00:00 - Andrew Mack (Guest)
There is a distinction to be made and it's a very important distinction between an edge which comes from an effect, and a model, which is a tool to help you identify the effect which can lead to an edge. It is not necessarily true that a model provides an edge. Model only provides an edge when it is identifying an effect that allows you to beat the VIG. 

00:19 - Rob Pizzola (Host)
Come on, let's go. You think I'm going to come on here and put one of the Aussies in the elite tier? I'd vote for Rob of $25,000. I wouldn't vote for you. I could basically just cheat and get the same bets that they're getting. It's weird because you can also pay your bills at the same place where you bet. I mean, you're a short guy, what? This guy talks a lot of trash. He's talked a lot of trash about me, rob, a lot of people in the community, but he's refusing to show his face, clouting yourself as a pretty good NFL gambler. I thought you were an idiot. Take testing, testing. I'm a cuck. One, two, three, four Aussies. 

00:57 - Andrew Mack (Guest)
Going head to head with the Aussies. That's what I grew up for. 

01:05 - Rob Pizzola (Host)
And I said get him Kirk. And they call me a mean-spirited name. I don't hang out with them. The Toronto Maple Leafs at 14-1 are staring me dead in the face. 

01:09
I'm already getting you know a lot of early this could be the best Circles Off episode that's ever been done. Welcome to Circles Off episode number 185 right here, part of the Hammer Betting Network and presented by Pinnacle Sportsbook. I'm Rob Pizzola flying solo in studio today. I shouldn't say that my producer, jacob Grumania, is over here with me as well. We're one week removed from one of my favorite interviews that we've ever done with Matt Metcalf, former sportsbook director at Circus Sports. If you haven't watched that yet, I would highly recommend you do so. The feedback has been great. The average watch time on that interview is insane, metrics wise. There is a ton that anyone can get out of that. If you haven't checked it out yet, check it out. We have a great guest on for today's episode as well, someone who I've known for a long time, interacted with for a long time. I think you'll get a lot of viable stuff about modeling sports in this interview, whether you're a beginner or an intermediate, maybe even advanced, probably learning something here today. Before we get to that, next week we have an interesting episode. Johnny will be back in studio with me next week. There's some announcements we're going to make, so you're going to want to set notifications here on next week's episode. Circles off is changing going forward In the new year, in 2025, there are going to be some differences. We're going to highlight them all in next week's episode. I know there'll be a lot of speculation. Probably don't know what's going on, but either way, wanted to put it out there. Make sure you watch next week's episode as well. 

02:43
Now, before we get into it today, I do want to remind you if you are in Canada, make sure you check out Pinnacle Sportsbook. Find out what pro bettors have known for the past 25 years. Pinnacle is where the best bettors play. I've been betting at Pinnacle for a very long time Great limits. If you're a sharp sports bettor, you're looking for higher limits. You can get them. Even if you're a top-down sports bettor, you get better limits on props than a lot of other spots, and you can still find edges at Pinnacle. It's one of the things that kind of flies under the radar. People are like ah, pinnacle, sharp book, how can I possibly beat them? Well, there's edges to be had every single day on props and things of that nature as well. Plus, they have great customer service. Easy to get money in and out. It's a no-brainer. Check out Pinnacle Pinnaclecom slash hammer if you want to do so, you must be 19 plus, not available in the US and, as always, can't preach this enough. Please play responsibly. 

03:38
Today we're joined by a very special guest. He's a sports bettor and he's the author of statistical sports models in Excel. He has spent years honing his craft, using statistical models to gain an edge in sports betting. His book provides a deep dive into how betters can build and use models in Excel to improve their strategies. But beyond Excel, he has expanded his expertise to tools like R, python and Stan, and we're excited to discuss his journey, his betting techniques and even the psychological elements of long-term sports betting. You can follow him on Twitter at GingeFaceKilla. We'll link that down in the description below Andrew Mack welcome to Circles Off. 

04:16 - Andrew Mack (Guest)
Thanks, rob, it's a pleasure to be here. 

04:18 - Rob Pizzola (Host)
Yeah, great having you, we finally got this going. I know it's been in the pipeline for a while, so appreciate your patience on that. I've heard your background before. I've listened to other podcasts involving you. We've had some private conversations as well, but this is a broader audience here. So if you can, in a few minutes, just share a bit about your background and how you first got involved in sports betting. 

04:40 - Andrew Mack (Guest)
Sure, it's kind of a long, windy road, to be perfectly honest with you, but I'll try to give the Coles Notes version, if anyone still uses Coles Notes these days. Dating myself a little bit, but I got into sports betting sometime around 2011 or 2012, I think, and at the time I was an electrician. The short version of my story is that I was a commercial journeyman electrician, eventually went to law school and then graduated law school, decided that being a lawyer probably wasn't going to be for me, doubled down on sports betting right as COVID started and then managed to sort of expand my repertoire into stock market trading, options trading as well, and now that sports is back running smoothly, I bet sports and trade stocks and options full time. Getting back to the origin, it was sometime in 2011 or 2012. I can't remember the exact date, but I do remember the exact bet, because I bet on the Philadelphia Flyers to win on the money line and the reason that I did that was because they had Jaeger on the power play. 

05:50
How dumb is that right? But you know, that was the first bet and I'll always remember it. One because it lost and two because that's got to be the silliest reason to place a bet in the entire world. But that was back when, you know, I was as green as you could be and I thought that the game had to do with who knows the most about sports. And I watched a lot of hockey, played a lot of hockey growing up and I thought, you know, hey, I know hockey, jaeger's on the power play, right, like these guys are going to, they're going to win and that was the start. Obviously, considerable improvements since then, but that that's the short version of the beginning. 

06:31 - Rob Pizzola (Host)
Yeah, it's funny. I think back to some of the bets I made in the early days as well, with I mean hate to call your reasoning dumb, but it frankly it was, and I did the same thing many times. It's funny now as sports bettors, when you like evolve over time to be able to actually look back on that stuff and just kind of have like a laugh and be like I can't even believe that that was me at some point. But such is life we've. We all place bad bets nowadays. You're modeling what. What role does modeling play in in your personal betting? Are you entirely modeling everything you're doing like? Would you ever place a bet nowadays without actually modeling the outcome? 

07:09 - Andrew Mack (Guest)
Well, it might surprise people, but the answer is yes, I would place a bet without modeling the outcome, and it relates back to something that I think might be a theme for our conversation today, which is that there is a distinction to be made and it's a very important distinction between an edge which comes from an effect, and a model, which is a tool to help you identify and quantify the effect which can lead to an edge. So people conflate a model with an edge, and it is not necessarily true that a model provides an edge. Model only provides an edge, not necessarily true that a model provides an edge. Model only provides an edge when it is identifying an effect that allows you to beat the VIG, and so there are situations where that model, as a tool, can be extremely powerful and very useful and you should absolutely use them. But there are also situations where you don't really need much of a model to harness that effect, and that's the reason for my answer um, you've been at this for a while now. 

08:08 - Rob Pizzola (Host)
How has your approach evolved over time in sports betting? How quickly did you get into modeling and in the early going? What did those models look like, um, relative to nowadays? Were they simplistic or were? Did you get really into the weeds in the early going? 

08:28 - Andrew Mack (Guest)
I didn't get really into the weeds in the early going because I didn't know what I didn't know. 

08:32
I had that beginner's enthusiasm, beginner's confidence, but also beginner's ignorance, which I'm sure many people can relate to. 

08:40
But the early models were very simplistic. With hockey specifically, I remember in Excel doing some basic stuff with goal differential and trying to basically like weight goal differential. But I didn't really know what I was doing and that sort of led me down the rabbit hole to statistical analysis and various different types of models which got me into rating systems, ranking systems, which at the time I thought was like, oh, this is great, and you know, look back on it now and you're like you know, a rating system is pretty, you know one-on-one sports betting kind of stuff. But that's where it all kind of started for me. As I got further down down the line, got into you know more advanced modeling techniques and and then obviously branched out from excel into r and using databases, uh, which which has obviously improved my efficiency tremendously as a, as a one-man show. I mean, it's very, very difficult to do this as an individual operator if you you're not having access to a database and the ability to scan a large number of potential bets quickly right? 

