NFL HFA by team

[Prop Man]

After reading an interesting blog post in the New York Times (http://fifthdown.blogs.nytimes.com/2009/09/18/belichick-on-giants-stadium/) with Belichick talking about a little about Giants HFA vs Jets HFA (or at least how it felt), I decided to look at the cumulative HFA for all NFL teams since 1994. I calculated HFA by looking at the difference between Home vs Away scores at Home vs Away vs Home scores on the road and then divded by two. The average team was around 2.8 as expected.

I was expecting to see the Jets near the bottom, but surprisingly, it was the Giants at the very bottom.

STL 4.3
MIN 4.2
KC 4.2
ARI 4.1
DEN 4.1
SF 4.1
BAL 3.9
JAX 3.9
DET 3.4
DAL 3.4
IND 3.2
GB 3.2
BUF 3.1
CHI 3.1
TB 3.0
SEA 3.0
CLE 2.7
HOU 2.6
ATL 2.6
OAK 2.5
MIA 2.3
NYJ 2.3
PIT 2.2
PHI 2.1
TEN 2.0
SD 2.0
CAR 1.9
WAS 1.7
CIN 1.4
NE 1.3
NO 0.9
NYG 0.8

[custer]

A better test might be to compare the correlations between the last 2 successive 5 year periods.

[Prop Man]

Interesting test. I like it.

I ran it. 2004-2008 vs 1999-2003. Correlation was 0.09

Pretty low.

[PerpetualCzech]

Correlation was 0.09

Pretty low.

Hmm, I'm not so sure about that. That result actually feels a little meaningful to me. For example, the correlation result I get for my strongest 2H "angle" is only +0.07 and it has historically produced terrific returns.

I realize my comparison may be a case of apples and oranges. But this is the problem I always run into when looking at these correlation co-efficients: I am pretty blind with my feel about what kind of result is worth getting excited about in any given test. I'd be interested in Comptrbob's opinion about this particular one.

[PerpetualCzech]

Maybe the first place to start when trying to test significance would be to figure out the SD for the 10-year result ...

[WarDekar]

Wait, so are you pairing games and only counting them if 2 teams both played each other home/away in the same year? If not, how are you adjusting for schedule?

[Prop Man]

Wait, so are you pairing games and only counting them if 2 teams both played each other home/away in the same year? If not, how are you adjusting for schedule?


I did it very simply. No adjustment for schedule, no pairing games.

[Prop Man]

For example, the correlation result I get for my strongest 2H "angle" is only +0.07 and it has historically produced terrific returns.



What are you comparing for the 0.07 correlation? Your estimated 2H difference versus the actual 2H result?

[WarDekar]

I did it very simply. No adjustment for schedule, no pairing games.

I'm not sure how useful of data you're going to get without somehow accounting for schedule - when accounting for average HFA as a whole across a sport, obviously it's not an issue. But when you start breaking it down to teams and only a couple years span (in a sport like NFL where they only play 8 and 8), the variations in scheduling are going to play a big enough factor in these #s that I wouldn't put much weight in them (among other reasons).

[Prop Man]

Good point, makes sense to me.

[PerpetualCzech]

What are you comparing for the 0.07 correlation? Your estimated 2H difference versus the actual 2H result?

I generate a rating for every team (for every game) based on a stat I create and then compare to the return on a hypothetical 1 unit bet on that team based on the closing 2H line. I have well over 3000 games to go by though so after thinking about it more I'm sure it's not a valid comparison to your rankings.

I'm sure a way to get a feel for significance of your chart is to calculate the standard deviation. With 32 teams, if the results are random approx. 5 of them (~16%) should lie more than 1 SD away from the average on either side. Eyeballing the chart, that corresponds to a SD of about 1.2. If the SD ends up being 1.2 or more then we can be confident the results are random (unfortunately I don't know how to calculate it ...)

