Degeneracy Instructions for Bodog Lines

I have dim opinions of the abilities of various betting agencies and/or their customers to set prices on hockey; but it is also common to say that making money consistently betting on hockey is hard. Purely for I-told-you-so purposes, I record here how I would bet if I were a gambling man given the posted lines on 2015-09-11, a few days after the Bodog lines came out. Not one of the lines is close enough to my model predictions to not give some value. My scheme is as follows: bet some fixed multiple of the 'value' entry for all thirty teams. That is, bet, say, 47 dollars on the under for Anaheim, 47 dollars on the over for the Coyotes, 32 dollars on the over on Columbus, and so on.

My intuition about where the model is good and bad is detailed in the previous post, but I very strongly believe that if one is to bet based on models, one should not permit oneself to make adjustments based on perceived model weaknesses. Model weaknesses should be fixed by altering the models construction, not the model outputs, ad hoc.

The Bodog odds are 53.5% for both over and under, which is, in passing, a shameful robbery of the degenerate public.

Team	Model Points	Model Std. Dev.	Bodog Line	Z-score of Line	Prob. of over	Prob. of under	Value in over	Value in under
Anaheim Ducks	94.7	8.3	105.5	1.29	0.0	1.0		0.47
Arizona Coyotes	91.7	8.3	80.5	-1.35	1.0	0.0	0.47
Boston Bruins	100.7	8.2	109.5	1.07	0.0	1.0		0.47
Buffalo Sabres	74.6	8.2	65.5	-1.11	1.0	0.0	0.47
Calgary Flames	87.5	8.3	72.5	-1.8	1.0	0.0	0.47
Carolina Hurricanes	88.5	8.3	81.5	-0.84	0.99	0.01	0.46
Chicago Blackhawks	105.4	8.1	111.5	0.76	0.02	0.98		0.45
Colorado Avalanche	85.5	8.3	95.5	1.19	0.0	1.0		0.47
Columbus Blue Jackets	93.5	8.3	90.5	-0.36	0.85	0.15	0.32
Dallas Stars	94.2	8.3	93.5	-0.09	0.6	0.4	0.07
Detroit Red Wings	97.1	8.2	91.5	-0.68	0.97	0.03	0.44
Edmonton Oilers	79.2	8.3	80.5	0.16	0.32	0.68		0.15
Florida Panthers	91.6	8.3	74.5	-2.05	1.0	0.0	0.47
Los Angeles Kings	106.8	8.0	106.5	-0.04	0.55	0.45	0.02
Minnesota Wild	90.9	8.4	98.5	0.91	0.0	1.0		0.47
Montreal Canadiens	89.0	8.3	93.5	0.54	0.06	0.94		0.41
Nashville Predators	90.9	8.3	80.5	-1.25	1.0	0.0	0.47
New Jersey Devils	100.6	8.2	86.5	-1.72	1.0	0.0	0.47
New York Islanders	89.3	8.3	87.5	-0.22	0.73	0.27	0.2
New York Rangers	96.8	8.3	95.5	-0.16	0.68	0.32	0.15
Ottawa Senators	91.2	8.3	79.5	-1.41	1.0	0.0	0.47
Philadelphia Flyers	91.3	8.3	88.5	-0.34	0.84	0.16	0.31
Pittsburgh Penguins	92.2	8.3	105.5	1.6	0.0	1.0		0.47
San Jose Sharks	101.6	8.2	102.5	0.11	0.38	0.62		0.09
St. Louis Blues	101.3	8.2	105.5	0.52	0.07	0.93		0.4
Tampa Bay Lightning	95.9	8.3	95.5	-0.05	0.56	0.44	0.03
Toronto Maple Leafs	72.3	8.2	88.5	1.99	0.0	1.0		0.47
Vancouver Canucks	95.5	8.3	88.5	-0.85	0.99	0.01	0.46
Washington Capitals	84.9	8.3	88.5	0.43	0.1	0.9		0.37
Winnipeg Jets	92.7	8.3	80.5	-1.47	1.0	0.0	0.47

Start of season predictions for 2014-2015

I've finalized the 2014-2015 prediction model (which I'll describe in a later post) but for today I've taken a hundred thousand simulations of the season and obtained my start-of-season predictions. Keep in mind that these predictions do not take any account of off-season moves, so they should be interpreted through that lens -- if you think a given team had an especially good or bad off-season, you should mentally bump them up or down as appropriate. In broad strokes, my own judgment agrees with the model (I trained it, after all) but after each division I'll give "my own" comments on the model predictions.

