Nba nerd

The NBAGeek Goes to Vegas

“Bond didn't defend the practice. He simply maintained that the more effort and ingenuity you put into gambling, the more you took out.”

Ian Fleming, Casino Royale

Recently, I went to Vegas for a long weekend. I had two key goals for that weekend. The first was to wrap up the code for this very website, to meet our ambitious launch schedule. The second goal was more capitalistic: to place some bets on the NBA.

The comments on this morning's post include lots of people telling us that if we were really so confident in our projections, we should be rich. We need only put our money where our mouth is.

Funny you should mention that. Recently, I was talking to other writers at the Wages of Wins network and we thought it would make an interesting exercise to track my bets over the course of the season; consider it a portfolio, if you will. Furthermore, it's kind of fun to "put my money where Arturo's mouth is." I wanted to do this last year as well, but I didn't make it to Vegas in time for the opener. For those curious, I'd have placed 6 over/under bets last year, and 5 of them would have paid -- the 6th didn't pay because of knuckle pushups.

This year I decided to diversify a bit. I placed a few longshot bets on some division winners, and 9 over/under bets. Here's a breakdown of the "portfolio": 

Team Bet Probability Bet Payout EV
Knicks Under 50 98.6% $1,000 10:11 $1,882.36
Hawks Over 40 85.6% $100 10:11 $163.42
Pelicans Under 40.5 86.8% $300 10:11 $497.13
Celtics Over 28 94.5% $500 10:11 $902.05
Pacers Under 55 91.4% $600 10:11 $1,046.95
76ers Over 17 98% $500 10:11 $935.45
Jazz Over 25 99.1% $800 10:11 $1,513.53
Warriors Under 51.5 90.1% $600 10:11 $1,032.05
Trailblazers Under 38.5 91.3% $400 10:11 $697.20
Celtics Win Division 19.8% $40 60:1 $483.12
Jazz Win Division 8.7% $40 100:1 $351.48
Grizzlies Win Division 26.6% $60 7:1 $127.68
Pistons Win Division 24.3% $20 17:2 $46.17
Cavaliers Win Division 15.6% $20 16:1 $53.04
76ers Win Division 12.9% $20 500:1 $1,295.16
Totals $5,000   $11,026.79

The Probability column comes straight out of our preview articles, where we listed the odds that a team would hit the over, the under, and win the division. The EV column is the expected value of the bet. This function is:  probability * (bet + (payout * bet)). Another note: you might look at some of the high percentages and wonder why I didn't play any parleys. That's simple: one can't play a parley on the over/unders. Or at least, at the Las Vegas Hotel there was no way to do so.

As some of you know, I had a brief stint as a "professional" poker player. I didn't earn earth-shattering money but it paid a lot of bills for a little more than a year. It's one area of life where my love for number-crunching paid off. A couple of lessons that I took away from poker include: 1) don't be afraid of pushing small profit edges (this is where bets like the Jazz or the Celtics to win their divisions come in), and 2) the risk of ruin carries a large opportunity cost. To understand the second point, imagine that you had a magic coin that could flip 60/40 for heads, and you had a person willing to bet on tails against you, and he'd never catch on. In this scenario, you'd be an idiot to wager your entire fortune on one flip -- even though the bet has a positive expectation, you will lose 40% of the time, and losing that bet means that you lose the opportunity to keep flipping the coin. You will make far more money in the long run by only flipping a portion of your bankroll at a time, to avoid the risk of ruin. That's where diversification comes in.

With this in mind, I bet more on things that I thought had higher probabilities of paying, but I still diversified. For instance, even though all of us at Boxscore Geeks think that the "surest" bet is the Knicks under, I didn't put all the eggs in that basket. The risk of ruin isn't worth it. Having said that, I should have put more on the 76ers.

The division winner bets are all longshots, so they don't warrant a big percentage of the portfolio. By the way, for these, the EV calculations are not exact; since some bets are exclusive of others (i.e. the Cavs and Pistions cannot both win the central, and the 76ers and Celtics can't both win the Atlantic, etc), the probabilities (and therefore the expected values) are not completely accurate, but the impact of this factor is low enough that I didn't bother with the math. (Editor Arturo's note: They actually are. I ran a full season sim 10,000 times and ranked all conferences and divisions for each sim and that's how the division odds were made. So there!)