[Rust]

Regardless of what you may think about the validity of your results, why are the Giants surprising? Eli's problems playing at home are well documented. A few others are where you'd expect them to be - NE, DEN, KC, MIN.

[trixtrix]

Regardless of what you may think about the validity of your results, why are the Giants surprising? Eli's problems playing at home are well documented. A few others are where you'd expect them to be - NE, DEN, KC, MIN.

eli has played every giants game since 1994?

[Rust]

He hasn't, but it's not like Prop Man is using the median or anything. Significant outliers will skew the #s.

[EdTeach]

In the last 15 years a number of new stadiums have been built...I have no data to back this up but I generally think the new Seahawks stadium has a much greater homefield advantage than the old Kingdome. I think the stadium differences are a relevant factor...

[custer]

Arizona would be another one on that list.

[goose1489]

are you inferring that the Tempe raucous crowds of 30,000 had no effect?

[Dreamer]

It's easy to make the mistake of confusing above average HFA with teams that just don't travel well..........

D.

[PerpetualCzech]

It's easy to make the mistake of confusing above average HFA with teams that just don't travel well..........

D.

If such teams actually existed (I am skeptical), for the purposes of the chart in the OP such an effect would even out over the long run.

[The Chaperone]

Wouldn't teams that 'don't travel well' often be teams with the better HFA? For example, Seattle is the most remote location in the NFL, therefore on average visiting teams have traveled the farthest to come to Qwest Field, but the flipside is that when the Seahawks play road games they are also on average traveling farther to those road games than other road teams.

I was surprised that SEA didn't rank higher in this list, but who knows, could be statistical noise. They've definitely dominated at home in games that count over the past few years and they're good for one or two eastern time zone games per year where they just don't show up at all.

FWIW, they actually keep track of 'Opponents' false starts' at Qwest and the Seahawks are first in that category since 2005. The Vikings are a distant second. Both stadiums are known for being loud.

Lastly, to EdTeach: back in the day the Kingdome was known as an extremely loud place with a huge HFA. The problem was the team was god awful in the 90s. I think they still did alright attendance wise, but crowds can be somewhat uninspired when the team is 2-14 or whatever. They also played in Husky Stadium for one or two seasons while Qwest was being built. While that stadium may be a good CFB stadium, it was quite subpar as an NFL stadium.

[Rust]

If such teams actually existed (I am skeptical), for the purposes of the chart in the OP such an effect would even out over the long run.

I could see the 'not travelling well' effect coming into play, mainly discipline issues/being mentally unprepared by away crowds, manifesting as penalties.

[Dreamer]

Wouldn't teams that 'don't travel well' often be teams with the better HFA? For example, Seattle is the most remote location in the NFL, therefore on average visiting teams have traveled the farthest to come to Qwest Field, but the flipside is that when the Seahawks play road games they are also on average traveling farther to those road games than other road teams.

I was surprised that SEA didn't rank higher in this list, but who knows, could be statistical noise. They've definitely dominated at home in games that count over the past few years and they're good for one or two eastern time zone games per year where they just don't show up at all.

FWIW, they actually keep track of 'Opponents' false starts' at Qwest and the Seahawks are first in that category since 2005. The Vikings are a distant second. Both stadiums are known for being loud.

Lastly, to EdTeach: back in the day the Kingdome was known as an extremely loud place with a huge HFA. The problem was the team was god awful in the 90s. I think they still did alright attendance wise, but crowds can be somewhat uninspired when the team is 2-14 or whatever. They also played in Husky Stadium for one or two seasons while Qwest was being built. While that stadium may be a good CFB stadium, it was quite subpar as an NFL stadium.

Giving a team a standard HFA is fine for bulk market analysis.

You don't care in the long run as you are getting most of your edge from getting a good number. Some games will be lower, some higher, who cares right. Long term you probably have an edge.

But if you are genuinely handicapping a specific matchup strength (vs a WA line) you need superior analysis to overcome tough to beat lines.

Its odd when I see people suggest that one team has a greater HFA than another team.