09:48 - Rob Pizzola (Host)
Yeah, I totally get it. I want to start with Excel a little bit because I think for the larger portion of the audience it'll make a little bit more sense and we can talk about some of the more powerful tools later. But you did write a book Statistical Sports Models in Excel. I do recommend it for anyone that ever asks me about a starting point. Describe it a little bit in the sense of who's it geared towards as a better. Does somebody need to come in with any sort of statistical foundation in order to understand it? Who did you originally write the book for? 

10:22 - Andrew Mack (Guest)
So I actually have a couple of books written written and two of them are in Excel. The first one Statistical Sports Models in Excel, volume 1, the blue book is it's all about generalized models, which essentially means that they are models that can be applied to any team sport, but they're looking at ratings from a team level, which is both their strength and their weakness, and it's designed for people that are looking to make that first baby step into statistical modeling, because it does show you some very powerful modeling techniques. But, as I say in the book, the application of those techniques is largely limited to low limit sports where there is not a lot of data, specifically player data, available. So if you take a Bradley Terry model, which is like a team rating model, and you try to use that to beat the NBA, you're going to lose and you're going to lose badly. But there are lots of low limit sports, which is what I would recommend, that beginning betters. Focus on that. You can successfully compete with that type of a model. 

11:25
And the example that I give in the book, as I recall, is Icelandic women's basketball, which is kind of funny, but I mean that's a, that's a low limit, obscure sport with like 200, $250 limit. Now, if you have a, if you're a better with a $10,000 bankroll, like everyone wants to bet the NFL point spread or the NBA point spread, and it's madness. At that point what you really want to be looking at are props that have a couple hundred dollar limit or you know Italian men's volleyball, like, like things that are lower limit that you actually have a chance to compete with a model on. And so that's the first book. There are a couple other things in there as well. 

12:00
We have, um, there's a basic Monte Carlo simulation for the NBA three-point competition during all-star weekend, which has been very successful for a number of years. There's a lot of angle on college basketball totals, which again, um, going back to effect, versus a model, it's, it's a way to quantify an effect that is very real and will probably get you limited if you attack it, but it's an absolute real edge. So that's the first book. The second book, the red book, we get into player-based modeling. So it's a bottom-up approach where we take team lineups and use them to construct lines, and then we also look at simulations for different player props in basketball, football and baseball. And those are the two Excel books. 

12:50 - Rob Pizzola (Host)
Excel is obviously a very common tool that a lot of people use outside of the sports betting space. I think I use a lot of Excel nowadays as well. I'm not a skilled programmer. Anything that I require programming for I basically outsource. I just don't have those tools in my arsenal. But a lot of people, I think from the outside looking in, they view Excel as like this incredibly simplistic tool, almost to the point where they see somebody posting something in Excel, maybe on Twitter, and they look down upon them in a way. I know you yourself are now using other programming languages as well. Do you still find value in Excel? And also, for someone who is modeling sports, do you think in the long run they would best be served to graduate from Excel to potentially a programming language? 

13:41 - Andrew Mack (Guest)
Well, it's true, a lot of people do look down on Excel because it's sort of simplistic, like you said. I would point out that many people in the financial markets, including people that work at reasonably large institutions, use Excel every day for some pretty important tasks, and the other thing I would say is that the benefit of Excel, in addition to it being an extremely visual user interface, which helps the learning curve substantially, but it's not the best tool for any given task that you might want to. You want to make a neural net in Excel, you can. If you want to make a Markov chain model in Excel, you can. It will probably crash your Excel. 

14:34
Need a lot of memory, but you could absolutely do it, and so it's that versatility that I think makes Excel a good starting point, and once you get used to the idea of using formulas and how the cells are updating, I do think that that's a very good baseline to get into a more efficient programming language, but it's a great place to start. You can still build models in Excel that make money in sports. They are going to be very specific types of models. Like I said, they'll be harnessing specific effects, but it's absolutely true that you can still use it to great effect. 

15:12 - Rob Pizzola (Host)
With your current expertise in other programming languages now, have you considered potentially writing up any follow up books or guides for those tools? Sports betting in R, sports betting in Python, things of that nature? 

15:26 - Andrew Mack (Guest)
Well, I don't do a lot of sports betting stuff with Python, but I do have a book out on R called Bayesian Sports Models in R, which has been available since July is when I released it, I think. But that basically goes through using R and Stan with R-Stan, the interface for Stan within R. 

15:46 - Rob Pizzola (Host)
And basically Is Stan. I thought it was a C++. Am I wrong about that? 

15:52 - Andrew Mack (Guest)
It is based on C++ but there are libraries to interface with it inside of R Okay got it. 

15:58
Yep. It also has an interface inside Python too for people that are interested. But that book is all about Bayesian statistics and and how you can make uh simulations for for games for the NBA, the NFL, mlb and the NHL and using Markov chain, monte Carlo, to estimate parameter values. It's um, it's a good. It's a good book If you have a little bit of a handle on R and you want to sort of take things to the next level. But I think's that's going to be it for for books for now on sports betting, at least for the next little while uh, why do you say that? 

16:30 - Rob Pizzola (Host)
out of curiosity there's like too much effort to put in or not wanting to give away too much that's valuable. 

16:35 - Andrew Mack (Guest)
I'm just curious on on why you say that um, mostly the burnout factor, like when you're trying to run a business full-time and then you're also trying to write a book any book it takes hours and hours and it's easy to get burnt out really quickly. And I released that book in July. And then, most recently, I released Retail Options Trading with Ewan Sinclair and that just came out in October and I basically told most people like I'm not going to do any podcasts until the new year, like I need some time to just just breathe and get back to betting and trading. So I mean, I took this one, but other than that, I think I'm good for the rest of this year. 

17:17 - Rob Pizzola (Host)
Got it. What are the main limitations in Excel compared to like R or Stan? So maybe I can rephrase that in the sense of let's say, someone was to take up sports betting in Excel and that's their starting point. When would they graduate onto something bigger, like when does Excel no longer become a viable option? 

17:39 - Andrew Mack (Guest)
Well, there are a couple of ways you could look at that. The first, I would say, is the level of modeling sophistication that's available to you. So, like our Python will have some incredible packages for machine learning and advanced statistical analysis, principal component analysis to try and identify these effects that I've talked about. The other thing is that, as a sports better, as I'm sure you know and many of your listeners know you're trying to turn this into like a production line, like you're not just trying to price two bets right, like you're never going to get anywhere doing that. You want to be able to look at the whole playing field and find a large number of bets that that have some potential. And so when you're trying to increase your efficiency now, you want to have a database. When you're trying to increase your efficiency now you want to have a database. Now you want to be able to scan hundreds of bets instead of just looking at one or two. 

18:34
So, excel, you know you might. If you want to price out, just for example, you're looking at NBA props and you're trying to price out player points over under, you can punch that out in Excel for sure, and you know you can probably do 20 or 30 bets, it'll take you a few hours, right? If you're trying to do that for every different prop market and maybe you're also looking at NHL, like goals and shots and things, and you're also looking at NFL player props, now you've got a problem because you're running out of hours in the day, and this is when R and Python are going to be very useful to you, because now you can take whatever your methodology is and you can use it to scan the entire playing field to find those weak points, which is really what the game is about. 

19:12 - Rob Pizzola (Host)
Andrew, one of the blessings and maybe curses as well of Circles Off is that we have an extremely broad audience. It started with mostly sharp bettors, many of which who have built their own models over time, and there's now a collective audience of people who are just getting into sports betting. I can see that based off the comments that are asked every week. I can see that based off of demographic numbers within YouTube as well and what other channels people are consuming that also consume circles off. I'm going to ask a very simplistic question here, and maybe there's a lot that goes into it, but there's, I would say, a majority of this audience right now that's listening or watching, that has never built a sports betting model before. They probably handicap games themselves using some sort of methodology that they've laid out for themselves. But when building a betting model, what is the first step you'd recommend to someone that is just starting out and wants to get involved in that? 

