Wednesday, December 7, 2011

Betting Market Power Rankings – Week 14

by Michael Beuoy

Here are the week 14 betting market power rankings. Here is a link to the first set of rankings I generated last week.

First off, thanks to Brian for making the Community site available as well as calling out last week’s post on the main blog. There was a lot of good feedback in the comments section. Most of it I’m still chewing over, but I have decided to incorporate a suggestion from Jim A, who has been creating a very similar set of rankings over at More on that later.


The original post has more detail, but here’s an overview:

The goal I had was to generate a set of point-based rankings that would best predict how the betting market would set the point spread for the coming week’s matchups. The better I was able to match the point spread, the better the rankings were a reflection of the market’s estimate of team by team strength. Through trial and error experimentation, I found that using point spreads for the most recent five weeks (with higher weighting given to more recent weeks), combined with an adjustment that accounted for actual game outcomes, generated the best predictive accuracy. More detail here.

The Rankings

Here are the Betting Market Power Rankings for Week 14:


LSTWK - The betting market rank as of the prior week (using the same methodology). It’s interesting to see who the big movers are.
GPF - Stands for Generic Points Favored. It’s what you would expect a team to be favored by against a league average opponent at a neutral site.
GWP - Stands for Generic Win Probability. I converted the GPF into a generic win probability using the following formula: GWP = 1/(1+exp(-GPF/7)). This gives a more direct comparison to the ANS rankings.
ANS RNK - The Advanced NFL Stats Team Efficiency rankings for the same week.
ANS GWP - The Advanced NFL Stats Generic Win Probability for the same week.

1 GB18.50.7730.71
2 NE28.50.7750.67
3 PIT47.50.7520.74
4 NO370.7340.69
5 BAL55.50.6970.62
6 SF840.64110.54
7 ATL63.50.62120.54
8 NYJ93.50.62130.54
9 DAL730.6180.61
10 TEX1010.5410.8
11 NYG1210.5360.65
12 MIA1510.53170.5
13 SD180.50.53160.5
14 CIN160.50.52180.48
15 DET130.50.5290.61
16 PHI110.50.51100.59
17 TEN19-0.50.49200.44
18 RAI17-0.50.48140.53
19 CHI14-10.47150.52
20 CAR20-1.50.45220.39
21 DEN22-1.50.45230.39
22 SEA25-2.50.41250.37
23 WAS23-30.39210.44
24 BUF21-30.39190.48
25 ARZ27-3.50.38240.37
26 MIN24-40.36280.33
27 KC30-40.36310.3
28 TB26-4.50.34290.33
29 CLE29-50.33260.36
30 JAC28-6.50.28270.36
31 STL31-6.50.28300.32
32 IND32-8.50.23320.27

Some observations:

• San Francisco moved up a couple spots on the strength of their convincing victory over St. Louis.
• Philadelphia continued their nosedive, dropping from 11th to 16th.
• San Diego moved in the opposite direction, climbing 5 spots to 13th.
• If you throw out Houston (the ANS model doesn’t know Matt Schaub is injured), the ANS top 4 are in line with the betting market top 4: Green Bay, New England, Pittsburgh, and New Orleans.
• Count the nation’s gamblers among the Doubting Thomases to the miracles of Tebow. Denver only moved up one spot to 21.
• My Colts are now only buried under 25 feet of crap (2 points instead of last week’s 4). Progress!

Predicted of This Week’s Point Spreads

See below for how well the ranking methodology predicted this week’s point spreads (that SF/ARZ spread puzzles me):

GamePred LineActLineDiff
 ATL @ CAR -1-2.5-1.5
 BUF @ SD 5.571.5
 CHI @ DEN 2.53.51
 CLE @ PIT 1514-1
 IND @ BAL 1716.5-0.5
 KC @ NYJ 10.59-1.5
 MIN @ DET 77.50.5
 NE @ WAS -10-82
 NO @ TEN -6-3.52.5
 NYG @ DAL 53.5-1.5
 PHI @ MIA 43-1
 RAI @ GB 12.511-1.5
 SF @ ARZ -6.5-3.53
 STL @ SEA 572
 TB @ JAC 000
 TEX @ CIN 1.531.5

