions on that.

 

[Sean Miller Fan]

I have no affiliation with Pitt. I’m just a fan of their sports teams who used publicly available information to perform this analysis. I wanted to figure out which factors “move the needle” and which do not.

For example, it has been argued that 3:30 starts draw better than Noon starts. This is actually not true. And it’s also not true that Thursday night draws worse than Saturday.

Again, I fully concede that announced attendance does not equal actual attendance. But the two move together. And i’m confident that we can draw conclusions from the data.

For example, one can ask whether the extra 4K in attendance for a P5 OOC home game (as opposed to G5) is worth the increased risk of losing, given that winning and being ranked are also proven to impact perception (and attendance) in a positive way

Is it better to play a 50/50 game with Iowa? Or a 90/10 game with Kent State? Reasonable people have different opinions on that.

Pitt should not schedule the Iowa's and Cincinnati's. Too much risk of loss with no attendance increase or "buzz."

These are the only P5s we should play

PSU
WVU
ND
Ohio State
Michigan
Oklahoma
Texas
Florida
LSU
USC

 

[pittchagg]

So the Steelers are good for Pitt attendance and the Penguins are bad for Pitt attendance? Move them to Kansas City!

Curious to see the correlation between Pirates games.

 

[Pitt79]

I went to the Pitt/UMBC basketball game and counted the fans. I counted 978 in total. Announced attendance was 7400. My count may have bee off by 100-200 either way but it certainly wasnt off by 6000+. You could have fit everyone in the arena in the Zoo section. I know this is basketball we're talking about but if they're wildly inflating basketball numbers, you'd be sure they're doing the same thing in football. However, this isn't just Pitt that does this. Most colleges lie about attendance.

Honestly, why do you care that much? Attendance for football is not much different than the 70s when they where great, Pitt football will be between 40-45K forever probably, as good as or better than college teams in an NFL town. And who cares what people see on TV? I don't go to other teams message boards BECAUSE I DON'T CARE THAT MUCH, but do people in Atlanta, Miami or anywhere else where they play in an NFL stadium agonize over empty seats seen on TV like Pitt fans do? As for basketball, it's proven we can fill that place and the reason we don't now is obvious, we don't need to guess.

 

[pittdan77]

Curious to see the correlation between Pirates games.

12:30 kickoffs in September.

 

[caleco's]

All of this shows me that a 50,000-seat new on campus stadium will satisfy our needs.

thats what she said

 

[TFBaum]

Hi everyone. I've been reading for years, but I'm quite new to the board as a contributor. Thanks for welcoming me to the discussions up to now.

I wanted to share some (amateur) analysis that I had done on a often-discussed topic around here: Pitt football attendance. I don't mean to dwell on it, and I'm kind of annoyed that this is the first thread that I've started, as I risk being called out as a troll. But I think that I have done some work on it that you might find to be different than the same tired arguments. Specifically, I did a statistical analysis of all 112 Pitt home football games from the opening of Heinz Field through the end of the 2017 season. The sample is small, and I need to revisit the model with data for the last couple years, but I was surprised to find so many factors that are statistically significant. I looked at all available data points that I could think of: team performance, time of day, time of year, opponent, weather, and several other factors. And here is what I found:

• Start with 25,400
• Add 2,000 if playing a G5 school.
• Add 4,000 if playing a conference game.
• Add 6,000 if playing a non-conference P5 school that is not ND, PSU, WVU
• Add 16,100 if playing Penn State, Notre Dame, or West Virginia
• Add 3,500 if the Steelers are at home the next day
• Add 4,500 if first home game
• Add 2,800 if Homecoming
• Add 2,600 if it’s a dry day (not raining)
• Add 2,700 is the opponent is ranked but Pitt is not
• Add 5,400 if Pitt is ranked but the opponent is not
• Add 8,100 if both teams are ranked
• Count how many wins the team has at the end of the season. Multiply that number by 1,600 and add it to the total (this doesn't help look forward, but it really increases accuracy).
• If not first home game, count how many idle/away weeks the team has had since it’s last home game. Multiply that number by 900 and add it to the total.
• Count how many home games Pitt has already had so far that season. Multiply that number by 1,100 and subtract it from the total.

There are some other things that correlate with attendance (they get 2,000 less when the Penguins have a home game the same day), but it doesn’t make the model any more accurate (it’s just correlation, not causation), so I left it out.

If this is found to be interesting, I can share more findings. If not, well it'll just get buried below the other interesting topics here, no harm done. Thanks for reading, if you've made it this far.

I am a Data Scientist that was a nice analysis

 

[TFBaum]

Honestly, why do you care that much? Attendance for football is not much different than the 70s when they where great, Pitt football will be between 40-45K forever probably, as good as or better than college teams in an NFL town. And who cares what people see on TV? I don't go to other teams message boards BECAUSE I DON'T CARE THAT MUCH, but do people in Atlanta, Miami or anywhere else where they play in an NFL stadium agonize over empty seats seen on TV like Pitt fans do? As for basketball, it's proven we can fill that place and the reason we don't now is obvious, we don't need to guess.

You know the only group that discusses attendance our cousins in Central PA. I with you do you think UCLA fans worry about attendance? No my in-laws all went to UCLA and have season tickets

 

[ProvoSpain]

So the Steelers are good for Pitt attendance and the Penguins are bad for Pitt attendance? Move them to Kansas City!

Curious to see the correlation between Pirates games.

Pitt averaged 1200 more fans at their game when the Pirates had a game the same day, but it was not found to be statistically significant by the model.

 

[TFBaum]

Pitt averaged 1200 more fans at their game when the Pirates had a game the same day, but it was not found to be statistically significant by the model.

Did you do a correlation matrix?

 

[ProvoSpain]

Did you do a correlation matrix?

I did a regression analysis using plain old Excel with attendance as dependent variable.

But I suppose that I easily could extend by a correlation matrix. Didn’t occur to me haha.

 

[HailtoPitt]

In summary, Heinz Field is too big for Pitt.