PFF: AFC South Win Projections

[stitches]

 

 

They project the Colts with the second best chance to win the AFC South, but all 4 teams are jumbled into a single win difference(8.1 for Jaguars, 7.3 for the Texans). 

[NewColtsFan]

3 hours ago, stitches said:

 

 

They project the Colts with the second best chance to win the AFC South, but all 4 teams are jumbled into a single win difference(8.1 for Jaguars, 7.3 for the Texans). 

Not sure win projections us the right title for this thread?

 

They only reveal the number for two of the teams and neither is the Colts.

 

They mostly talk about the odds or percentages for winning the division.

 

It was stunning to see the Colts with the second best odds when they literally said the Colts have the worst roster in the division and it’s all about Luck.    You don’t need a computer to figure that out.

 

This video is a head scratcher for me...

 

[Smonroe]

Interesting.  There’s a good chance that no team will win the division.

 

 

[Lawrence Owen]

So with them having Indy @ 25% chance to with the division,...that would mean a  slightly better chance to go to the playoff's. 

Don't tell USA TODAY this.............

[stitches]

1 hour ago, NewColtsFan said:

Not sure win projections us the right title for this thread?

 

They only reveal the number for two of the teams and neither is the Colts.

 

They mostly talk about the odds or percentages for winning the division.

 

It was stunning to see the Colts with the second best odds when they literally said the Colts have the worst roster in the division and it’s all about Luck.    You don’t need a computer to figure that out.

 

This video is a head scratcher for me...

 

In what way? They said what the top and bottom projection in the division are and they are within 0.8 wins... the Colts and Titans are probably somewhere in the middle(around 7.6-7.7). To me it doesn't matter that much... the important part is just how even the division is. 

 

It's just what their formula spits out. QBs are the great equalizer in the league and their numbers consider Luck one of the best in the league. He's been carrying trash rosters to the playoffs and at least .500 record since he was drafted in the league. Him dragging this team to a + record will not be the exception - it's been the norm for his first 5 years in the league. 

[NewColtsFan]

I think PFF sometimes puts out really interesting stuff.    And sometimes like this they put out stuff that makes me say.....

 

Wait!    What?!?

 

i think if you’re using a computer formula that spits out nonsense then you might have a flawed formula.   You know, garbage in, garbage out.

 

The Colts have the second best odds of winning the division?    Come on.   Texas is last by a good margin?   Come on.    The most wins for any team in the division is 8?    Come on.

 

I’d give the system more credit if they showed how their formula did last year, or the year before.   Does this formula gave a history of accuracy?  The two guys didn’t say so, so I’m highly skeptical. 

 

I think the chances that their predictions hold up are pretty small.    Perhaps we’ll remember this at the beginning of the year.

 

[Smonroe]

13 minutes ago, NewColtsFan said:

I think PFF sometimes puts out really interesting stuff.    And sometimes like this they put out stuff that makes me say.....

 

Wait!    What?!?

 

i think if you’re using a computer formula that spits out nonsense then you might have a flawed formula.   You know, garbage in, garbage out.

 

The Colts have the second best odds of winning the division?    Come on.   Texas is last by a good margin?   Come on.    The most wins for any team in the division is 8?    Come on.

 

I’d give the system more credit if they showed how their formula did last year, or the year before.   Does this formula gave a history of accuracy?  The two guys didn’t say so, so I’m highly skeptical. 

 

I think the chances that their predictions hold up are pretty small.    Perhaps we’ll remember this at the beginning of the year.

 

 

To be fair, they’re basically saying all the teams are pretty even, with the Colts rated that high because of Luck. 

 

You can’t put injuries into the calculations, unless you do a Monte Carlo type of prediction, then you have to weigh each injury’s value.  So that’s the wild card.

 

I think a couple things are clear, the Jags have the best defense.  IMO the Colts have the best QB.  Texans and Titans both have better D’s than the Colts, but that’s because we don’t know the Colts D right now.  Also IMO, offense wins games (how did the Jags D do against Pittsburgh?)

 

Kind of hard to say anyone’s predictions are unlikely unless they’re completely outrageous.  I don’t think they were.  

 

 

 

 

[stitches]

1 hour ago, NewColtsFan said:

I think PFF sometimes puts out really interesting stuff.    And sometimes like this they put out stuff that makes me say.....

 

Wait!    What?!?

 

i think if you’re using a computer formula that spits out nonsense then you might have a flawed formula.   You know, garbage in, garbage out.