Proposed Methodology Change

In the comments, Jim A mentioned that he uses point spreads for future weeks as well (since these are available more than a week in advance in a lot of cases). I think this makes sense and have recalculated the rankings using this approach. Under the new approach, I gave both this week and the next a weight of 4, and prior weeks descending weights of 3, 2, and 1. The model seemed to work best with 5 weeks of “interconnectedness”, so the new approach keeps 5 weeks of data, but now has the advantage of being able to give more weight to the latest betting market information, as opposed to stale information from prior weeks. Here is how the new rankings compare to the rankings above:

RankTeamGPFGPF NewDiff
1 GB8.590.5
2 NE8.57.5-1
3 PIT7.57-0.5
4 NO77.50.5
5 BAL5.560.5
6 SF451
7 ATL3.53.50
8 NYJ3.52.5-1
9 DAL330
10 TEX10.5-0.5
11 NYG11.50.5
12 MIA110
13 SD0.50.50
14 CIN0.50-0.5
15 DET0.510.5
16 PHI0.50-0.5
17 TEN-0.500.5
18 RAI-0.5-0.50
19 CHI-1-1.5-0.5
20 CAR-1.5-10.5
21 DEN-1.5-10.5
22 SEA-2.5-3-0.5
23 WAS-3-4-1
24 BUF-3-30
25 ARZ-3.5-2.51
26 MIN-4-3.50.5
27 KC-4-4.5-0.5
28 TB-4.5-4.50
29 CLE-5-50
30 JAC-6.5-7-0.5
31 STL-6.5-7-0.5
32 IND-8.5-8.50

The biggest change is clear separation of Green Bay and New England, although the ranks remain the same at 1 and 2. For future weeks, I plan on using this approach, so the “Last Week” column for next week’s rankings will reflect this new approach, rather than the ranks in the first table above.

Future Investigations

Another source of insight into the betting market is the Superbowl futures (also pointed out by Jim A). Unfortunately, these are not only just a function of inherent team strength, but also a reflection of how difficult a team’s path may be to get to the Superbowl. If they’re in a strong conference, or if they’ve underachieved thus far in regards to wins and losses, this will reflect poorly on their Superbowl odds.

To get around this difficulty, I was thinking I could use Chris Cox's ridiculously awesome NFL Forecast tool. I could plug in the betting market GWP’s into the tool (I believe it allows for that type of customization), and then use those to generate Superbowl probabilities for each team. These could then be compared to the probabilities implied by the Superbowl futures odds.


Jeff Fogle said...

On the SF/AZ line you said was puzzling. Two things in play I think...

*Kolb is back, and the market respects him more than Skelton...which is why last week's like with Dallas moved down once it was clear Kolb was going to play (SF is just 1 point better than Dallas in your initial it's kind of a replay scenario).

*Things get tricky in the area right above the key number of 3. That's a common result, and the money tends to come in toward the three if a line is nearby (a 2.5-point favorite will get bet to -3...a 3.5 point favorite will see money come in on the dog). When that DOESN'T's telling you that respected money (based on respected ratings) are forcing things the other way. You see this above with both N.O. and S.F. in your data...your ratings have them further from their opponents than the market...but the market is sitting on a 3.5...which is A LOT higher than -3 in terms of differentiating teams market-wise.

I used to be involved with a project a few years ago at a wagering forum where we were putting together "community power ratings" amongst the field that we would then compare to the market prices (using the wisdom of a sharp crowd to see if we could outdo the wisdom of the full crowd...worked well the first year, fairly generic the second year). Similar things would happen in this regard. Gambling author King Yao came in and encouraged us to use win percentages off the moneylines rather than integers. I'm too old school and I can only think in power ratings...but Yao's approach helps dodge this little headache.

Something for you guys to think about maybe since BB's work already speaks the language of win percentages. Try estimating/representing the market win percentages off the moneylines as a way to guage market rankings rather than poinspreads. Just a thought.

Michael Beuoy said...

Thanks Jeff. I forgot about Kolb coming back.

Unfortunately, I don't have easily available access to moneyline odds, so I'm not sure how feasible that is.

Horseracing said...

Great ideas that I have never thought of-thank you!

Ed Anthony said...

Here's a great site for money odds. They also have historical data include money lines, spreads and over/under numbers.

Jeff Fogle said...

Hunted around, and it looks like Ed Anthony's guidance is the best in terms of finding past moneylines. You see this week's on that link he provided...then scroll down a bit and there will be links for Week 13, Week 12, etc...going back. Also noticed that on THOSE links...they'll show you Week 13 from last year, 2009 etc...if you scroll a very good historical tool. Thanks Ed!

