But a calculator never misses 2 + 2 = 4.
WE: So, if one is looking to specific splits or trends or whatever else to find a spot to bet, is it even worth the time to have your own simple model-that is, a way to come up with an estimated spread based on scoring and whatnot. When you look at the king of stat driven sports/analysis, you’re looking at baseball….which is a much more static game. You hear about “betting models”, some of which are talked up so much they almost seem like mythical scrolls.
In essence, we use a model. These trends, these stats, this schedule, these match-ups. We can at least rule things out, and as a sports bettor anytime I can clear out some noise, I feel like I’m helping myself.
MS: Right. And I think the problem with that is that really most models are weak models. Basketball is messier. It’s like those crazy coupon hunters who end up finding a way to make money shopping.
WE: So is it even possible to develop a non-statistical approach to consistently beating a book? You might be able to spot something like ‘Isaiah Thomas is a sparkplug, Boston is a team that’s hungry for fresh blood’ on one night, but these things can be maddeningly random. I think you can pick up on certain things. We’re basically hand-pumping a model. So the lesson is you can’t predict basketball. So in the end you’re looking for an edge, is a simpler way of putting it.
MS: Well, yeah, I think so. A lot of times they are designed to produce *a result* but not necessarily a predictive result. But even still, in baseball you can generally narrow things down, like you can play the percentages of a pull hitter versus the way you can align your defense, and you know the statistical outcome. These are things that books generally pick out, but not always, and the deeper into it you get, the more likely you are to find things which may not be factored into a line. That’s why it’s gambling.
When you’re thinking of a successful investor, sure, they are going to react as an investor of say, Folgers, if Folgers is successful. As part of that series we asked two friends of the program, Matthew Stewart and Will Eisenberg, for the perspective on the overlap between gambling and analytics. On the other hand, it’s also easy to come up with “just-so” stories for this type of trend-watching. But that’s it. Except when you can. Finding the point at which to abandon strategies, and the moments when you need to stick to it and hope the numbers start coming up better for you. For example, maybe you know that player X has had some problems scoring in X arena. Then you have to hope they still haven’t caught on. Sixty years ago, for a local bookie, this just meant being more informed than his clients, which wasn’t particularly hard. And you can bet (PUNS) that they are using modelling-they’re not going to risk vast sums of money on a gut feeling about how scrappy Aaron Craft is. And even the best read can get blown apart by a couple of free throws made or missed. if what you’re saying is true, that basketball is just a slightly organized kind of chaos, are we just rubes? If it is that messy, is it predictive enough that one can use available tools to predict it consistently? This is like the million dollar question for any semi-serious sports handicapper.
MS: Right. And you can tell that only through the repetition of something which resembles a model. Let’s go back to the investor/pro handicapper analogy for a second. So… I know it’s been beaten to death on the internet but here is the important point. Will is a classical musician from Minnesota. It’s chaotic. I’ve bet basketball for about the past 6 years, and you the last two, and we’ve worked together for a majority of that. Okay, and I think where the mythology of the “model” comes in is from the fact that when you’re talking to a gambler on Twitter or a forum or something or if you’re — yikes — listening to one of the 1-900 number touts, you’re hearing about models that are both ridiculous and imaginary. But I also know I have a vested interest in that opinion.
MS: Well, in reality everything is going to be a model with human error, regardless of what you do. That was a real thing, evidenced even by him basically admitting he tried harder. So what you have is things from that which you can pick out. And they’re going to react in a very specific way based on their investment principles. Just noise. Of course, the book’s lines can be way off on any given night too. — are not using models so much as playing markets. You hear a lot of bonafide get rich quick schemes and perfect systems. But the randomness definitely hits us harder, because we have to choose the spots we bet. Figuring out how to ride out bad spells is definitely something I want to get better at. And if anyone tells you something other than that, there’s a reason and it’s not good news. We know this if for no other reason than the truly successful and verified pro handicappers in the world — Billy Walters, Haralabos Voulgaris, etc. They’re something like 10-1 now. Or is it simply better to trust that the book is going to be “about right”, and your only goal is to either find a side that has an edge or a line you know can get value on?
WE: Is basketball harder to bet because of the way in which team play and system affect a player’s stats?
MS: So before we get too much further into the model discussion: I think we should get the elephant out of the room. Follow him on Twitter @TLMBERWOLVES.
But yes, if I’m telling a novice sports handicapper how to take themselves more seriously, I don’t say to them that modeling is completely bunk, even if it’s not always particularly helpful. The Warriors were for the majority of this season unbeaten against the spread when they had 2 or 3 days of rest before playing. The Bookmaking business in the United States has always been a sort of statistical operation, but in the end what the bookmaking business has been is just being better at whatever the current method is. Without giving away the whole recipe, we go through a check list. At the same time, you can risk going too deep to where it is all bad noise.
