Nba nerd

2 Games, $1400

I thought I was done writing about Vegas, but a funny thing happened. Two season-long over-under bets have come down to each team's final game of the season (and, for the record, it doesn't look good for me in either game):

Game 1: Indiana vs. Orlando

If Orlando beats Indiana, my under bet on the Pacers will push at 55 wins, meaning this game is essentially a $600 game for me. And frankly, this makes me feel more than a little bitter. We Box Score Geeks will insist that the last 6 weeks have been regression to the mean, and that during the first 20 games of the season, Indiana was a perfect storm of "everybody defies age, has a career year, or overachieves at once", and won a few more games than they "should have". Meanwhile Pacers' fans (who were pretty damn vocal at the beginning of the season but have strangely been pretty silent in the comments lately) will insist that the beginning of the season was the "real" Pacers team, and the most recent stretch is the actual outlier.

Here's what I am hoping:

  • With any luck, since Indiana can no longer lose the first seed, Paul George, Lance Stephenson, George Hill, David West, and Roy Hibbert spend tomorrow on the bench in nice suits. Unfortunately, I think Indiana feels too much like they must prove something to do this, but a man can hope.
  • I hope the Pacers run hot in the playoffs and goad Vegas into setting the over/under at about 57 next year.

I think there is a real possibility the over will be too high for this team next year, but it will depend on the offseason, of course.

Game 2: Utah vs. Minnesota

If Utah wins, I push the $800 over bet on 25 wins.

Frankly, I'm ticked off at Utah right now. After beating the Sixers on March 8th, they had 22 wins. So with 19 games to go, I felt good about them winning 3 (for a push) or 4 (to win). They promptly went 2-16.

You Had ONE JOB

One ****ing job.

I had to turn off the Utah game against the Lakers last night. They displayed a level of apathy/distractedness that would make a mid-February Sixers fan cringe in agony. And now I'm left hoping they will beat Minnesota to push. And Minnesota, in a stroke of just brilliant ****ing luck for me, sits at 40-41. Think Minnesota really wants to win and finish 0.500? And Minnesota couldn't beat Sacramento last week to prevent this deathmatch from happening!?

Vegas Next Year

So I'm getting angry watching basketball. And I've got $1400 riding on two NBA games. Which makes me uncomfortable, because a) the most I've ever bet on a single game is $200, and b) I'm a stats nerd. I HATE SMALL SAMPLE SIZES.

Next year, I'm thinking of making a different portfolio, where I simply place $150 on every team (betting the over or under as our forecasts indicate), both because this diversifies a lot more and because it's probably more fun for readers to follow this 'portfolio' over the course of a season. It also keeps me a lot less invested in certain teams (although the Sixers win in Portland was a lot of fun to watch!).

The downside is that I know there will be an equivalent to this year's Knicks. A team where the over/under is so detached from reality that I'll be itching to make a bigger bet.

What about you, readers? Which would you prefer to track? A weighted model where I bet lots of money on some teams, or an evenly distributed model with the same bet size on every team?

"With any luck, since Indiana can no longer win the first seed…"

EIther you mean Indiana can no longer LOSE the first seed, or you have been watching a different basketball league. The Pacers clinched the top seed in the East several days ago.
I keep my fingers crossed for you tonight!

I would suggest a mixture of your two approaches for next seasons. If you find a second "Knicks" case, in which the betting market is really inefficient, I would suggest betting a higher sum on the team (e.g., 500$ instead of 150$). For all other teams, you place your 150$ bet.
If you're a maths geek and want to make wagers, then the correct answer is to basically follow the Kelly Criterion: http://en.wikipedia.org/wiki/Kelly_criterion
Doing every bet, and using the Kelly Criterion to figure out the amount to put into every bet (it will naturally advise you make a bigger bet on the ones where you are more sure), looks like a great answer.
The sample size on these bets is NOT SMALL (using your all caps). On the over-under it is 82 games. I thought your model captured this and gave you some measure of confidence.
Fractional Kelly might be a better choice. FWIW, Kelly will tell you not to bet if you don't have an edge.

The flat betting approach seems foolish. If you want to keep track of how well Vegas does - or how it differs from your predictions - there's no real need to throw money away to do it.
It makes sense to weigh your bets based on the delta between Vegas and your analysis, especially considering how much people hate on you when you're predictions are wrong on this site. That said, it's always more fun when there is some follow up involving your home team, so even small bets across the board would be fun to track throughout the season.
"If you want to keep track of how well Vegas does - or how it differs from your predictions - there's no real need to throw money away to do it."

In fact, even if you did happen to throw money away this season, you can still use the Kelly approach to see how you would have done.
The profit maximizing strategy is to only place bets when you think you have a really strong edge. If you are stressing too much over outcomes, you can lower the bankroll to a more comfortable level.
With respect to several posters, Patrick did not ask for advice on how to maximize profits. He asked us what betting strategy we'd be most interested to read about. The more you bet, and the greater the number of teams, the more interesting the story becomes. Of course, budgets are limited. Maybe put 250 on each of the 10 teams you think most likely to exceed the total (most underrated) and 250 on each one of the 10 teams you think least likely to exceed the total (most overrated)?

With respect to several posters, Patrick did not ask for advice on how to maximize profits. He asked us what betting strategy we'd be most interested to read about. The more you bet, and the greater the number of teams, the more interesting the story becomes. Of course, budgets are limited. Maybe put 250 on each of the 10 teams you think most likely to exceed the total (most underrated) and 250 on each one of the 10 teams you think least likely to exceed the total (most overrated)?