20:14 - Andrew Mack (Guest)
Well, no one will take my word for it, so they'll probably learn this the hard way regardless. But what I would say, to save you a considerable amount of time, is what I said earlier. To reiterate you need to focus on effects, because a model is just a tool to try and harness an effect. So what most people do and I know this because I did the same thing is you start out and you say, okay, I want to know who's going to win this game, because you're a beginner and you still haven't quite got your bearings yet. So you take a whole bunch of data, you go to basketball reference, you go to hockey reference, you take all of these publicly available team stats, you throw them into a model, you run a regression, you do a simulation and boom, you have a model and the model doesn't make any money. In fact, you lose money. 

21:07
And the whole reason that it sucks is because you started building a model without any effect in mind. 

21:09
So the assumption that you had baked into that process was that some part of that model was was going to provide an effect that you could capture. So maybe you know the uh, the, the technique, the model technique itself is is, uh, more advanced than the the aggregate market, or maybe there was data in that model that the aggregate market that doesn't include. Neither of those things are true when you're using Excel and you're using basketball reference to get your data right. So so there is no effect. And so you're trying, you're you're basically just going through the footsteps of of people that are already betting in the market, and while you may get some overnight line stuff, just because you're able to get down earlier, because you have a lower threshold for what you, what you need to get down, otherwise, you basically have nothing. Um, so you really want to focus on an effect, and an effect is like uh, scoring in the second period of a hockey game tends to increase because the goaltender switch ends and the teams are closer to the opposing goaltender. 

22:11
That's an effect. It has a direct relationship to the scoring environment, and then you can go about building a model to try to quantify that effect. How much does the scoring change and is the market pricing it in? That's the appropriate approach to get the most out of modeling is starting with an, an observable effect, and then working away from there, instead of starting with the model and hoping that you get lucky somewhere along the way. Um, that's the hardest way to do it. Everyone likes to do that. You know, the latest and greatest machine learning model comes out and people just jam that in hope it works. But but really, if you start with the effect, you're going to be much better off at the end because you're starting with something that at least has a chance to provide you with an edge yep, really well said. 

22:58 - Rob Pizzola (Host)
Uh, you talked about building out a model, uh, you reference machine learning model and uh, just jamming, jamming it in and hoping that something works. Can you talk about the importance of testing and validating your models before you actually use them to place bets? 

23:15 - Andrew Mack (Guest)
Sure, this is kind of a gray area a little bit. I'm not the biggest fan of backtesting as it's normally conceived of, although it's useful to do, you should know how to do it. But the assumption that because something's worked in the past it will work in the future is a dangerous one and you can end up going around in circles a little bit with that. It is important to test a model. I think the best way to test a model is a walk forward optimization, so basically you fit the model and then you start. You start iteratively uh testing it on new data as it's presented. So like um, I think that like fitting the model and then testing it on current in-season data as the games come out is actually a really effective way to discard bad models before they hurt you. And one of my favorite techniques for that actually comes from plus EV analytics. 

24:09
He did an article on pinnacle uh many years ago called towards a theory of everything, part two, where he talks about uh, a Bayesian updated technique for Uh using the Kelly criterion, where you start with a prior of like uh, kelly says 10% is what you should bet, and then, as the bet results come in, it adjusts the Kelly criterion and when it gets to zero, you can effectively discard that model and say, okay, so this probably isn't something I should risk my bankroll on, and I do think that that's a very effective way to sort of stress test a model. The other thing that's very useful is Monte Carlo techniques, but with respect to the effect that you're trying to harness rather than the model itself. My preference is for doing something more like that, but ultimately you do want to at least test that the effect you're trying to harness is there and it appears to have enough of a difference that you can beat the VIG and hopefully more than that, obviously. 

25:07 - Rob Pizzola (Host)
Makes a lot of sense. I do a lot of backtesting myself. I do find value in it. I do echo your sentiment of just because it's happened in the past doesn't necessarily mean it's going to happen in the future. I think a lot of people do backtesting in the quote unquote wrong way, where they make these incremental, incremental what they deem as improvements to the model, but it's just basically throwing shit at the wall, right. It's like I have this baseball model and, um, here are my error metrics here's the roi against, uh, closing lines and so but then they just tweak something that doesn't actually make any sense in terms of how it's applied to the sport. That actually improves those metrics, but it's probably just sheer randomness. I think that's like one of the biggest mistakes I see people make when when they're backtesting stuff. 

25:55 - Andrew Mack (Guest)
I agree with you. I mean, that's a sort of a data mining error right, where you you just tweak the parameters until it looks good. But but have you, have you harnessed or identified or quantified the effect that you're trying to go for in any conceivably better way? Not really. And so that's the whole idea A model or a strategy or an angle, whatever you want to call it. It has to have some kind of appreciable effect. There has to be some reason that this makes a difference. And then you go towards well, has the market priced this in? 

26:26
But yeah, to get back to your point about backtesting, I mean the markets adjust, the market is not a static thing, and so backtesting assumes a static market, assumes that you could go back in time and you could place all these bets and nobody else would notice, nobody would tell you, nobody would. The market makers, the sports books, would not adjust to what you're doing. They're just going to let you make millions of dollars and nobody will change their strategy when they see this happening, and that's not very realistic. The market will adapt. And so I think that's important to remember in the context of backtesting is that as soon as you start hitting the order book. Things are going to slowly start to change. 

27:09 - Rob Pizzola (Host)
And that's where we see edge erosion over time too. Right, I was going to say that I've experienced this firsthand, especially with other originators who have potentially come to me in the past and said hey, Rob, I've built this out, it wins. I've tested it against the last five years worth of data. It's got more accurate error metrics than the pinnacle closing line. I mean if I had a dollar for every time, I'd heard that I wouldn't be rich, but I mean I could buy a pretty decent meal. Happens all the time. Then you actually free roll these people, you put them and you realize that exactly what you said, I mean you're testing against stuff that I mean I don't wanna say say it's fictional, those lines existed, but once you actually start to enter the market, things change. You might have back-tested and had an edge on one team in the NHL on 30 of their 82 games over the course of the year, but guess what? You start betting them three, four games in a row, market corrects and all that stuff happens. 

28:06 - Andrew Mack (Guest)
So I think that was really really well, said andrew um, I wish I would add something to that, actually, which is that, um, there's this, uh, at least from what I see online, there's this, um, ignorance about the fact that the game that you're actually playing is a market game, and the market game is like an, an abstraction layer on top of the sport and so, and that that market layer is always changing because it's competitive and everyone is trying to get edge. So, you know, the idea that this is a static thing, is an, is an assumption, that it's a math problem, that it's a has a closed form solution, and one of the first things that you should write down about markets is that they have no closed form solution, and that's a great thing, because it means that when some edges are starting to fade, there are new edges that are always emerging, but change is the most constant part of the entire process. 

28:59 - Rob Pizzola (Host)
You've talked about now using R or Stan versus Excel previously. Could you share a practical example of how you've used R or Stan to develop a model for a specific sport or a specific market? 

29:15 - Andrew Mack (Guest)
Sure, I mean I use R basically for all of the sports markets and financial markets as well these days, but I mean we could really pick any of them right Like a basketball model. I started with props, so props is like kind of where I got my footing on programming and coding and everything before eventually getting into the main lines. But basketball, for example, is extremely dependent on player minutes and lineup rotations and things like that, and so you really want something that has a database that allows you to track player minutes and rotations specifically. So this is really important for, like, uh, people that are looking at like first quarter and first half type of stuff. But, um, when coaches decide to pull players out, um, like there's the starting lineup, usually takes a break sometime around the just before the end of the first quarter and then they're out for most of the second quarter before eventually going back in the game. Um, there are effects around that that you can attack, right, uh, but you're going to have a really tough time, um, analyzing all of that and making sense of it in Excel, like our Python is really the preferred tool for for that that level of detail. So I've done a lot of work on that, and obviously I have models for NHL, nfl all in R Stan. 