A lot of places (covers, vegas insider, don best) will show you this week's numbers. Past ones are tougher to it's very handy to have that spot ed tipped us to.

If you don't have time MB, would you like one of us to calculate what the numbers were for the last five weeks? Actually, I posted last week's Sunday morning ones in a comments section here somewhere...and I can throw down this week's here. Wouldn't take too long to figure the prior few weeks. Not sure if your machinery would be ready yet to absorb those...but some of us can assist in the tracking down process at least.

From high to low (BB's in parenthesis)

Baltimore 93% over Indy (86%)
Green Bay 86% over Oakland (75%)
NY Jets 85% over KC (78%)
Detroit 84% over Minnesota (81%)
New England 79% over Washington (66%)
San Diego 74% over Buffalo (60%)
Seattle 73% over St. Louis (63%)
New Orleans 65% over Tennessee (67%)
SF 65% over Arizona (60%)
Denver 64% over Chicago (55%)
Dallas 64% over NYG (54%)
Miami 61% over Philadelphia (Phi 52%)
Cincinnati 58% over Houston (53%)
Atlanta 57% over Carolina (same)
Tampa Bay 55% over Jax (Jax 61%)

Used BB's replacement numbers for Houston/Chicago from his NYT piece.

For people wanting to figure these on their own if the prices change over the weekend, I'm using the no-juice moneyline (split the difference between favorite and dog prices), then posting the break-even win-percentage that the no-juice price represents (that number...divided by the sum of that number plus 100). A -200 favorite has to hit 67% to break even for an easy example to visualize.

Daniel B said...

How much have you guys found that big or small fanbases (which I assume bet more often for then against their own team) affect the betting line vs what the line 'ought' to be if all humans were dispassionate?

Anonymous said...

You are calculating the no- vig line wrong. You don't split the difference unless it is even money.

Michael Beuoy said...

Thanks Ed for the link. It may take some time to scrape the moneyline numbers from that site. Although if there's one thing I've become an expert at, it's scraping data from websites (it's a necessity if you want to do any work with sports stats). However, I notice that for future weeks, all they have are point spreads, no moneylines. Given that that future week provides a lot of valuable info, I'll probably stick with using point spreads.

Jeff - It's a fair point that all points aren't created equally. I was thinking I could come up with some sort of transformation of the point spread to a metric that better represents win probability. Such a transformation would probably have a bigger jump from 3 to 3.5 than from 2.5 to 3. In other words, I can come up with an implied moneyline from the spread.

Daniel B - I recall seeing some sort of study on that a while back, I'll see if I can dig it up. I wonder if the bias is stronger for moneyline, rather than spread bets. At least with a spread bet, you could bet against your team, but still want them to win (just not win by too much).

Ed Anthony said...

Future odds are difficult to obtain. No one wants to commit too far into the future as bets will get placed and then injuries affect outcomes.

Here is another site where you can get some futures. This is week 15. The board is incomplete so it may not be of any use for what we're trying to do here. If you scroll to the bottom of the page you can select other weeks.

Jeff Fogle said...

anon, what's your suggestion for calculating the no-vig moneyline in Oakland/Green Bay for example? As I write this, Green Bay is -800, and Oakland is +600 at the Las Vegas Hilton. What's the best way to calculate the no-vig moneyline off that, and then convert it to a projected win percentage in your view?

Jeff Fogle said...

Sorry, forgot to answer Daniel B's question. The Vegas and offshore markets tend not to be influenced by local fanbases. Maybe individual bookmakers in certain regions are with their local teams (an illegal bookmaker in NY may have to inflate the line a bit in Giants and Jets games for example, or a guy in Texas might have to do that with the Cowboys).

Vegas and offshore oddsmakers have made it very clear that their biggest task is defending against the public's tendency to bet favorites the vast majority of the time. There's a myth about "splitting the action" out there because it makes sense on paper...but the real world dynamic just doesn't allow for it.

Imagine public money representing a mountain, and "smart money" from professional wagerers representing a hill. Now, imagine that the mountain bets almost exclusively favorites. The mountain looks at the card and says "What favorites should I bet on based on those lines?" rather than looking for true line value.