 

The Colts have the second best odds of winning the division?    Come on.   Texas is last by a good margin?   Come on.    The most wins for any team in the division is 8?    Come on.

 

I’d give the system more credit if they showed how their formula did last year, or the year before.   Does this formula gave a history of accuracy?  The two guys didn’t say so, so I’m highly skeptical. 

 

I think the chances that their predictions hold up are pretty small.    Perhaps we’ll remember this at the beginning of the year.

 

I don't know why you would assume it's garbage in garbage out formula. Those are smart people who do statistical regression modeling for their livelihood and people that largely know what they are doing. All 32 teams pay for their services. 

 

I am not certain if this specific formula has been successful in the past(or if it has been used in the past?). In their latest podcast they said last year was the first time they used data modeling(in addition to their film grades) to try to project different game outcomes, over/unders, picking against the spread, picking against the total, etc. 

 

In their lock picks against the spread they went 14-6 last year and they went 57% in all of their bets against the spread and on totals. They also entered a fivethirtyeight competition against over 20 000 other people/organizations/formulas and they finished 26th for the season. And this was on their first try with the formula. Again, I reiterate - those are not amateurs that simply throw around numbers for the heck of it or pull numbers out of their butt.

 

I cannot confirm that, but from the discussion in their latest podcast it seems like they are using the formula(or some derivative/development of it) that had those success rates above in order to build their projections for this coming season(the ones that give all 4 teams projections between 7.3 and 8.1). 

 

I don't get your disbelief. What is hard to stomach in those projections? I actually think they are extremely reasonable. Even the by the eye test, without any calculations or advanced analytics, I think this is one of the closest divisions in the NFL and I think the teams are close. Texans being last "by good margin" is actually not that good of a margin. It's 0.8 wins between them. It's like...  13% difference between the first and last. This is NOT a big margin. For example the margin between first(Atlanta) and last(Tampa Bay) in the NFC South is 40% vs 5%. Not every division is projected as close as ours. It just happens that their numbers come out close for all 4 teams. 

 

Colts second best? Well... yah... by like.... 0.05 wins over TEN and by 0.4 over the Texans... this is nothing. And again... I don't see what you think is so unbelievable about it? When Luck is healthy(seems like that's what they assume), the Colts have never had a season under 8 wins. What's so shocking for their projection to come out at 7.5-7.7 or whatever in that range the formula is spitting out? 

 

Texans last? Yah... at 7.3... that's 3+ more wins than last year. Watson was due for regression, they have the worst offensive line in football and their CBs are not great eitehr. There are things in Watson's statistical performance that were unsustainable last year. Just an example - his receivers had only 2 drops for the 7 games he played. This is something he has no control over and this is something that's largely random. If you just give him the league average drop rate his QB rating drops by 10 points. It's also worth pointing out that even while they had Watt and Watson on their roster and while Watson was on fire and playing like a man possessed, they still were not winning games... they were still 3-4 while Watson was playing way over his head. 

 

[NewColtsFan]

44 minutes ago, stitches said:

I don't know why you would assume it's garbage in garbage out formula. Those are smart people who do statistical regression modeling for their livelihood and people that largely know what they are doing. All 32 teams pay for their services. 

 

I am not certain if this specific formula has been successful in the past(or if it has been used in the past?). In their latest podcast they said last year was the first time they used data modeling(in addition to their film grades) to try to project different game outcomes, over/unders, picking against the spread, picking against the total, etc. 

 

In their lock picks against the spread they went 14-6 last year and they went 57% in all of their bets against the spread and on totals. They also entered a fivethirtyeight competition against over 20 000 other people/organizations/formulas and they finished 26th for the season. And this was on their first try with the formula. Again, I reiterate - those are not amateurs that simply throw around numbers for the heck of it or pull numbers out of their butt.

 

I cannot confirm that, but from the discussion in their latest podcast it seems like they are using the formula(or some derivative/development of it) that had those success rates above in order to build their projections for this coming season(the ones that give all 4 teams projections between 7.3 and 8.1). 

 

I don't get your disbelief. What is hard to stomach in those projections? I actually think they are extremely reasonable. Even the by the eye test, without any calculations or advanced analytics, I think this is one of the closest divisions in the NFL and I think the teams are close. Texans being last "by good margin" is actually not that good of a margin. It's 0.8 wins between them. It's like...  13% difference between the first and last. This is NOT a big margin. For example the margin between first(Atlanta) and last(Tampa Bay) in the NFC South is 40% vs 5%. Not every division is projected as close as ours. It just happens that their numbers come out close for all 4 teams. 