30:38
Specifically, the Bayesian approach is very useful for a number of things. It's not I wouldn't say that it's necessarily better for, like, the median value of the line, but what it's really good at is producing a full posterior distribution that allows you to answer most of the questions you would have about any one game. So you've got the median line, you've got the shape of the distribution, the standard deviation, so it allows you to look for things like middles and basically it's going to give you a lot more options in terms of how you want to attack one game, right? So, like I said, you might find that the volatility for this game should be a lot less than the market is implying, and so you could. You can kind of like make your own middle by taking the top and the bottom values and betting both of them, expecting it to remain inside um, things like that uh, I want to pick up on that basketball example because I think it's an interesting one. 

31:33 - Rob Pizzola (Host)
So the overarching question I'm going to ask you is do you think that everything in sports can be modeled? And the reason I'm going to ask you that is because you picked up, you know you specifically mentioned player minutes for the NBA and lineup rotations, and I know a lot of betters. There are some that model NBA, but there's also some that I know who are successful that they basically say to me there's such limited samples in some certain player rotations that their human mind can process this better than any type of model. So essentially the notion that well, if this player gets into foul trouble early and they got to go to the bench, I know that if the team is down by x amount of points, this is likely who will come in if they're up by x amount of points? It's basically all these things that exist in their head from watching the sports. How do you react to something like that? 

32:30 - Andrew Mack (Guest)
actually I agree with it because I personally know two examples that do just that, and so I, and so I can't denigrate that approach because I've seen it work. 

32:38
I know one guy that does basketball props strictly by watching basketball games like an absolute machine and does very well at it and basically is looking for exactly what you said, like the coach's reaction to performance. 

32:53
He sees a player maybe getting sat back down on the bench because they did something dumb and and so you know he makes a little mental adjustment and absolutely crushes uh points and rebounds and things just by watching the game. And I know another guy who, um, he's actually really good at fantasy sports but he basically can like visualize a game script in his head because he's been watching football for so long and and and has been very successful with player props based on the game script about, like you know, is the passing game likely to be like over the middle of the field or to the edges, a cover two versus cover three. Who does that open up, who gets more opportunity? And he's, it's, it's, it's an instinctual thing, but he's effectively doing the same thing that I would be doing with a model, and so I've seen both. I've seen both ways work. Again, I think that's because the effect is the effect, and if you have the, the, the bandwidth to keep it all on your head because you're a massive sports fan, then you're more power to you, right. 

33:51 - Rob Pizzola (Host)
Yep, absolutely agree with you. Uh, you've talked about some financial market trading. Uh, are there any lessons or techniques that you can take from financial modeling that you found that are applicable to sports betting? 

34:06 - Andrew Mack (Guest)
the main one that I found is the concept of volatility in trading. Um and volatility in trading the closest parallel is the standard deviation, uh, for you know scoring for the game total, for you know the margin of victory, uh. But basically, volatility in trading, uh, especially with options trading, is an extremely important component because it's how options are priced. But you know to keep this more strictly on sports, if the you can attack a sports line more than uh, for more areas than just the median line, and it's like I just previously said, if the size of the distribution is larger than the market implies or smaller than the market implies, there are things you can do. 

34:47
So if you think, if say the uh, a basketball total, uh, the market is implying that the standard deviation is like 14 and a quarter or something. So maybe you've done some modeling and you think that for these two particular teams the standard deviation should be like 22 or something, you may want to take an alternate line, right, you may want to sell the total up to a higher number, same with the margin of victory. If you have a model that suggests that basically the spread of that distribution, the standard deviation, is going to be larger, you might want to sell some points and then vice versa. If you think that it should be smaller than the market implies, you might want to create yourself a middle using the top and the bottom. So there are different ways that you can attack volatility as an angle itself, rather than you know the median of the line yep, I I actually love that. 

35:42 - Rob Pizzola (Host)
It's uh one of the ways I think about sports betting nowadays, especially with alternate lines, and I think it's becoming more public knowledge nowadays that, like not every uh nfl game in the in a week has the same distribution. There's pre-existing injuries to quarterbacks that could matter where you might have potentially an alternate line that's valuable on the underdog. So I really like that style of thinking, andrew. It makes a lot of sense. Question I get a lot which I'm going to pose to you. People are often reaching out to me for beginner modeling questions. How often should bettors update their models in order for them to uh to stay relevant? 

36:19 - Andrew Mack (Guest)
well, that's going to depend an awful lot on what the model is trying to do like if, obviously, if you have a win total model, you're not going to update that after every game. 

36:27
Um, it's going to be more sensitive to uh you know, the starting, the starting rosters and rule changes and things like that, so you know you can let that one sit for a while, whereas a player prop model is going to be extremely sensitive to recent changes in the players playing opportunities via minutes and also who they're playing with, which can change the global offense potential for that team in basketball, as an example. So it really depends it also from like a team perspective will change by sport. You know, if you look at an ELO model for teams, like we did in statistical sports models in Excel volume two, you can see that the K value, the updating sensitivity of those models, changes based on the sport. So you know a sport like basketball has a pretty high rating, so you should be more sensitive to recent changes. A sport like soccer or hockey is less sensitive because individual outcomes mean slightly less in those sports. So it's all very contextual and hard to encapsulate in a single sentence. 

37:26 - Rob Pizzola (Host)
How often are you reviewing the actual logic itself, though within a specific model? 

37:32 - Andrew Mack (Guest)
Okay. So the logic itself. Hopefully that's something that you've got relatively ironed out before the season begins, because if you have to, you know if your f1 race car blows, blows up in the middle of the track, it's going to be a nightmare trying to fix it and also stay competitive for the rest of the season it's. That is not a fun thing when that happens. Um, so hopefully the logic you know is fairly rock solid. Before you start, I would say that you know there are injuries you have to account for. Rule changes typically don't happen in the middle of the season, so you don't usually have to worry about that. But it is good to review these things, usually at least once a week, to make sure that the metrics that you're monitoring are within kind of a range of acceptability. Obviously, if it looks like it's going off the rails, then you probably should pull it and take a look. 

38:24
I will add to that also that not all models work well during all parts of the season. It's a mistake to think that the entire season is uniform in terms of how a model will perform. There are opportunities at the start of the season when there's less data available that do not exist after the trade deadline, for example, and so you should really test a model, like in the distinct parts of the season, to make sure that this thing is worth running the entire time. Some models you might have that only work in October, november, some might be much better after the trade deadline, and then some might work even better in the playoffs, but but absolutely uh, you know, bomb and during the regular season, and so that's. That's an important uh segmentation for test or tracking bets and model performance as well. 

39:09 - Rob Pizzola (Host)
I'm really glad, glad you brought that up because, um, honestly, I don't think I've ever had a guest that's spoken 185 episodes in that has spoken about something that rings so true to me from a modeling perspective, because for years I had a very successful NHL model that sucked out of the gate and sucked post-trade deadline and I just assumed that it was random variants for the first two or three years. It's like, well, I'm betting all these big underdogs, post-trade deadline here in the NHL you know these numbers are off market is wrong. You know just the casual, stupid things I used to say to myself. And then after a while I really started to dig into it and there's a lot of components at those specific times in the season that are very different. Beginning of the year for NHL, you're working with only previous year's data and there's players that switch teams and they play within a new system that they're maybe not accustomed to and all these things matter. So, yeah, that really resonated with me, andrew, because I've experienced that firsthand. 