How to read these charts

The four green shades show the chances of making the playoffs in first through fourth in the division. The four red shades (three in the west) show the chances of missing the playoffs in fifth through eighth (seventh). The yellow bar shows the chance of making the playoffs in fifth; the orange bar shows the chance of missing the playoffs in fourth.
The large number is the predicted number of standings points. The smaller numbers above and below show one standard deviation each.
For example, the model predicts that the Bruins have around an 85-90% chance of making the playoffs; which breaks down as about 40% of finishing first in the Atlantic; around 25% to finish second, about 15% for third, about 10% of making the playoffs in fourth or fifth, and around a 10% chance of missing the playoffs altogether.

Atlantic

• The Leafs were extremely fortunate to be in the playoff conversation at any point last year, and while I think they've improved a little on-ice (and a lot off-ice), they'll remain terrible as their goaltending regresses and their shot suppression remains woeful, at least until they fire Carlyle (Christmas-ish?) at which point they might get a little better.
• The Sabres, who were abysmal last year, may actually have gotten (deliberately) worse; they are "rebuilding" the old-fashioned way, that is, burning their team to the ground and amassing a stockpile of draft picks the likes of which I've never seen. Expect McDavid.
• Habs possession numbers were middling this year and I think they'll drop back like the model suggests, although perhaps not quite as far as sixth.
• Senators predicted to be the bubble team that I think they are; most commentators seem to think they've taken a big step back this off-season but I don't see it. Could be surprising, could be unnotable also-rans.
• Panthers have surprisingly good possession numbers and now also Luongo and Ekblad. I suspect they'll be on the outside when the playoffs roll around but I expect them to be a nuisance to good teams and not an easy two points a-la Buffalo
• Lightning won't get Vezina goaltending from Bishop again but they are a very, very good team and deserve every bit of the third-place prediction. Also, they'll have Drouin this year, in case anybody forgot.
• Red Wings probably not as good as their numbers suggested last year as their best players are starting to suffer heavily from age regression. Probably a bubble team.
• Bruins had a bit of a complacent off-season and are in cap trouble now. Probably still enough raw talent to win the division as predicted.

Metropolitan

• Devils were incredibly unlucky last year relative to their possession. Their horrific shootout luck will go away and Schneider replacing Brodeur in goal is an enormous upgrade. I won't be at all surprised if they win the division as modelled.
• Rangers remain good, second seems about right.
• Bluejackets also not a flash in the pan; plenty of good players at every position. Third seems fine.
• Penguins have several extraordinary players but poor possession numbers, stemming largely from a dreadful bottom six. Teams like Toronto (!) this off-season have demonstrated how easy it is to find cheap, good bottom six players every single summer; it's frankly an unforgivable mistake from the management in Pittsburgh. I lean to the high side of the prediction but not by lots.
• Flyers seem like a bubble team; forwards are good and improving but defence is old and goal very suspect. Probably out for mine.
• Islanders had an amazing off-season and I expect them to be a huge pain. The logjam in the middle of the Metropolitan could be lengthy again.
• Hurricanes have gotten a little worse this offseason, I think they'll probably be squeezed out fairly early and will start trading what they can for picks.
• Capitals are, frankly, dreadful, and no quantity of Ovechkin can stop it. Last with a bullet.

Central

• Blackhawks are great and everybody knows it, including my model.
• Blues are also great but not as great as the Blackhawks. Everybody knows this and so does the model.
• Stars were surprising last year, they will surprise nobody this year. A fully healthy Spezza (doubtful...) makes them a cup contender; I put them in a knife-fight with the first two for the division.
• Jets have a lot of underrated players and their success will largely hinge on player usage. A severe-ish injury to Ondrej Pavelec would probably propel them into the playoffs, an age regression and a lot of Pavelec starts probably put them in the McDavid sweepstakes. I don't mind the "bubble-team" prediction.
• Minnesota with a quiet off-season after a lucky season. They'll win a lot of games but miss out in the end.
• Predators will have a new coach this year for the first time ever and I suspect it'll be rough. They'll need a pretty heft overhaul before they're really good but they have too many good defensive pieces to lose many games. Your favourite team will lose to them and you'll wonder why.
• Colorado probably the biggest against-the-grain prediction here. They can expect many things to not go as well for them this year: their goaltending will regress, their shooting percentages will regress, MacKinnon will get the attention he deserves, Roy's (generally good) strategy of pulling his goalie early will not work quite as well; most of all, they will be deeply complacent after being the luckiest team in hockey last year. Prediction seems solid to me.