Cedric:

"The sample size on these bets is NOT SMALL (using your all caps). On the over-under it is 82 games. I thought your model captured this and gave you some measure of confidence."

Go back and re-read the whole thing in slow-motion replay. If you watch carefully, you might see the joke I was making.
And for what it's worth, I'm more than comfortable with the amounts I have bet this year. I just find it a little unlucky that a couple of the large bets have come down to the very last game of the season (unlucky in the sense that I don't want to pay that much attention to the results of one game).

Yes, you can all point out that all the other 81 games had an equal impact on the outcome. It must be so fortunate for you that, unlike me, your awareness of a cognitive bias makes you immune to it!
Because of the awesome team previews we already know the Boxscoregeeks position on every team (assuming the previews continue). I would be more interested in following Patrick pursue a betting strategy that makes him more money.
RyNye,

Good catch, lol. I meant that the game has no impact on the seed, but brain-farted and wrote win instead of lose.
You guys do a preview on all 30 teams, there's nothing wrong with playing a bet on eat. But yeah, probability weighted is probably a lot better than equal weighting. If your model is more confident you should be more confident. Take a combo of line deviation from model predictions and model confidence and weigh your bets that way. If you have 99% confidence in a result, please bet that way.
You guys do a preview on all 30 teams, there's nothing wrong with playing a bet on eat. But yeah, probability weighted is probably a lot better than equal weighting. If your model is more confident you should be more confident. Take a combo of line deviation from model predictions and model confidence and weigh your bets that way. If you have 99% confidence in a result, please bet that way.
Just a quick question, when you listed the probabilities, was that assuming everything went well? What would you list probabilities as knowing that injuries are not knowable? It seems like that should be something that could be baked in (not for a specific team getting injured but as a differential for every team in case of significant injury). I think fewer people would be snotty if the confidence was 45-95% depending on injury with the stipulation that 95% is if things don't go wrong.
I wanted to point out that Vegas lines tend to be pretty sharp, so we should expect these Over/Under lines to come down to the wire every relatively frequently.
I think you should avoid the long shot bets, bet confidently on 15 over/under tallies however you choose, and bet one team to win the championship. That should be enough to constrast WP and your models with other analytics and popular public opinion.

Also, you should bet on every playoff series this year and next. From memory, Arturo came very close to picking the Finals winner in the correct amount of games, and only missed on a few picks (Denver almost beating LA).

If you make money consistently, who can argue with that? Lots of people! What fun.
Patrick,

It would be a lot of fun if there was a contest with funny money (say $1,000 per user) to make bets on over/unders for the season and then there would be a winner amongst your readers.

Full disclosure, I've never written code before and this might be far too difficult for a free site.
I like watching the profit maximizing strategy so do it like this year. I also love the long shot bets because they are so +EV. Plus, it'll be cool to see when you do hit one.
"Next year, I'm thinking of making a different portfolio, where I simply place $150 on every team (betting the over or under as our forecasts indicate)"

lol that'd be even worse since, for the third straight year, Vegas crushed your projections (rmse bovada 8.5, yours 10.8).

"A team where the over/under is so detached from reality that I'll be itching to make a bigger bet"

The Knicks are going to finish 12-13W under their line. There were 8 (!!) of your projections that were more detached from reality than the Vegas's Knicks one. Voila

PS. You can talk about all the regression you want with the Pacers and trash their fans but you still missed them by 16W.
PS. 35 wins for the Sixers...lol
I'm with Josh, putting the same bet on every team doesn't differentiate this from the season preview enough, focusing on a smaller number of bets where the WP projections are way different from conventional wisdom is much more interesting.
I don't see how you can call what happened with the Pacers regression to the mean, unless you are saying that they didn't have selfish iso and off-court distractions prior to the all-star break, and after the ASG they've had a ton so it evens out.

Their offense has gone from middle of the pack to near last, because their system is predicated on playing inside out, passing to distort the defense, and ZERO wing iso. They've done the exact opposite in the last 3 months. Those are deliberate, selfish actions of the ballhandlers, not "they were really hot early, now they are missing shots they hit before". They are, quite literally, not taking the same shots they used to.

More than anyone else, Lance Stephenson is asked to sacrifice shot attempts relative to his talent level. Meanwhile he sees Paul George breaking the offense and taking more and more FGAs. Then Lance got snubbed for the ASG, while George went, and that pretty much broke the "sacrifice for the team" mantra. Deliberate choices, not regression. Couple that with Lance being in a contract year and not getting an extension, and you've got a classic "I'm going to take the shot whenever I touch the ball" scenario. And as we all know, taking a large number of "I'm going to waive off the play and dribble drive into 3 defenders" plays is not conducive to an efficient offense.

The numbers say what happened, they don't say why it happened.
Hey, the Utah one worked at least! Splitting the two wasn't that bad.
"rmse bovada 8.5, yours 10.8"

What's the site that provided the rmse of all the various stat geek predictions? Hope they do that again.
I think it would be asinine of you to make a bet where you don't have the edge. Why bet $150 on something your model shows has an expected return of $140 for our entertainment (i.e. it's -115, your model predicts 35.7 wins and the line is 35). You expect a higher level of schadenfreude of the readers than we have!
@Al_S

http://weaksideawareness.wordpress.com/2013/04/23/checking-2012-13-nba-win-predictions-projections/

Sign in to write a comment.