40:14 - Andrew Mack (Guest)
That really resonated with me, andrew, because I've experienced that firsthand. Well, I've seen I mean, I've talked to people that have lost everything they made in a season after the trade deadline. There are people that still don't understand that it's not a uniform season and it's a really important point. Some things just don't work when there's enough data out there available. And then there's other effects after the trade deadline too, right, like there is a low key tanking in various sports where some teams they start you know they've traded away some pieces to try to make some moves for the future. There are others that are in win now mode, but that has an effect on you know in who the starting goalie is going to be and what the lineups are going to look like. 

40:53
It has a huge effect on basketball, uh, because basically, like teams are teams that are not competitive by the trade deadline, like they basically just implode Right, um, and so it makes a really, really big difference. If you're, if you're trying to do this the modeling way, um, you have to segment by, and I would recommend months. You just look at, you know, segment by months and try to see where the leaks are coming from in terms of the profitability. 

41:17 - Rob Pizzola (Host)
Yeah, I like to do game ranges, so like game one to 10, 11 to 20, so on and so forth, but there's many ways that you can do it. I also think, like you know, there's and I hate to bring it so hockey specific, but we're two Canadian guys and this is my bread and butter at this point but, like motivation, lots of people just dismiss that in sports as well, where they're like they're professional athletes, they're giving it 100%, and I think, in hockey specifically, there are some late season scenarios that really impact the game where, let's say, teams that are out of the playoff mix, you might get players that are a little bit less reluctant to maybe block a shot on a penalty kill or maybe make that extra back check, and it took me a long time to actually realize that. You know it is what it is. 

42:12 - Andrew Mack (Guest)
But no, no, I agree with you, I think. Um, well, first of all, let assuming that everyone's playing 100 all the time uh, just ignoring injuries for the moment assumes that they're like robots that don't have, you know, feelings and motivations. Not everybody, nobody, can play 100, 100 of the time there are. You know, there are different, different things that go on in everybody's lives and things, and so there's variance in individual player performance, and thank goodness for that, because otherwise it would be an easy thing to solve in terms of a market problem. But I agree with you. I mean, performing at 100% hurts in every sport. You're going to take the hit, you're going to block the shot, you're going to risk breaking your foot in hockey, all kinds of things, and so if you don't have guys that are motivated to go out there and take the hits, yeah, you'll see some serious variance in their performance. There's no question. 

43:03 - Rob Pizzola (Host)
How often are you seeking out new data? And when new data becomes available, are you the type of guy that's immediately going to start digging into it and start incorporating it into the models? I mean, there's several examples of this. Baseball went from, you know, the steamer and zips projections to stat cast very quickly. Hockey is a great example where all these expected goals models started to surface. There's even private data that's available via sport logic that is different from what's publicly out there. How often are you seeking out new data sources? 

43:38 - Andrew Mack (Guest)
Well, I think it would be foolish not to consider new data sources when they come out, because we're constantly playing a game of OutRun, if you remember that video game when, in markets, like, if you do not make incremental improvements, you will eventually be put out of business. 

43:53
So you have to at least have that mentality of like making constant incremental improvements to whatever it is that you're doing. I actually like to try and find the problems with the new data, because frequently there's a chance that new data getting incorporated allows overfitting. So, as an example, I had a theory for several years that the expected goals models were overfitting in. So, as an example, I had a theory for several years that the expected goals models were overfitting in hockey and that there was actually considerable value to fade specific spots when it looked like the expected goals model was favoring a team too much. And so when new data comes out, it's not always the worst idea to try something old and see if you can find significant distortions always the worst idea to try something old and see if you can find significant distortions. 

44:40 - Rob Pizzola (Host)
Love that. The Corsi models that I had outperformed the expected goals models for several years but the market kind of drifted towards expected goals because it's the new and improved and there were actually a lot of data integrity issues with expected goals, where shot locations weren't being tracked properly in certain arenas and things of that nature, and that's also a consideration I always have to. It actually happened with StatCast as well. I can't remember which MLB stadium I want to say it was the Angels Stadium where the data wasn't being recorded properly, and sometimes simplicity is better. Just because you know Corsi, it's like, yeah, was there a shot attempted at goal? It's really hard to mess that up. Now we start tracking shot locations, the speed of the shot, it's a lot harder to mess that up and to really trust the integrity of that data. So I've experienced that firsthand, andrew. 

45:33 - Andrew Mack (Guest)
Yeah, I agree. Um, I, I've experienced that firsthand, andrew. Yeah, I agree, I think that, um, it's always good to at least try, try to find the holes in in the new data or the new method, just to see if, if people are overreacting to something new. Um, that's, that's my kind of go-to general mentality on that stuff. And then the off season normally I would look to see if I can incorporate it, see if it makes a difference, but, um, but there's a lot of value to be found in fading the potential for overfitting when something new comes out. 

45:59 - Rob Pizzola (Host)
I've had a lot of experience with diving into certain betting markets that I haven't had success with for various reasons, but because I find them just extremely difficult to model. Formula One is one for me. I know you mentioned that earlier, but a lot of the practice times in particular are really challenging to understand. Whether there were intentionally bad laps, essentially, whether the car is good or not, there's people that can do it. I'm not suggesting it's not a beatable sport, I'm just saying I've personally had trouble with that. Are there any specific sports or markets that you have avoided in the past because they're too unpredictable or too difficult to model? 

46:48 - Andrew Mack (Guest)
There are. Yeah, just to get back to your Formula One, I helped a reader of my first book out with a NASCAR model, um, using the three point competition Monte Carlo simulation from the first book and and he he did quite well with it. His name is, uh, christian he's. He's on Twitter somewhere I can't remember his handle, but um, that's about the only experience I have with it. Um, looking at kind of Monte Carlo stuff where you're looking at uh, seconds per lap and then the, the variance around seconds per lap and then conditional on the qualifying times. And I've also fielded some questions over the years about translating that into horse racing and a Greyhound racing. I've never touched that stuff but uh, that's. That's about the full extent of my knowledge on those particular things. 

47:35
There are a couple of sports that I'm not really interested in anymore. Baseball is one of them. When StatCast came out, it became a bridge too far. I just sort of said, yeah, I'm going to back off of this, and actually it's been really nice because it gives me the summer off of betting, which allows me to make improvements to the other models and gives me a little bit of a mental break while I just stick to trading the other one that I don't get into much is soccer, english Premier League, scottish League, those kinds of things, mostly just because I feel like with hockey, basketball and football I'm pretty overextended as it is and I've made I've made mistakes in the past where I thought I can take a model that is working somewhere and I can sort of shoehorn it into a new sport with a few tweaks and and it never went well for me. The lesson I learned from that was you know, don't don't be arrogant and think that each sport doesn't deserve, you know, its own level of respect in terms of trying to attack it. 

48:34 - Rob Pizzola (Host)
Yeah, I like what you said about baseball there with StatCast, because I felt the exact same way. First it was an exciting problem and then, after a while, burnout almost set in with trying to figure all of that stuff out. And when I reached out to you for this interview, andrew, I hope you're okay with me sharing this, but you basically said that you watched or listened to the solo version I did of Circles Off recently and that that personal story resonated with you a lot. I want to talk a little bit about the psychological aspects of betting, because long-term betting requires a lot of effort, persistence. There are highs and lows. How do you stay motivated with sports betting right now? 

49:18 - Andrew Mack (Guest)
Well, it's true, you know, I thought that that was a very brave episode for you to do, because not a lot of people would be that forthright in front of a camera about, like, the reality of sports betting, and the truth is that you will get burnt out, like this particular vocation lends itself to burnout. Number one, because there is always something else you could be doing, the work is never really finished, and also because it's it's an inherently solitary activity, relatively speaking. Those two things you combine together and then you don't get enough sleep and it's very, very easy to burn yourself out. Staying motivated is, I guess, a slightly different issue. I would say that for me, like you know, I was a journeyman electrician for many years. I've worked some pretty awful manual labor jobs in my life, so for me I stay motivated mostly by having a sense of gratitude about the opportunity that this is and remembering that the worst day that I have, betting, is better than many of the best days I had on a job site, on a, on a ladder, you know, 30 feet in the air, twisting wires together. 