For those curious about an individual breakdown of all the bets, almost everything you need to know is captured in our season previews, but I will add a few thoughts:

Yes, I bet on the 76ers to win the Atlantic. No, I don't think it's going to happen. Again, I'm pushing a small edge here. 500:1 is a ridiculous payout when the Nets are the only team in that division that I'm truly confident will have a better record. So consider this basically a longshot bet that a couple of Nets go down with injuries, which would create a situation where a .500 team can win that division. I grabbed the Celtics for the same reason. The payouts here seem way out of line with the odds.

Note that the Jazz will be a better team than the 76ers, but they are far less likely to win their division. It's interesting that the vegas line doesn't consider things like that the Jazz would have to beat both OKC and Denver. In other words, if I were making the odds, the Jazz division win would pay out bigger than the 76ers.

The Grizzlies' over was one of the better bets in our projections, but on the day I was in Vegas it went up 3 wins and I decided to pass. I debated placing $100 on them just to make a statement (i.e. that I believe everyone is sleeping on the Grizzlies), but in the end the money seemed better spent elsewhere, and the division win bet already makes that statement.

If there were any bets here that I'd be "worried" about, it's the Pacers and the Celtics. The Pacers because I worry about 2011-12 Hibbert rearing his head, and the Celtics because I worry about minute allocation. I worry about things like Danny Ainge trying to lose on purpose by keeping Rondo out for the whole year.

So what are your thoughts on the bets? Did any of you make it to Vegas and put money down? What were your bets?

Nice article dude. Glad to see you guys are putting all your hard work into something that is (potentially) quite profitable. I laid some bets down myself and mine look like this.

NYK / GSW / LAL (kills me to bet against the Lakers) / NOP(e)

TOR (Raptors fan, had too) UTA / PHI

TOR to win division.
Not wagering on 70-75% projections (Spurs) after explaining how much of a value represent a 60/40 event shows how much you trust your "Wins Produced" stat.

Anyway good luck with your bets, i'm on some of them too :)
Well I'm taking the over for the sixers but I also want to bet on the raptors making the playoffs even though I didn't pick them to make it. Vegas has underrated them. I might spread the wealth or decide to spend more. Fantasy Basketball contest are profitable also. Im trying that this year.
Perfect - this is the best way to test. Look forward to seeing how this unfolds. Good Luck.
Awesome to see you putting this to the test. I actually picked Knicks under, Jazz over and 76ers over as well, and diversified with a couple longshots (Rockets or Spurs as champs, Rockets western Conference winners, Grizzlies/Rockets highest placed team excluding Heat & Thunder), picked Raptors for the division at 1:35. Oh and the Celtics with a better finishing position than the Lakers at 1:3.65 as more of a safe pick as well. The previews were of great help, awesome work guys!

So sorry my bankroll isn't $50k and I couldn't make EVERY profitable bet I wanted. It must really suck for you to read my stuff, I'm sorry to put you through this.
Do you think it is safer to bet an under than an over? It seems like the things that can go wrong to miss an over are more likely than vice versa. Since wins are typically tied to a couple star players, one injury can derail the entire bet. Contrast to betting the under for the Knicks this year. Barring an early season trade for LeBron, there's no way they are winning 50 games.
I personally view the O / U as equal risk. An Injury might take out a player who gets lots of minutes and force a coach to play a bench player that is actually quite productive and the opposite. Players over and under perform all the time. Throw in a GM making a trade or two and the original roster is out the window.

I.E. If PHX had traded Gortat 2 weeks from now it might have seriously altered your futures bets
mdietz has a nice point. Presumably, the lines are set to account for the chance of impactful injuries, such that any line reflects the consensus estimated probability that the players on a given team will suffer injuries.

That said, it makes intuitive sense, using the Warriors, let's think about how it works.

It's basically impossible to set the lines to account for an injury to Curry. Sure, on average, if we assume there's a 20% chance he misses 40 games, it'll drop their expected wins by, say, 3 wins.

But the injury is a discrete event. If he actually does get injured for half the season, it might drop them by 15 wins, way more than the false conservatism expressed in an expected wins prediction.