Pacific

• Kings are great and everybody knows it.
• Sharks are great and everybody except the Sharks themselves know it.
• Canucks were unlucky last year and are not the force they once were but will do reasonably well this year. Bubble team.
• Anaheim as a team were very fortunate in shooting percentages; that should run out. Also a bubble team.
• The newly minted Arizona Coyotes remain middling, surprising nobody.
• The Flames have more talent than people realize but it's still not much. Brian Burke still ruining hearts and brains.
• The Oilers young players are probably better than in previous years, but the organization is still dysfunctional. Probably won't be last by quite as much as the model suggests.

Recap of 2013-2014 predictions

Using a model trained on 2009-2012, I made the following predictions for the finishing places in 2013-2014. Before the next post (predictions for 2014-2015), let's look at how the old model did.
Let's look at these in order of how they actually finished:

• 117 Boston: Finished with an easy first-place, as predicted. Bang on.
• 101 Tampa Bay: Just edged out Montreal for second, predicted to be 5th-6th. Badly wrong.
• 100 Montreal: Third place as predicted. Solid.
• 93 Detroit: Well-ensconced in fourth; predicted to be top three. Iffy.
• 88 Ottawa: Fifth with space on either side; predicted to make the playoffs fairly easily. Mostly wrong.
• 84 Toronto: Clearly in sixth; predicted in 7th or 8th. Mostly right.
• 66 Florida: Well back in seventh; predicted to be a spot or two higher. Mostly right.
• 52 Buffalo: Deep in the basement at 8th, as predicted.

• 109 Pittsburgh: Alone at the top. Predicted to be a few spots lower. More right than wrong.
• 96 Rangers: In second as predicted. Exactly right.
• 94 Philadelphia: Predicted as near the bottom. Badly wrong.
• 93 Columbus: Predicted at eighth, badly wrong. (Note for future: whoever wins the Vezina every year is going to play for a team that finishes better than predicted.)
• 90 Washington: Tightly packed in at fifth/sixth, predicted at sixth. Close as makes no odds.
• 88 New Jersey: Just behind Washington at fifth/sixth, predicted to win the division with an enormous lead. Shootout record was outlandishly unlucky but the prediction is still hideously wrong. Dramatically bad goaltending the likes of which sane teams do not tolerate was another issue.
• 83 Carolina: Finished in seventh but predicted at fifth. Not bad but not great.
• 79 Islanders: Alone at the bottom but predicted in third/fourth. Wrong.

• 112 Colorado: Just edged out the Blues for top spot. Another team with a Vezina candidate outperforms the prediction; which in this case was "lowish bubble team". Badly wrong.
• 111 St Louis: Predicted as close second. Right on the money.
• 107 Chicago: Finished third of the trio way out ahead of the other five. Predicted to do better. Mostly right.
• 98 Minnesota Completely alone in fourth, as predicted. Spot on.
• 91 Dallas: Finished in fifth, predicted to be level for sixth/seventh. Mostly right.
• 88 Nashville: Predicted in sixth/seventh, wound up in sixth. Pretty much spot on.
• 84 Winnipeg: Finished at the bottom but predicted in third/fourth. This is mostly wrong, but in my defence, the Jets hugely underused their own personnel by playing Byfuglien away from D and giving the worst starter in the league many starts.

• 116 Anaheim: Finished in reasonable comfort atop the pacific. Predicted to be a distant fifth. A bad mistake.
• 111 San Jose: Finished second as predicted. Bang on.
• 100 Los Angeles: Finished alone in third; Predicted to run away with the division. More right than wrong.
• 89 Phoenix Finished fourth, predicted to edge out Vancouver for third/fourth. Basically right.
• 83 Vancouver Predicted to come in fourth, came fifth, although the model suggested a 75% of the playoffs and they missed. Right-ish.
• 77 Calgary: Finished alone in sixth, exactly as predicted.
• 67 Edmonton: Finished at the bottom, just as predicted.

On the whole, not bad. Predictions for next year will contain means and standard deviations for standings point, so judging next year will be a little more quantitative.