WE: Studies are inconclusive.
When we think of the “model” that I think a non handicapper is thinking of, it’s something akin to an excel spreadsheet or dataset in which you can put in numbers and you can spit out a percentage based result that over time will definitely win you money.
MS: I think that’s the crux of the whole thing. So, we had a conversation.
So if it isn’t, we’re kind of wasting our time. They find the cheaper prices on games and take advantage of that via a vast network of runners, etc. For example, one of the tougher things baseball has dealt with over the last fifteen years is honing down defensive measurements. Matthew is a grad student from Austin, and can be found on Twitter @_MatthewStewart. So. And when you’re not working with some way to get a feel on which games are worth betting, you’re probably doing a huge amount of legwork to come up with a small number of bets. I think it is. And so on.
WE: It’s been totally up-and-down this NBA season. So is scrounging the best prices and catching the best numbers just the only viable option?
The idea with this is that modeling is still a relatively new thing in sports betting, even if statistical modeling in general isn’t. At some level. Until we are able to measure every fathomable aspect of a sport, there is going to be some human investment in whatever system you can create to compensate for things we think might be valuable but cannot measure, including a market based system. There is situational coaching like that in basketball, but I’m not sure it’s as…..widespread? Maybe that’s not the right word.
For the past two years I’ve been betting basketball in tandem with Will Eisenberg. When asked to answer the question of whether or not models can actually work, I couldn’t give a complete answer without including him. It can be tough to lose a game on the last possession, but it’s even worse in my opinion when you feel like you have a really good read on a team, and you’re just completely wrong about that night. That was a lot of words.
Let me ask you…. Well of course they are!
MS: I don’t think it’s the only viable option. But more importantly, this means on a broader scale, I think it can be much harder to objectively gauge an individual’s impact solely through numbers.
That said, we’ve taken a beating in the second half of the NBA season after a very good NBA first half, and when that happens, it really can all seem sort of futile, can’t it?
WE: I think one area I’m definitely guilty of is being more interested in whether a trend is robust than if it is explicable. And yes, they’re aided by computer models, but they’re generally the same models the books have.. Betting basketball is hard. Which again, puts you at the mercy of the randomness of basketball betting. Chaos means ‘sensitive dependence on initial conditions’ and that reigns in basketball. For example, there is no doubt that Reggie Miller played in a different way against the Knicks than random mid-December games against random team. Over the course of even a small amount of time spent interacting with sports handicappers, you hear a lot of things. Which is sometimes.
This is kind of a “the best way to go broke slower” idea, but I do actually believe that the people who are able to be the most efficient in that way are actually profitable. Maybe in the end it all just about evens out, but switching up my bet size after a string of successes definitely backfired for me. As someone whose interest in basketball was kind of driven by a love of analytics as well, do you see there being more “immeasurable” things than in a more “static” sport like baseball?
Freelance Friday is a project that lets us share our platform with the multitude of talented writers and basketball analysts who aren’t part of our regular staff of contributors. Or other splits. Of course on any given play, there are an infinite number of ways the players can interact, and a ton of different possible outcomes, even in ‘resetting’ the game. Once every measurement has an answer, you no longer have a model, you have a….calculator? After all, I am sure I have missed many “best” prices on the hunch that a better price was still going to come along. This could be entirely pointless, or it could be helpful, depending I guess on your perspective on the whole thing. And though you will hear of it often, the elusive perfect betting model still obviously alludes us, since sportsbooks have not exactly yet gone running for the hills.
WE: So, I guess my point is, you don’t need to be able to model basketball perfectly, right? You just need to be able to do it a little bit better than the books do it. On a larger scale though, they are always going to be very formidable to beat. So you are continuing to limit. 1 day travel records? We know we can count them. But! They are also going to understand how the rest of the market is going to react, in order to know when is the best time to maximize value.
WE: Baseball moves from a fixed position (or really, six main scenarios) and so it’s predictable, which makes measuring impact easier-it’s also, essentially, a one-on-one sport. I think what you just described is where the efficient sports handicapper exists.
If you’re a “serious” handicapper — whatever that means — you are indeed always fiddling with the method you use to place your bets. For example, do we really understand the impact of 2 days v. This is still relatively opaque to me.
Phew, okay. Some think it is impossible. Models can be a very good way of telling us what is bad information, even if they are not as good at telling us what is actionable. Are the Warriors genuinely getting a boost from practicing with Steve Kerr and having a chance to rest their legs so much bigger than other teams in the league that it’s being systematically undervalued? On the other hand, if you backed that trend early in the season you probably made out well.
In reality, that’s not how real handicapping works