And when you compare this to the effect of injuries to opponents, the picture becomes still rosier for the under. It only takes one severe injury to a key player to ruin the Warriors' season. But in order to derive a benefit equal to the amount that the Warriors stand to lose to a Curry injury, each of their opponents would need to lose a similarly productive player. It's fair enough to assume that the odds on that happening are much lower than the odds that any given player goes down.

Players do over- or under-perform their expected per-minute production. But there's no similar equality on the minutes played side of things (at least when we're talking about the starters). You either play the maximum amount of minutes you would play un-injured, or you play less. Now, that doesn't hold for a player like Drummond, who could end up playing 34 minutes a game, but for most of the most-productive players, it's true. They play as many minutes as they can, or less.

In short, the average expected wins estimate can't accurately value the impact of an injury to a star player, because those events either happen or they don't, so averaging out a team's expected wins, even if you accurately predict the probability of injuries, doesn't account for the severe impact such an injury would have.
Patrick, you should use some of bankroll formula to reach the goal and making money or more money than u did. For example simple Kelly formula. Probability * odds as a part if your bankroll. Boston have almost 20% to win div with 60 odds ! (I bet above 100 odds btw) It's 1200% in long run. Your "best" bet on NY give u only 90%. 40$ to 1000$ in this situation are not fair. Someone told you before - you can not believe in this percentage unfortunately ;)
man i'm jealous. i was excited to bet on sports for the first time, but i'm poor, don't live near vegas, and don't trust gambling sites. blazers under knicks under and jazz over are free money, and blazers under had the best payout where i was looking so i was considering consolidating my bet into that because, if you can replace jj hickson with robin lopez and win 6 more games than last year, in the west, i should be more worried about the apocalypse than a lost bet. good post. also hilarious that someone is chiding you for not betting on a 72% chance and that this somehow indicates a failure in the model... the spurs are a machine, but under 3/4ths a chance isn't good enough for me; i don't want a gamble, i want free money. not to mention, as someone was saying, this was an over bet so a moderate injury keeping an important player out for just a few weeks could spoil it.
Buttonik, I'm not sure you understand the Kelly criterion as well as you think you do. If the Knicks truly are a 90% favorite on a 10:11 proposition then the Kelly fraction is much higher for them than for the Celtics at 20%, 60:1. I'm not going to run the math but with an eyeball test Patrick's bets are not way out of line (although the 98-99% teams should all be about the same size bet.) The real problem is that I don't think he really has such an overwhelming edge over Vegas :-)
Betting on sports is like playing poker in Vegas; you're not playing against the house, but against everyone else. The house just wants to maximize action.

I tend to agree that the odds are not quite as good as your numbers lead you to believe (last year's Timberwolves and Nets should be cautionary tales about tail risks) but most of these are still very solid bets. I'm most skeptical about the longshot odds, as it is very, very difficult to distinguish between a 1 in 50 and a 1 in 500 chance without enormous reams of data.

Yes I understood quite well. Of course should be big difference between those bets but it should be circa as 5 to 1 . If 1000 on Ny about 200 on Boston - not 40...
Awesome that you are trying this out. I don't have a large bankroll so I decided to place some nba east/west/finals championship bets (mainly on Houston and Memphis) so I could get the larger payoffs. I'm going to use some of your division picks as well. Would be interesting to figure out how the profitability of the championship futures compares to the win/loss over/unders. I suppose you also have to consider that your money will be locked up for a longer period of time with the playoff futures opposed to the reg season win/loss bets so that carries a cost as well.
Why not try foreign Markets? They are usually different because the population in a country is usually disproportionately loyal to one team, e.g. Australians tend to over bet on the Dallas Cowboys. has only champion odds, but it is something to consider...
Nice post Patrick. It will be very interesting to track these as the season plays out. Just a few comments:

- The expected value calculations you made are actually correct. ie. E[X + Y] = E[X] + E[Y]. This holds even if X and Y are correlated.

- It would be interesting to not only see the EV, but also the distribution of results. For example, how often will you lose money? If it's shows a low probability and it happens, that could be an indication that your models are not as good as Vegas'.


Not looking great! You might need the 76ers to come through and win the division to bail you out.

Sign in to write a comment.