That info is totally useless unless you are able to analyze if that advantage is already built into the line or not.

It's the same with situational bettors..........they have no way of knowing if the lines/market has caught up and they no longer have an edge.

D.

[kgbted21]

"Its odd when I see people suggest that one team has a greater HFA than another team."

Your post implies that you do not factor team by team HFA in your own numbers. Maybe the market doesn't use these either which would mean it matters (if the numbers are statisically significant.)

[custer]

are you inferring that the Tempe raucous crowds of 30,000 had no effect?

Those crowds were often raucous, but they typically rooted for the visitor.

[Dreamer]

"Its odd when I see people suggest that one team has a greater HFA than another team."

Your post implies that you do not factor team by team HFA in your own numbers. Maybe the market doesn't use these either which would mean it matters (if the numbers are statisically significant.)

Classic quote "out of context"

The real meat is the line that follows the quote.

"That info is totally useless unless you are able to analyze if that advantage is already built into the line or not."

We make bets vs a market line.

I know many will cry "soooo fucking obvious" .......yet..........I never see it mentioned in bulk analysis.

It's as useful as situational betting........... hold your nose and hope you have an edge going forward.

IMHO Superior analysis keeps winning going forward.
IMHO Market bettors will have a much tougher time.

D.

[kimlee]

That info is totally useless unless you are able to analyze if that advantage is already built into the line or not.

Ummm, isn't that true of any information? Where is Captain Obvious when we need him?

[PerpetualCzech]

I could see the 'not travelling well' effect coming into play, mainly discipline issues/being mentally unprepared by away crowds, manifesting as penalties.

Sure, but all you are doing here is simply describing HFA. We know that exists and that it is worth on average 2.5-2.7 points per game (depending on how far back you want to go). I might even accept that some players get affected by this phenomenon more than others and it affects their performance to different degrees.

What I find difficult to swallow is the theory that this quality changes in a player simply because he moves to another city and puts on a different uniform. If Team A "doesn't travel well" today for any given reason, there is no apparent reason to me that the same team with different players will be affected in the same way 5-10 years later.

[PerpetualCzech]

It's the same with situational bettors..........they have no way of knowing if the lines/market has caught up and they no longer have an edge.

D.

Of course they do. There are very basic techniques of analyzing this. It simply involves just looking at what the lines were historically and comparing them to what they are today.

[Dreamer]

Ummm, isn't that true of any information? Where is Captain Obvious when we need him?


I know many will cry "soooo fucking obvious" .......yet..........I never see it mentioned in bulk analysis.

WOW............

Are you really this dumb?

Yes, of course you are.........

Stick to gambooling on craps with Wong..............

http://www.youtube.com/watch?v=6rdjPB-vzBY

D.

[Dreamer]

Of course they do. There are very basic techniques of analyzing this. It simply involves just looking at what the lines were historically and comparing them to what they are today.

Really?

The likes Of Bob and Kellen have virtually given up on situational betting in CFB..........."After" poor results.......... not before........

What happened to all those touting NFL home dogs in 2004 as money in the bank................oh!

As for "basic techniques" for analyzing if a line has changed from one period to another.........

Really.........LOL.

Read the next bit carefully................

There is nothing "basic" about having to compare the matchup strength of two teams, at a very specific POINT in the season in say week 7 of 2004 and in week 3 of 2009.......
Just having a basic boxscore database is not enough IMHO...........

But hey, what do I know?

D.

[PerpetualCzech]

There is nothing "basic" about having to compare the matchup strength of two teams, at a very specific POINT in the season in say week 7 of 2004 and in week 3 of 2009.......

This is not at all what is meant by "situational handicapping". It's just a specific analysis of a given matchup. Situational handicapping is identifying some kind of phenomenon at work over a large sample of games and then deciding whether the market knows about it or not. And yes, the techniques you can use to decide this can be trivially simple.