 

Colts second best? Well... yah... by like.... 0.05 wins over TEN and by 0.4 over the Texans... this is nothing. And again... I don't see what you think is so unbelievable about it? When Luck is healthy(seems like that's what they assume), the Colts have never had a season under 8 wins. What's so shocking for their projection to come out at 7.5-7.7 or whatever in that range the formula is spitting out? 

 

Texans last? Yah... at 7.3... that's 3+ more wins than last year. Watson was due for regression, they have the worst offensive line in football and their CBs are not great eitehr. There are things in Watson's statistical performance that were unsustainable last year. Just an example - his receivers had only 2 drops for the 7 games he played. This is something he has no control over and this is something that's largely random. If you just give him the league average drop rate his QB rating drops by 10 points. It's also worth pointing out that even while they had Watt and Watson on their roster and while Watson was on fire and playing like a man possessed, they still were not winning games... they were still 3-4 while Watson was playing way over his head. 

 

 

You don't have to defend PFF to me.    I typically sing their praises more than most.

 

There is a group of posters who will never accept PFF no matter what.    They scoff and roll their eyes at the mention of PFF.    These posters think they know better than PFF.    I'm always at odds with those posters.

 

I don't hate everything PFF does.    I don't even dislike most of their work.    I mostly like their work.    But every once in a while I think they drop a stinker and I think this is one of them.

 

Look,  I haven't looked at the other 7 videos for the other 7 divisions.    Maybe I'd like at them and think they're pretty good.    My comment is based on how they view the AFC South.    I think their work here is just silly.    And I'd love to discuss it again in 5 months when the season is over.    See how close they turned out.

 

I'd love to see how they did with their divisional predictions last year.    Maybe they did pretty good.   But typically,  when that happens,  the website will tell you.    PFF didn't share how they did last year.    That's a Red Flag for me.    Let's revisit this at the end of the season.    You think the winner is going to take the division with 8 wins?     I'm guessing no.    9 wins?    I'm guessing no.     We'll see at the end of the year.

 

[NorthernBlue]

1. Colts 10-6

2. Texans 9-7

3. Titans 7-9

4. Jaguars 6-10

 

dont @ me

[stitches]

43 minutes ago, NewColtsFan said:

 

You don't have to defend PFF to me.    I typically sing their praises more than most.

 

There is a group of posters who will never accept PFF no matter what.    They scoff and roll their eyes at the mention of PFF.    These posters think they know better than PFF.    I'm always at odds with those posters.

 

I don't hate everything PFF does.    I don't even dislike most of their work.    I mostly like their work.    But every once in a while I think they drop a stinker and I think this is one of them.

 

Look,  I haven't looked at the other 7 videos for the other 7 divisions.    Maybe I'd like at them and think they're pretty good.    My comment is based on how they view the AFC South.    I think their work here is just silly.    And I'd love to discuss it again in 5 months when the season is over.    See how close they turned out.

 

I'd love to see how they did with their divisional predictions last year.    Maybe they did pretty good.   But typically,  when that happens,  the website will tell you.    PFF didn't share how they did last year.    That's a Red Flag for me.    Let's revisit this at the end of the season.    You think the winner is going to take the division with 8 wins?     I'm guessing no.    9 wins?    I'm guessing no.     We'll see at the end of the year.

 

No projection is perfect or even close to it and PFF is no exception by any means. If you get to about 55% success rate over the long haul on bets that usually have 50% success rate you are doing pretty good.  This still means you will get 45% of them wrong. 

 

There is probably a probability distribution of some sort for number of wins for all those 4 teams, and the number you see here(because it's a single number - 8.1 for Jags, 7.3 for TEN, etc) is either the expected value or some mean value on that probability distribution curve.

 

(I don't have the exact numbers, I'm just giving those specific numbers for illustration purposes) This means that lets say in 25% of the situations Jags will get over 9.5 wins, and in 18% of the situations Colts or TEN will get over 9.5 wins and in 14% of the situations Texans will get over 9.5 wins.... So chances are still that you will get a team over 9.5 wins, just the chance for any single one of them is relatively small(25%  and under?). This means you will very likely not get 8-8-8-7 wins for the 4 teams...  in fact... whatever result you get for this particular division you cannot really say whether it was a good projection or not. You will need much bigger sample for that and the measure would have to be how much the big sample deviates from the actual results compared to other projection models. 

 

Just an example from a statistical projection that we've had 3 years of data on - Kevin Pelton's yearly NBA RPM win projections. Here's a thread on it on reddit. 