50:23
And so you know, I've been. I've been outside, uh, in a rock climbing harness, dangling off the side of a building. Uh, in minus 35, right it's. Even if you're having a tough time with a model, you're inside, you're at your computer chair, you've got a cup of coffee. I mean things are pretty good. 

50:45
But I guess, getting back to the larger picture of that thing, there are going to be ebbs and flows in terms of your motivation, your joy, your, your excitement. But it's also unrealistic for anybody to think that you know something that is like more of a marathon than a sprint is always going to be like this wonderful, exciting time. The truth is that when you get good enough at this to make money, it's going to be much like a job. It's just going to be a really great job because you have autonomy over your personal time. You have the opportunity to make a lot of money and do something that is intellectually challenging and stimulating and rewarding, and so you try to develop a sense of routine, you try to develop a sense of perspective that allows you to just keep putting one foot in front of the other. 

51:30 - Rob Pizzola (Host)
Yeah, it's all a matter of perspective a lot of times as well. And yeah, I wouldn't be wanting to be dangling off the side of a building in negative 35. 

51:40 - Andrew Mack (Guest)
Celsius either. I've had some bad jobs, but that sounds like my nightmare actually, yeah, you don't wake up for days like that, being super stoked, right, you just do them and do them and you go home and, uh, you know, if you can handle that, though, you can handle a bad day, you know, in the NFL, right? 

52:00 - Rob Pizzola (Host)
Yep, totally agree. Um personal question here. But what's the biggest psychological challenge that you have personally faced as a sports better? 

52:11 - Andrew Mack (Guest)
Um, personal, um, personal, that's a really good question. 

52:24
I um, I suppose the burnout, like, is a real factor, like, even when you like what you do, um, it can, it can be exhausting, um, and especially, I would say especially when you're getting your feet under you, when you're building models, when you're trying to build it like some semblance of a business for yourself, you have to work a lot harder and a lot longer than you would at almost anything else. 

52:44
Like, if somebody told you when you were starting out how much work it was going to take to get to where you are now, you probably never would have even started, and so the amount of work can really, um, it can, drain on you. In the beginning, um, and certainly I I dealt with burnout where, you know, I was just trying to do too many things at once all the time and and then you collapse for a day or two and sleep 10 hours a day. But, um, I would say that that that was a pretty big psychological challenge. I had to learn to sort of structure my days with a routine so that I had like artificial constraints where I would get away from the screens, get away from the computer and sort of connect back with the real world and let your mind kind of wander a little bit. I found that that was really really helpful. 

53:29 - Rob Pizzola (Host)
But when you're starting out and you're still trying to build some kind of a baseline for yourself, you don't always have that luxury, and so I think the burnout factor is very real are you competitive by nature, and the reason I asked that is because, um, I think we share a lot of of similarities in the sense that I suffered burnout in the past several times before, but in sports betting I became obsessed with being the best at my craft. 

54:00
Right, it's like I want to be the best NHL better, and if I'm not around at my computer in the afternoon from these hours I might I might lose a piece of goaltending news or some random injury or coach says something about this, and that's going to eliminate one play that I could have had. And it became this obsession in maximizing Evie. It's like if there is an edge, I want to have it for every single game. I'll do whatever it takes. And then, in hindsight, that was really dumb. It was really dumb of me to put that much time in. It wasn't worth the amount of effort, but it was because I was so competitive in nature and I'm curious if, if, you're like that as well, or if, or if you don't really think about those things. 

54:47 - Andrew Mack (Guest)
I'm definitely like that. I've had to, like I said, I had to enforce artificial constraints on it to create a routine for myself that was livable, recognizing that this is more of a marathon than a sprint because, like you, I will w without like some kind of a routine. I will sit at the screens for 16 hours a day and just like attack everything that I can find and at some point your health starts to suffer, your, your psychological health starts to get a little squirrely and um, and then all of a sudden your plus EV edges start to look like minus EV life. And I'll relate a story that I had from the job site actually there was, because it sort of relates. I was on one job site one time where there was a guy that an electrician that walked around and always made fun of other people's work. You know that looks like dog shit. Who the fuck made this? You know whatever. And um, I was walking past a couple of guys talking about this guy one day and it's like so-and-so thinks he's the best electrician. There is no best. Either the job gets done correctly or it doesn't. 

55:50
And you know there's a lot of truth to that. Right Like you're either getting what you want out of this or you're not. But the idea that there is like an objective, best like we have a leaderboard out there on a website and it's like, uh, you know, you're moving up the ranks. It's like you're getting what you want out of this or you're not. You're making the amount of money that you want to make or you're not. You're living the kind of life that you want to live or you're not. And if you're achieving your goals, congratulations, you're winning. And if you're, if you're not achieving your goals, then you got to do some work to to create that. But sometimes that means working harder on the models, and sometimes that means spending more time with your wife or your kids or your family, et cetera. Right, and so there's there's sort of like a broader scope of of what your goals are. 

56:33 - Rob Pizzola (Host)
Yep, that's really well said, really well said. Do you have a favorite sport to model or build models for, and why would that be the case? 

56:43 - Andrew Mack (Guest)
You know, for the longest time, I probably would have told you basketball. I like basketball because the distributions are relatively smooth compared to other sports, so it lends itself very well to modeling, which is both a good thing and a bad thing. I would say that more recently. I would also tell you that I really like modeling football, just because the clustering of scoring around key numbers lends itself to a lot of really interesting effects that are very actionable, and I've found that very intellectually, intellectually, um, interesting and also profitable, um, so those would probably be the two that did come to mind. Um, yeah, I mean, I like hockey too, but they all have their own. They're all equally annoying for different reasons, and and that is actually a really good thing, because every time there's some pain in the ass to get something figured out, that's a barrier to entry, and every one of those that you can jump over puts you in a better position. So, ultimately, the little annoyances for each of these sports that make them unique are wonderful things for someone that's trying to model. 

57:47 - Rob Pizzola (Host)
Yep Agree with that. Can you share a time when one of your models either performed exceptionally well or failed completely, and what you would have learned from that experience? 

58:03 - Andrew Mack (Guest)
um well, maybe I'll share one where where it failed completely, because those are always more entertaining I was hoping for that. I was hoping for that. So so, right, right when covid hit, I was. When COVID hit, I pretty much decided I'm going to take a run at doing this for a living. 

58:22
And basketball went squirrely right Because Rudy Gobert breathed on a microphone and all of a sudden, everything went completely nuts right, but now you have lineups that are just going completely berserk. You don't know who's playing, people are getting pulled into the COVID protocol like 10 minutes before the game starts. And so I thought, well, this is actually where of my hockey models and I'll rework it so that it can work for the English Premier League, because the Premier League was still going. And I thought, well, this will be fine, right, this will be fine. 

59:05
And I was totally arrogant about it. I didn't give it the respect it deserved at all and I absolutely got my ass handed to me over the course of about a month and at the end of a month I was like, yeah, no, no, thanks, take, take EPL back. I'm, I'm good, um. But I guess what I learned from that is is just not not thinking that success in one particular modeling type or sport is going to automatically give you success anywhere else, that you need to approach each one as a new project and, while you can take some of the things you've learned, if you just lazily try to, you know, shoehorn something or don't think that there's going to be, you know an incredible amount of work you're going to have a bad time, and I certainly did Now. Now I stick to this stuff that I'm having success with, which is is going much better. 

59:53 - Rob Pizzola (Host)
I I do the same. Although you know, listen the failures, they are regrettable in the sense that you know you've wasted time, you've lost money, but I actually do. I mean this is absolute cornball cheesy stuff, but like the whole notion of what doesn't kill you makes you stronger type of thing, there are valuable lessons in failing. I firmly believe that. In fact, I don't think I would have gotten to where I did in sports betting without the failure story at the beginning. So I mean I like hearing that I do. Yeah. 