Suppose HFA in the NFL has been 2.5 points for the entire history of the league (I don't know whether this is true or not, it's just for example purposes). You go back and find that the Home Team was favoured by an average of 1.5 points in 1950, but since 1960 the average has consistently been 2.5 points. You can easily conclude that at some point between 1950 and 1960 the market learned the true size of HFA and reacted to it.

Saying there is "no way of knowing if the lines/market has caught up" is just not true, as this example shows.

[Rust]

Sure, but all you are doing here is simply describing HFA. We know that exists and that it is worth on average 2.5-2.7 points per game (depending on how far back you want to go). I might even accept that some players get affected by this phenomenon more than others and it affects their performance to different degrees.

What I find difficult to swallow is the theory that this quality changes in a player simply because he moves to another city and puts on a different uniform. If Team A "doesn't travel well" today for any given reason, there is no apparent reason to me that the same team with different players will be affected in the same way 5-10 years later.

What if we were to describe discipline as 50% innate and 50% as a result of organizational policy?

[Dreamer]

This is not at all what is meant by "situational handicapping". It's just a specific analysis of a given matchup. Situational handicapping is identifying some kind of phenomenon at work over a large sample of games and then deciding whether the market knows about it or not. And yes, the techniques you can use to decide this can be trivially simple.

Suppose HFA in the NFL has been 2.5 points for the entire history of the league (I don't know whether this is true or not, it's just for example purposes). You go back and find that the Home Team was favoured by an average of 1.5 points in 1950, but since 1960 the average has consistently been 2.5 points. You can easily conclude that at some point between 1950 and 1960 the market learned the true size of HFA and reacted to it.

Saying there is "no way of knowing if the lines/market has caught up" is just not true, as this example shows.

The ONLY way you KNOW if the market has caught up to a situation is if you have a method which can (reasonably) acurately predict the TRUE line of any given matchup compared to the current market line at the time of the bet............

So what if the AVG home fav changes from 1.5 to 2.5!

You ONLY know that the market line has changed.

What if they OVER adjusted in certain spots? How would you know?

Answer..... with simple bulk game analysis.......you don't....... untill you give up after a period of losing.

The old "it works till it doesn't"

Unless you have a predictive matchup strength model you cannot possibly tell if the TRUE line has changed over time in the same situation.

It's that simple.

The problem is the approach............

Datamining is fine up to a point to indentify market inefficiencies but you have to take it to the next level or 2

Ask not only WHY is that inefficiency created but how can I predict the true matchup line at the time of the game.

THEN create a matchup model that can predict that "true line" instead of the market line at the time of the matchup.

D.

[PerpetualCzech]

The ONLY way you KNOW if the market has caught up to a situation is if you have a method which can (reasonably) acurately predict the TRUE line of any given matchup compared to the current market line at the time of the bet............

You can repeat yourself and use all the caps you want, but until you introduce some evidence your conviction is little more than my old favourite, an argument from personal incredulity (http://en.wikipedia.org/wiki/Argument_from_ignorance).

For example, I have a favourite situational subset for 2H betting with data going back to 2002. It's not enough to bet alone but does add 3-4% to a bet if I happen to like it for other reasons. Here is the average line for a team belonging to this subset for the last 7 years (subset of 256 games per year):

2002: +0.2 points
2003: +0.2
2004: +0.4
2005: +0.3
2006: +0.4
2007: +0.3
2008: +0.3

Eyeballing that I can easily conclude that the market has not learned or adjusted for this situation the last 7 years. By monitoring lines week by week this year I can do similar analysis. I have many other subsets I monitor in the same way.

It's simply not the case that bettors "have no way of knowing if the lines/market has caught up". I know in this case the market has not caught up with as great a degree of confidence as anyone can have about any line they want to generate themselves. And I don't have to make my own line myself to know it.

[PerpetualCzech]

Just stumbled across this thread, can't believe I didn't remember it before posting my own (much less comprehensive) HFA thread the other week ...