Quote

 

2015-16:

Win Deviation: 7.7; Vegas: 8.9

Over/Under: 22/30 correct

Coefficient of Determination (r2 ): .703; Vegas: .586

2016-17:

Win Deviation: 4.9; Vegas: 5.1

Over/Under: 18/30 correct

Coefficient of Determination (r2 ): .816; Vegas: .790

2017-18:

Win Deviation: 7.6; Vegas: 7.3

Over/Under: 18/30 correct

Coefficient of Determination (r2 ): .609; Vegas: .650;

 

 

In other words Pelton has 64.4% win rate against the Vegas lines. If this holds up on a larger sample, this is insanely good projection model, but this is with a sport that has developed much better advanced analytics models. 57% success rate (what PFF claimed they have on 50/50 bets) is still pretty damn good if they can extend it to their season total win projections. 

[BR-549]

I hated statistics in college.  For that matter I hated statics too.  Dynamics was much more enjoyable especially thermal dynamics.  I didn't like Fortran though no matter how much I wanted to, probably due to the poor English speaking TA's in large lecture halls we had to deal with .... give you any idea of when I went to school?

 

It probably didn't help that my statistics professor brought a bamboo bowl to class every day.  No it didn't contain soup …. what is the probability that it would?  No, it had water in it and he drank out of it like he was sipping the last bit of sugary milk left over after a bowl full of frosted flakes.  He also had an annoying way of saying "little sigma tilde"  I cant reproduce it with typed words … use your imagination.  

 

Long story short ... he ruined any enthusiasm I may have mustered up for statistics and these days I usually just ignore anything to do with the subject.

 

 

[MacDee1975]

On 8/4/2018 at 4:05 PM, stitches said:

I don't know why you would assume it's garbage in garbage out formula. Those are smart people who do statistical regression modeling for their livelihood and people that largely know what they are doing. All 32 teams pay for their services. 

 

I am not certain if this specific formula has been successful in the past(or if it has been used in the past?). In their latest podcast they said last year was the first time they used data modeling(in addition to their film grades) to try to project different game outcomes, over/unders, picking against the spread, picking against the total, etc. 

 

In their lock picks against the spread they went 14-6 last year and they went 57% in all of their bets against the spread and on totals. They also entered a fivethirtyeight competition against over 20 000 other people/organizations/formulas and they finished 26th for the season. And this was on their first try with the formula. Again, I reiterate - those are not amateurs that simply throw around numbers for the heck of it or pull numbers out of their butt.

 

I cannot confirm that, but from the discussion in their latest podcast it seems like they are using the formula(or some derivative/development of it) that had those success rates above in order to build their projections for this coming season(the ones that give all 4 teams projections between 7.3 and 8.1). 

 

I don't get your disbelief. What is hard to stomach in those projections? I actually think they are extremely reasonable. Even the by the eye test, without any calculations or advanced analytics, I think this is one of the closest divisions in the NFL and I think the teams are close. Texans being last "by good margin" is actually not that good of a margin. It's 0.8 wins between them. It's like...  13% difference between the first and last. This is NOT a big margin. For example the margin between first(Atlanta) and last(Tampa Bay) in the NFC South is 40% vs 5%. Not every division is projected as close as ours. It just happens that their numbers come out close for all 4 teams. 

 

Colts second best? Well... yah... by like.... 0.05 wins over TEN and by 0.4 over the Texans... this is nothing. And again... I don't see what you think is so unbelievable about it? When Luck is healthy(seems like that's what they assume), the Colts have never had a season under 8 wins. What's so shocking for their projection to come out at 7.5-7.7 or whatever in that range the formula is spitting out? 

 

Texans last? Yah... at 7.3... that's 3+ more wins than last year. Watson was due for regression, they have the worst offensive line in football and their CBs are not great eitehr. There are things in Watson's statistical performance that were unsustainable last year. Just an example - his receivers had only 2 drops for the 7 games he played. This is something he has no control over and this is something that's largely random. If you just give him the league average drop rate his QB rating drops by 10 points. It's also worth pointing out that even while they had Watt and Watson on their roster and while Watson was on fire and playing like a man possessed, they still were not winning games... they were still 3-4 while Watson was playing way over his head. 

 

 

Excellent points that nobody ever seems to bring up.  Watson is in no way, shape or form an established entity.  He has a sample size of less than half an NFL season, against garbage defenses, where nobody was gameplanning for him.  Not to mention he has now had 2 serious knee injuries before the age of 23.  He is as much a wild card as anyone else going into next season.