Comments on Advanced NFL Stats: In-Game Home Field Advantage

I know this is old, but I hope someone responds. ...

2009-11-18T16:32:27.037-05:00

I know this is old, but I hope someone responds. I just read the 2007 article about why passing is relatively more important for road teams. One theory offered up, but not defended, was that road teams have to come from behind more, a situation in which passing is more useful.
Could this large home-team 1st half advantage explain this phenomenon? Visiting teams dispraportionately fall behind early, and thus need good passing games to catch up.

Hi Brian, You can further demonstrate the importan...

2009-07-14T10:03:07.775-04:00

Hi Brian,
You can further demonstrate the importance of the game position by looking at the proportion of points accrued by home favourites in the final two quarters in different situations.

When tied at the half home favs average 59% of the 3rd quarter points and 57.4% in the 4th.(On average they get ahead in the 3rd and they keep the pressure on in the 4th).

When leading by a TD or more at the half they average 56.8% of 3rd quarter points and 54% of 4th.

Lead by 2 or more TDs and they get 55.5% of 3rd quarter points and just 52% of 4th.(They keep their opponents at arms length in the 3rd and coast through the 4th).

So the extent of their average supremacy and the amount that it tails off appears to depend on the size of their half time lead and you get the decline in scoring per quarter that's observed in a league wide sample.

However,if instead they trail by one or more TDs at the half home favs average 58% of 3rd quarter points and they up this to 59.8% in the 4th.

Trail by 2 or more TDs and they score 59.6% of 3rd quarter points and 62.1% in the 4th.(They try to assert their superiority in the 3rd and such was the size of the half time deficit they have to up their game even more in the 4th to either protect or create a lead).

And we see a reversal of the more commonly seen scoring trend by quarter.

Interesting. Thanks, Denis. My 2nd graph in the li...

2009-07-14T06:51:50.603-04:00

Interesting. Thanks, Denis. My 2nd graph in the linked-to article bears this out. The effect is only present when home teams are behind.

If you look just at games where the home side goes...

2009-07-14T06:16:56.632-04:00

If you look just at games where the home side goes into the game as the underdog,you don't find this effect.

Since 1989,home underdogs generally score 45% of the points in the first quarter and 46% of points in the fourth.So they actually do best in the 4th.

Divisional home dogs show an even more extreme split.They gather 47.3% of the 4th quarter points,but only 43% of first quarter ones.

Home favourites do show the progression you report.They score 60.5% of 1st quarter points,58.3% of 2nd,56.9% of 3rd and 55.3% of 4th.

This means when you look at the league as a whole the progression exhibited by the more numerous home favourites swamps that shown by the home underdogs and it appears that home advantage declines with time.

I think the effect is most likely driven by game position.The better side more often aggressively gathers points and then moves to a more passive clock control mode.

To say that hfa declines throughout the game by looking at points scored/allowed you have to be sure that both teams are playing at the same competitive level through all four quarters.....and you carn't always be sure of that.

If you look at games that you expect to be low scoring,there's a reduced chance that the better side can build up a lead they're comfortable with.Therefore,its less likely that one team will play a passive role late in the game and in these type of matchups the proportion of points scored by home and away sides remains relatively constant over all 4 quarters.

Dave-I agree with what you say, but a generalized ...

2009-07-13T13:34:09.222-04:00

Dave-I agree with what you say, but a generalized model of NFL football is very useful in many ways.

Yeah, that's true. I guess we'd need the s...

2009-07-13T13:17:20.128-04:00

Yeah, that's true. I guess we'd need the same regression but with points instead of win probability as the dependent variable.

But doesn't Doug's third post just show th...

2009-07-13T13:12:54.496-04:00

But doesn't Doug's third post just show that the less time you have left in the game, the less time the HFA has to come out? That's already known: if football games were 600 minutes instead of 60, HFA would be much higher. So if you're talking about (the last) 15 minutes instead of 60, the observed effect of HFA should be lower.

The inherent problem with this analysis is it only...

2009-07-13T13:10:18.098-04:00

The inherent problem with this analysis is it only applies if the teams are even. The 56.5 chance of winning is the avg. over the whole league. Therefore if 2 even teams played the one at home would therefore have 56.5% chance of winning. You have to incorporate the % chance of winning for each team to have accurate reading in one particular game. You can get this from the odds (subtracting out juice) or your model and then factor that into each individual game.

These two posts at PFR have some interesting thing...

2009-07-13T12:52:39.961-04:00

These two posts at PFR have some interesting things about how the decreasing HFA effect may be due to environmental familiarity.

http://www.pro-football-reference.com/blog/?p=1208

http://www.pro-football-reference.com/blog/?p=1227

But this third post might be what ultimately convinces. In it, Doug Drinen has a regression of team win probability holding for team strength. Holding all else equal, the coefficient for home field diminishes for each quarter.

http://www.pro-football-reference.com/blog/?p=691

One of the reasons I have a lot of faith in that post, and what it says about HFA, is that the HFA coefficient was incidental to the point of his study. Doug wasn't really investigating HFA at all.

In any case, the main point of my own post here was to figure out how to empirically model HFA as it's observed. I can't say that these results confirm any of the speculation about why it happens.

Yes, that's the one reason I'm confident--...

2009-07-13T12:38:39.298-04:00

Yes, that's the one reason I'm confident--the effect appears to hold even when the score is tied. But, as you point out, a home team that's tied in the 4th quarter is probably the weaker team. I wonder how big an effect that would have. There is so much random noise in any one game (good teams can be tied with bad teams, and vice-versa).

I wonder if this effect also holds in baseball ......

2009-07-13T11:30:21.933-04:00

I wonder if this effect also holds in baseball ... the evidence in MLB is that there's no such thing as clutch hitting, so if there's a natural tendency for home and road teams to converge late in the game, teams should score a more equal number of runs in the 8th inning than (say) in the 4th. You'd have to adjust for lineup, though.

Actually, if you averaged all the teams, lineups would probably even out.

I read the other article, and I'm not sure how...

2009-07-13T11:22:20.875-04:00

I read the other article, and I'm not sure how you can be sure that HFA decreases as the game goes on. Couldn't it be that teams change their style of play based on the score? You do include a chart for when teams are tied, but, as you say, a home team that's tied in the fourth quarter is probably the worse team ... and the fourth quarter is always weird in other ways.

Yes. Check out the article linked-to at the top of...

2009-07-13T09:20:30.737-04:00

Yes. Check out the article linked-to at the top of this post. The same effect exists in the NBA. I believe the Pro-Football-Reference guys independently found the same thing in the NFL.

Are you sure this is right? It seems like HFA woul...

2009-07-13T09:05:14.570-04:00

Are you sure this is right? It seems like HFA would have a near-constant effect on scoring likelihood, but the ratio of the information it provides to other information would decrease as the game goes on (0-0 before kickoff tells you nothing, 37-3 with 2 minutes left in the 4th tells you everything), thus causing the decrease.