01:00:28 - Andrew Mack (Guest)
I think there's value in training yourself to take losses like not on purpose. You don't quote your way to lose on purpose, but the truth is the first couple of times that you lose a significant amount of money to you, whatever it is at the time, you're going to have an emotional, like visceral, reaction to it and you can train yourself not to not to have that kind of a reaction Like you can. You can train yourself to be more Zen, more detached over time. That is valuable because it allows you to stick to a system or an angle or a model that's working, even when you're experiencing variance, because what most people do when they experience that is they start immediately changing things in a panic mode or they hop to a different sport. They do all of these things that are very emotional. That ultimately will just keep you going around in circles. So I agree there is value in learning to train yourself from having those kinds of reactions, as much as they're not fun when you're going through them, and you're quite sure that you'd rather not. 

01:01:27 - Rob Pizzola (Host)
Yeah, listen, I mean just to expand on that. I do live watch-alongs for NFL primetime games on the Forward Progress channel that people watch and I often get people who message me on the side and they're like ah, you know, did you actually have a lot of money on that game? Are you saying that you did? Because they don't get the reaction from me that they might get from other content creators, who are like taking the TV off the wall and smashing it on a Hail Mary or something like that. And I really just credited that to literally losing on sports for a decade. And now I win and I know that I win and I I'm almost numb to the pain. It's really weird to say like you still experience, you still don't want to lose and you'll go through prolonged losing streaks where you're like, ah, this kind of sucks, but you don't get the reaction out of me, where I'm like I'm just absolutely losing my mind. 

01:02:16
I watched a live Cowboys Bengals game Monday Night Football where I had the um, the Cowboys plus six, and the, the. There was a, a punt that was blocked. Basically all the things that needed to happen in that game were preposterous for me to lose the spread, but I just kind of sat there in silence for about a minute because of previous experiences with losing at sports, and I'm actually thankful for that, because I see other people react nowadays to losing streaks. I can see the people who are messaging me like 25 times on Telegram sweating out the last two minutes of a game like everything is on the line for them and it sucks. And had I not been a losing better before, I think I probably would be wearing the emotions on my sleeve a lot more than I actually am now. 

01:03:09 - Andrew Mack (Guest)
Yeah, it affects your decision making and since the entire nature of the game is is making a good decision when you're presented with an opportunity, um, the worst decisions you will ever make as a better are going to come after a massive win that produces an emotional reaction, or a massive loss that produces an emotional reaction. So part of you know the operator skill is smoothing out the waves so that you don't have big emotional reactions, either positive or negative, so that the next decision will also be a hopefully a good one. Um, you know, and if you're ever feeling bad, just remember that somebody picked the colts to win the other day, uh, and then the broncos absolutely embarrassed them I talk about having. 

01:03:53 - Rob Pizzola (Host)
I had under 44 in that game now. Now someone might say, well, rob, you could have got a 44 and a half earlier in the week and they would be correct, but I got 44. It closed below there. I actually don't think either team averaged more than three and a half yards per play. In that it is one of the most frustrating losses. Denver 3.2 yards per play. Colts 4.3. Success rates 30% and 35% respectively, which are extremely low for those who don't. Anyways, I don't want to relive my bad beats on air here, but yeah, that was not a fun one. 

01:04:26
Shifting gears here. Just to the future of sports betting and modeling. How do you see sports modeling evolving over the next five to 10 years? 

01:04:35 - Andrew Mack (Guest)
Yeah Well, I expect that machine learning type of stuff will continue to grow in prominence. Obviously, it's already a big deal. I think that if you're not at least familiar with concepts like random forest, xgboost, light, gbm, you're already probably four or five years behind. So it's already a big deal. It will probably continue to be. I would expect that we'll see some trends, like we see in the finance markets, which will include alternative data sources, which are a big deal in finance these days, and you'll also see a little bit of AI stuff coming into prominence. I would expect that those would be the two general meta trends of the next little while. I'm sure there'll be new algorithms and things like that too, but those would be the main two things. 

01:05:23 - Rob Pizzola (Host)
Are you experimenting a lot with AI and machine learning now? 

01:05:27 - Andrew Mack (Guest)
I do quite a bit with machine learning Again sort of broken record. But the machine learning algo is not the magic. The magic is the effect and the machine learning algo just helps you to identify that when you have a tremendous amount of data, where it's not an easy thing to do to build a model by hand, right, and when there is an effect it's great, it will help you to find it. When there isn't an effect, you know you've done a whole bunch of linear algebra for absolutely nothing. 

01:05:56 - Rob Pizzola (Host)
Yep, yep. I've seen that firsthand before, with people who just throw as many things as possible into like this black box and hope that the model is going to spit something out for them that's going to be valuable and guess what? It never is. Yeah, interesting For someone who's not strong in math or in computer programming. Do you think it's still possible to leverage statistical models, do you? 

01:06:20 - Andrew Mack (Guest)
think it's still possible to leverage statistical models. I mean yes and no. I think that you could. Maybe a better question is if you would like to take this seriously and make money and you've also decided that modeling is going to be part of your toolbox, what exactly is the rationale for not learning what you need to learn to be good at it? Because I would say that, as with most things in life, one of the keys is to figure out what the price is for success. However, you define that and then be willing to pay the price. 

01:06:50
Uh, when I you know I didn't know how to program, um, you know what? Five or six years ago, right, I was doing mostly Excel. I learned how to program in R, I learned Python, I learned a database infrastructure and all that web scraping, and I did that not because I had this burning passion to learn how to program. It was because you could see the writing on the wall that this is the way the industry is going. Marco Bloom from Pinnacle, I think, in like 2014 or something, talked about using R and machine learning algorithms, and so you knew that this is the trend right, if you want to remain competitive, you either are going to do what it takes or you're not. I would suggest that if you want to take this seriously, it's worth taking the time to do it. 

01:07:37
Can you benefit from these things without not knowing a lot about them? You can, but it makes it hard for you to determine what is a good thing to be incorporating into your process versus what is a garbage thing. You you'll have no way to discern value from noise. Um, you're just looking at other people's models. You're looking at moneypuckcom, you're looking at whatever and you're deciding this. Looking at whatever, and you're deciding this is this is good or this is not good, but you have no basis to make that decision and that that means that it's going to be a bit of a crapshoot, right? 

01:08:08 - Rob Pizzola (Host)
yeah, uh, yeah, completely agree. Uh, you mentioned moneypuckcom and I laughed. I actually like the site in terms of the layout and things of that and just like visuals for other people out there, but they do have these pre-game win percentages up there that some people use as gospel and I'm like I I wouldn't uh be avoid avoiding these public models as much as possible. 

01:08:30 - Andrew Mack (Guest)
Uh, in all likelihood, go ahead I went to law school with a guy that that bet hockey off money puck, like while he was in class. Right, he's still a lawyer, you know. It's just saying. 

01:08:44 - Rob Pizzola (Host)
With R and Python and learning a programming language. A lot of people find this overwhelming. So you're a great example of someone who just kind of taught themselves this because they saw that you needed this skill, to acquire this skill, in order to be successful what you wanted to be successful at. How did you teach yourself um? Just did you take any online classes? Did you read books? If you could go back to um how you started to learn a programming language, how did you do it so? 

01:09:13 - Andrew Mack (Guest)
the way that I did this in the beginning probably wasn't the best, but I followed my enthusiasm, which, um which at least got me to where I am now. But I started out basically realized machine learning models are the trend, so I want to learn about machine learning models. So the first course that I took in R without knowing how to wrangle data or clean data or anything, was a machine learning models course, which sounds kind of insane but did you did. 

01:09:39 - Rob Pizzola (Host)
You find that online? Andrew Like did. How did you even come across that course? 