Given that real world dynamic, oddsmakers try to charge a tax on favorites that gives them an additional edge over time to the 11/10 they already enjoy. But, NFL action is so heavy that a half-point tax is enough to give them the best of it. If they charge too big a tax, betting goes way down (the mountain just isn't going to bet dogs!). If they don't charge any tax, they're way oversided on games and they're not maximizing their possible edge. The sweet spot in the NFL they tell me is just a half a point or so...and some creative movement around the critical numbers of 3-7-10, etc...

So, even though there is some adjustment for public influence in the market lines, those market lines still end up being the best known predictor of results. All studies that I know of have shown this.

Note that other sports will see slightly bigger taxes. College football may see as much as 3-4 points added to the very tall lines when a power is playing a patsy. College hoops will have as much as 1-2 points in there as well on the TV games (the mountain tends not to bet small conference non-TV games, so those lines are more on the money...hoping to hold their own against the most serious wagerers). The NBA isn't as widely bet as the other sports, but will have a tax against teams like the Lakers or Heat (the mountain loves betting on Kobe and LeBron).

So, anyway, in the NFL...the public's tendency to bet favorites is in the mix...but any pollution has yet to push the numbers away from a point where they offer the best predictive value.

And, if you hear somebody talk about "the oddsmakers job is to split the action," please let them know that this isn't something that's feasible very often in the real world. Oddsmakers try to make the public take the worst of it by charging them 11/10 on their losses AND by making them lay at least a half a point more than they should be. Over time, those edges really add up for the sportsbooks.

Those of you who have tried being small bookmakers amongst your friends have seen this play out...or, if you're in college...ask your fraternity bookie how his clients bet. It's just the nature of the public to bet favorites. Smart money wagerers take up some of the slack by betting underdogs because that extra half point or point is in their favor. The 11/10 isn't though...

Anonymous said...

Jeff's comparison between the Vegas winning percentages and BB's is interesting to me. As you can see, if you were betting money lines solely on BB's percentages, you would be betting almost all dogs this week (13 out of 14, one game has no difference between the Vegas and BB winning percentages). I think this tendency has also been present in other weeks; BB's weekly numbers typically show much greater chances for upsets when compared against a non-vig Vegas money lines. Jason

Jeff Fogle said...

Agree with you about that Jason. And, normally that's a good sign for a methodology. Maybe it will be today as well.

Last week's projections, in my view, didn't fare well compared to what the market was saying about those games. Kind of depends on how one decides to grade them. Just in terms of picturing which difference better captured how the game ultimately played out...

*He had Indy closer to NE than the market did (91% to 95%), and that ended up being a closer game than expected after garbage time.

*He had NYG closer to GB than the market did (57% to 71%) which proved prescient, and helps make up for the 51% or 52% assessment at Detroit on Turkey Day when the market had that one much higher.

*He had Houston beating Atlanta even after adjusting for the backup QB, while the market had Atlanta.

Those are the only three real highlights though. Going the other way:

*Philadelphia was 70% over Seattle in a game they lost badly (outgained 6.7 to 5.9 ypp while losing turnovers 4-0).

*He had Buffalo at 61% over Tennessee compared to 52% for the market...and Tennessee had the better of it throughout the game.

*He had Miami at only 51% over Oakland, compared to 62% for the market. Miami was in control the whole way.

*He only had Baltimore at 64% over Cleveland, when the market had 74%. Baltimore was in control the whole way, and led by 21 until a very late score.

*He only had the Jets at 51% in Washington, while the market had 60%. Jets won by 15.

*He only had NO at 66% over Detroit, while the market had 78%. New Orleans led 24-7 at the half and ultimately won by 14.

*He only had SF at 74% over St. Louis while the market had 89%. St. Louis never scored, and was just squashed. Tough to suggest a team with a horrible offense is going to win on the road one time out of four with against a top defense. Market had it more like 1 out of 10.

*He had Dallas up at 73% vs. Arizona, while the market was lower at 65%. Arizona won in overtime in a game that played out very close.

*He had Jacksonville at 52% over San Diego. The market had SD at 60%, and they were up 24-14 at the half on the way to a stat blowout.

(I didn't mention Carolina/TB because TB's had a late QB scratch, or Cincy/Pitt, KC/Chicago and Denver/Minnesota because BB and the market were very close there).

Maybe it's a good week for dogs and things float back toward equality in this late season sampling today. Destined to be an exciting Sunday if BB's conservatism is correct...

Post a Comment

Note: Only a member of this blog may post a comment.