Comments on Advanced NFL Stats: Full 2006 Season Now Available

JMM-Yes. There were several methods suggested, and...

2009-06-08T18:22:36.589-04:00

JMM-Yes. There were several methods suggested, and I've been testing some of them in various forms with the NBA and NHL games this year. What I've learned is that the different sports require their own method due to the nature of the scoring.

For example, a hockey game that's 1-1 would have very little total line movement, but would be relatively exciting. For hockey I use a measure of how 'balanced' the graph is during the game.

For football, I like the total movement of the line the best so far. There's nothing scientifically perfect about it--it just makes sense.

Great suggestion! I think it will be the one I end up using in the long run.

Regarding crossing the .50 line. I think that's a good idea too, but total movement will capture that on its own. As the line nears/passes .50, the importance of every play is magnified, so that the movement of the line will be higher.

A while back you said you were looking for ways to...

2009-06-08T17:00:19.895-04:00

A while back you said you were looking for ways to capture the excitement level of games. At the time, I suggested (what turned out to be) the length of the line.

After looking at several games I remember for being exciting, I also suggest "how many times the line crosses the 50% line" and the "game time of the last crossing of an arbitrary line- say 50%."

Those 3 seem to distinguish the exciting games, although in different ways.

Ian, I understand your point. I used the Det-NE as...

2009-06-05T07:49:29.120-04:00

Ian, I understand your point. I used the Det-NE as an exaggerated example, of course. My real point is that when you look back at real games, I think the current measure of WP tends to overstate the improbability of comebacks by good teams against bad teams. For example, last year the Lions were up 7-0 on Carolina at the end of the first quarter. Did Det really have an 85% chance of winning that game? I'd say no way.

Don't get me wrong, I think that WP is an interesting way to look back at games and compare comebacks. But, I think that it would be a better *predictor* of what's *going* to happen if it considered who the teams are. Going back to my example, no matter how the Lions managed to get that 4th quarter lead on New England, I think they'd be much less likely to hold onto that lead (given their horrible D) than, say, the Steelers, Eagles, Ravens, etc. Of course, if WP considered who the teams are, it wouldn't be 50% at the opening coin toss of every game.

Wayne You have to remember that the WP graph isn&...

2009-06-05T06:20:22.645-04:00

Wayne

You have to remember that the WP graph isn't just the odds of a team winning when put into a situation, but also accounts for how the rest of the game has already gone. i.e. if I said to Detroit "ok, you start 7-0 up with 15 minutes left to play" then yes, you'd expect New England to win. However, the WP says "having played 3 quarters of football against new england, the Lions lead by 7 and have a WP of X".

The 'lead' the Lions have in your situation isn't a headstart or a handicap, it's a lead they have built up over 3 quarters. Thus, you would expect that they could keep it up for one more quarter.

The reason the WP seems off is that the Lions rarely found themselves leading by 7 with one quarter to go, so you're trying to predict how a team would do based on a situation they barely faced.

Colts-Texans #1 last year had a 99% swing. In the ...

2009-06-04T20:14:44.140-04:00

Colts-Texans #1 last year had a 99% swing. In the last 8 minutes Indy's WP dipped below 5% 3 different times. Then Sage Rosenfels self destructed (with some help from a tremendous play from Robert Mathis and some killer Indy hurry up).

You may have already addressed this elsewhere, but...

2009-06-04T17:06:14.748-04:00

You may have already addressed this elsewhere, but what program do you use to obtain the current game status (downs, yds, score, time)? Is there a particular website that you get the "feed" from? I'm interested in doing WP updates for tennis, but I'm much more experienced with the mathematical programming than I am with programming as it relates to the web.

Wayne-You're correct. I use league-wide data f...

2009-06-04T11:22:01.882-04:00

Wayne-You're correct. I use league-wide data for the system. The method I use is almost completely empirical and relies on very large data samples. Dividing it up into smaller subsets for home/away and for various team strength levels thins out the data too much to be useful. But there are ways to make educated adjustments, which I hope to implement in the future.

Oops. Somehow I left out some the 2007 playoffs wh...

2009-06-04T10:30:01.320-04:00

Oops. Somehow I left out some the 2007 playoffs when I added 2006. I'll fix it this evening.

Thanks for posting these. It's kinda fun to lo...

2009-06-04T08:02:38.620-04:00

Thanks for posting these. It's kinda fun to look back at the graphs and remember some classic games. One question... as I understand it, WP calculations do not take into account who the teams are, correct? I'm not sure, but it didn't sound like WP considers who's playing at home, either. For example, imagine that the Lions are playing at New England. The Lions lead by 7 at the start of the 4th quarter. It seems like that would give Detroit over an 80% WP. But, IMO, their chance of winning would be much less than 80%.

miles-I sure can. That's the "CBF" s...

2009-06-03T21:11:40.933-04:00

miles-I sure can. That's the "CBF" stat down at the bottom. It's the lowest odds against the eventual winner. So if the lowest WP was about .01, the CBF would be about 100.

Wonder if you have a way to look for the biggest s...

2009-06-03T20:17:06.745-04:00

Wonder if you have a way to look for the biggest swings in the win probabilities. For instance -- in the 2007 dal-detroit game, it looks like the percentage touches 99% Detroit but quickly swings to 100% Dal.