01:09:44 - Andrew Mack (Guest)
I did. It was online, um, the guy's name is Jason. The course is called machine learning mastery. It's like an ebook with uh, with code snippets in it so that you can basically run the code and then like see what's happening. Um, so basically I used the, the dummy data that came with it, and I ran these models and that's how I I first started using R and I I'll admit the first time that I stared at a blank console, like in our programming. I hadn't looked at a blank console since playing the Oregon trail on a Commodore 64, you know like it had been a long time since I'd looked at a blank console. But, um, so I ran that stuff and sort of got okay, this is how it works, whatever, and actually then learn data cleaning and wrangling afterwards, which is generally backwards. 

01:10:27
Most people would tell you that you should learn the basics first, um, but that's how it started for me. And then I enrolled in an online master's degree program. Uh, so I did an online master's in data science with James Cook university on Australia, um, but ultimately it was pretty expensive and I don't think that it was necessary, because that particular program is about training data scientists to get employed, so they teach you a little bit of R, a little bit of Python, um little bit of database, little bit of Stata, a little bit of um, um, what's the other one, the other main? Anyway, they, basically they teach you enough, just enough about all of these different, uh, data science interfaces that you could be gainfully employed with someone else. And really what I wanted to know was how can I use R to make money at SwapArts, right, um? And so you end up. You end up kind of self-teaching most of the rest of the way, and there are some pretty good sites for that. There's Coursera. 

01:11:26 - Rob Pizzola (Host)
There is geez, what's the other one called. God forbid you recommend a class to anyone out there, andrew, by the way, I just you know it's not. It doesn't get received. You know you can find all this information for free online, but apparently getting a tightly packaged class is a real problem in this industry nowadays. 

01:11:43 - Andrew Mack (Guest)
Well, yeah, you know, people are funny about that stuff, from what I can tell just seeing the chatter online, the average sports better could probably use all the help they could get. So I'm confused about the level of heat that you guys took for that. I don't really understand it myself. But yeah, there are a lot of different places that you can self-teach these days and you should like. This is a skill that is very applicable for a lot of things. And getting back to your course congratulations, by the way, I would find it hard to believe that a course that features you and Jack and Rufus and Gina and I forget the other people involved but I find it hard to believe that a course involving those people would have no value for anyone. Like I just don't. I don't believe that. So I don't understand where where all the heat is coming from. But I got a lot of the same stuff Every time I release a book. People have really nice things to say in my comments too, so it is what it is. 

01:12:38 - Rob Pizzola (Host)
I talked about this a bit on Circleback on Tuesday, so I don't wanna I won't dive into it too much. I understand the heat and I get it. It is a very basic entry level course which I think benefits like 98% of the betting population on it, people who don't understand what a betting market work, how a betting market works. This was me in my 20s. It's a course basically designed for me in my 20s. And sure, some of the personalities in there are not sharp, they're not intended, they don't portray themselves as sharp, they're just there to connect with the audience. So, anyways, I digress, I'll move on. I don't want to make this all about me, but the sharp section of gambling Twitter, twitter god forbid you put something out there that's behind a paywall. 

01:13:24 - Andrew Mack (Guest)
You will just, uh, suffer the consequences I mean, in fairness to what you just said, right, um, if you're looking at the stats that are that are being released, like the average better in the us is losing money at a rate that is unheard of right in terms of the um, the the level of minus EV thinking and decision-making out there. If there's a course out there for a hundred bucks that basically explains to them all the things that they're doing, that are guaranteed to lose the money, I don't see how that's a bad thing personally. 

01:13:53 - Rob Pizzola (Host)
Yeah, I mean. The course technically is on sale for eight bucks a month, which could be canceled after the first month. This is what I got grief over, by the way, with other courses. Anyways, I digress, we will move on. I promise I don't wanna make this all about me. Appreciate everything here today. Andrew, it was great talking to you. Before we move on to our final questions here, where can our audience find your books out there, or any other resources that you might be offering right now, or any other resources that you might be offering right now. 

01:14:20 - Andrew Mack (Guest)
Well, all of my books are available on Amazon worldwide. You can find them in the link in my bio on Twitter, or X, I guess, as it's called now. 

01:14:32 - Rob Pizzola (Host)
No one calls it. X. You don't have to give in. You do not have to give in. It's always going to be Twitter for us. 

01:14:39 - Andrew Mack (Guest)
Well, we'll keep it with twitter, then yeah, that's that's where you can find them. 

01:14:42 - Rob Pizzola (Host)
They're, uh, they're there if you're interested all right, let's uh end here with a few questions. We'll start with plus ev and minus ev, so this doesn't have to be related to sports betting. It could be related to anything in life, something that you think is plus ev, something that you think is minus ev something that you think is minus EV, sure, plus EV. 

01:15:09 - Andrew Mack (Guest)
You should outsource things that are not part of your day-to-day business operations to increase your efficiency. So a lot of people they imagine, if they're making a lot of money, that they would then hire a data scientist or something to like increase their ability to make money at sports. Right, and it's the exact wrong thing to do. You don't want to give away anything that's proprietary that you're doing. You want that to remain pretty tight to your chest. What you should outsource are things that will improve the quality of your life that are not directly related to the day-to-day operations. So, for example, get yourself an actual accountant and not like some CD, like backroom. Like get yourself a chartered accountant that works at a firm that knows all the changes and all the um deductions that you're eligible for and you send, you send all your documents to them and they deal with everything. That's an amazing, amazing upside. 

01:15:59
Another example uh, house cleaning services. Right, if you value your time at a hundred bucks an hour and you can pay someone you know 20, 30, 35 bucks an hour to clean your house and you save yourself three or four hours um, doing that massive upside to that because you get to stick to doing what you're good at and everything still gets done around the house. Um, I think that most people have that backwards and they think about outsourcing the stuff that they're actually doing, which makes no sense at all For minus EV. Let's see how about arguing with people online? 

01:16:37 - Rob Pizzola (Host)
This has been used on the show before. It's not proprietary to you, but it can be reiterated. I think it's a good one. 

01:16:44 - Andrew Mack (Guest)
So the way that I look at this is that there are two groups of people that are saying really, really inflammatory or dumb stuff online. One group of those people saying them on purpose to drive engagement, and if you play into that, you're just helping them achieve whatever their end is. I assume that's making money on on Twitter. Um, now, but but. Again there's no dollars added to your bank account and you've wasted precious time that you could probably spend doing something else. And then the other group of people are just they. They're very confident and they don't know anything and you're not going to convince them, you will not persuade them and again you're just losing precious hours that you could be doing something much more constructive with. So I generally look at that and say there's not a lot of upside to trying to correct someone that is wrong and extremely confident People that are open to having their minds changed will ask questions. They will not assert statements that are ludicrous. 

01:17:45
That was my general perspective on that. 

01:17:47 - Rob Pizzola (Host)
That was very succinct and I really liked the way you. I actually might just cut that clip and play it back to myself once a day, because I fall into that trap as well. All right, let's end on this. Andrew, If you could go back five years and talk to a previous version of yourself, what advice would you give to your former self? 

01:18:07 - Andrew Mack (Guest)
There is no edge in a model that does not identify and quantify an effect. Modeling is not an inherent edge in itself. It is only a means to an end it andrew mack. 

01:18:20 - Rob Pizzola (Host)
Everyone follow him on twitter. On twitter at ginge face killer again, link in the description below for andrew's twitter. Check out his books as well. I do recommend them. I actually I got I'm. I'm not gonna lie, I didn't know that you actually had a book in r, so that's going to be the next thing that I recommend for people who've been requesting that, and ding me on the side as well, but I have read the Excel stuff and highly recommend it to anyone else that's out there. If you enjoyed today's episode, smash that like button down below. Make sure you're subbed here on Circles Off. It goes a long way for YouTube recommending us to other people as well. I see a lot of people in the comments every week that they just found us. Thank you very much for finding us. Make sure you hit that subscription button down below. We're back next week for a brand new episode. Johnny will be here in studio with me. Peace out. Thanks for watching Circles Off.