Transcript

Andy:
All right. We are live with, I think it’s the third installment of our weekly strategy sessions. This week we’re going to be diving into the two most popular Esports, League of Legends and CS:GO. My name is Andy Baldacci. I’m the CEO of SaberSIM. We’ve got Jordan who is still working on the title. We can call him the coach. We can call them marketing strategists. We call him many things, but he’s usually running the show with these office hours.

Andy:
We also have Max and Wil. Max has, you didn’t do a ton of Esports until COVID happened, but you really dove all in and have a pretty strong background for League of Legends, specifically. I don’t know how much you’ve messed around with the other ones. Then Wil is one of the newest additions to our team. He is a full-time data scientist for SaberSIM, works with us to improve all of our models. One of the big things that we got in addition to Wil was when he joined he also brought with him his CS:GO model. This is, I don’t know if, maybe Rog is listening to this, but this is, I would bet anyone, by far the best CS:GO model that is out there. We’ll can talk a little bit more about that.

Andy:
Yeah, that’s the team. We’re going to be just jumping into League of Legends, CS:GO strategy. We’ll start with more of the basics and just work our way up to some more advanced questions and see if we can level up all the way to some of Maverick’s really intense meta stuff. Yeah, do you want to start things off, Wil, a little bit and just talk at a high level, how your CS:GO model works?

Wil:
Yeah, so I only really started getting into Counter-Strike around the pandemic as well, just when there wasn’t much to do. And so the way my model works is actually inspired by, because I’ve obviously been a SaberSIM user for longer than I’ve worked here, trying to replicate a simulation style in a smaller, more condensed setting. In Counter-Strike there’s a lot of different variables. You can’t really simulate if a team is going to run up the middle or run to the right. That that much data isn’t available.

Wil:
The way the model works is by looking at stats and different maps and different win scenarios and different outcomes, and looks at a whole bunch of different variables and then creates different simulations by randomly sampling those variables. For instance, if a really good player, like a S1imple or ZywOo, who’s one of the elite players in Counter-Strike, they may average 25% of their team’s points. But if they have a really good map, they could get up to 35%. If they have a really bad map, they could get down to 15%. It’s just about taking as many of those different complex variables where we know what the average is and how different variations can come together and create a wide range of outcomes and really see how a slate might play out.

Andy:
One of the coolest parts to think about it is you also simulate the picking the maps and the banning and all of that. Can you explain that a little bit more?

Wil:
Yeah, so I actually created …

Andy:
Well, I guess first, can you talk about just what that process is like for someone who might not be familiar with CS:GO?

Wil:
Yeah, so the way CS:GO works is there’s a map pool of seven maps. Each series, they call it, they play a best of three series. First to two wins out of those maps. The way it works is each team gets to ban one map right ahead, so they get to remove their weakest map. Then each team gets to pick their favorite map of the remaining. Then they ban two more. Then there’s the leftover, the last one. And so I thought there was going to be a lot of edge in that when I first started in figuring out not only what maps are likely to be played but also how that affects win probability and how that affects the different players statistics.

Wil:
I’ve created a model that essentially calculates a team strength on each given map. And so with that, I can estimate, there’s 70% probability to win on this map and 55% to win on the next one. With that, we can figure out the likelihood that they’re going to pick their stronger map first over their second map and using that combining their opponent strengths and everything like that, I’m able to get a pretty accurate look at what maps are likely to be chosen and how that might play out for the contest.

Andy:
How [crosstalk 00:04:24].

Max:
Can I also just add, Andy?

Andy:
Yeah.

Max:
Could I just add something really quick is and you can clarify this for me, Wil. Maps are kind of like, it’s like parks in baseball too, as well, right? Some maps are smaller, so teams will be more aggressive on them or bigger or there’s factors that the map has influenced on the team matchup dynamics too, as well, right?

Wil:
Yeah, absolutely. So that’s a big thing. It’s constantly evolving, I think, very similar to League as far as how that meta works. There’s a bunch of maps where there’s certain sites that are harder to try and recapture from your opponent. That usually results in teams not even attempting it, which lowers the kills for the winning team. You have an interesting dynamic of a team that consistently picks a map with higher save rates. They’ll end up getting lower kills despite a higher win percentage. It’s those little nuances that can really impact projections.

Andy:
I mean, this is the perfect person asks this, because you actually just did a lot of work on the park factors for our MLB model. Compared to a sport like baseball where parks have a noticeable, measurable impact, how I’m assuming it’s a bigger impact that the maps have in CS:GO, how does that look? How much does that actually impact things?

Wil:
Yeah, so it’s slightly different in the fact that teams specialize in specific maps. A baseball team, there’s impacts in the park factors, but by and large, they’re usually a similar skill level in all parks. Whereas, a team on their favorite map pick versus their least favorite map pick can be a wildly different game. I mean, they could be an elite best team in the world on one map and look horrible on another, so they’ll just always ban it. They run into certain issues when they face teams that always pick that map or there’s certain game theory in it in figuring out where teams are trying to exploit where their opponents are weakest and where they’re strongest.

Andy:
And one thing, so Jordan, I’m actually going to go call a little bit of an audible on this. I think what will be easiest for people to follow is if we talk about CS:GO for a bit first and then transition into League, I think bouncing back and forth between the two might get a little bit confusing. Just kind of jumping into it, Wil, this is something, I know with our sports, that we have all the simulation data for in the builder and that we’re automatically counted for correlation and variance and everything else, we try to avoid rules of thumb. With CS:GO, we do have simulations powering the projections, but we haven’t fully incorporated that into the builder yet, and so with that in mind, what are you looking for, just at a high level, for what makes a good lineup? Are there rules of thumb that you follow when building your lineups for CS:GO?

Wil:
Yeah, for sure. I’m a firm believer that pretty much regardless of the slate size that you want to at least force one three stack in there. Most of the tournaments nowadays, with the current this summer season, a lot of the slates that DraftKings are offering are two games slates. And so this will be four teams, two games. There’s really strong correlation in Counter-Strike. It’s usually you just want to have teams that win. There’s obviously occasional times that you’ll have a losing player that scored really well, but there’s a really strong correlation in the winning teams and negative correlation to the losing teams. Forcing a three stack is something that I always do. Even from two game slates up to six game slates, I think I always have 95% of my lineups at least have a three stack force in them.

Wil:
I’m a strong proponent also of just running a three-three, which is two three stacks. The total lineup is just two teams. I think those are often overlooked as far as upside in VPPs. People try to get too cute in my opinion. They’ll play two one stacks. Then you’re introducing too many variables. Usually, maybe having a higher ceiling, but they’re sacrificing, they’re like 75th or 85th percentile, which is usually all you need to win a 2,000 or 2,500 person contest.

Andy:
For people that aren’t as familiar with these lineups, yeah, you’ve got, it’s pretty standard what’s for other sports to showdown format where you’ve got a captain and then you have five flex positions. For CS:GO, they don’t, people might have different roles, but in drafting in FanDuel, there’s one position. That’s something to keep in mind. With that weird component to this, what are you looking for in choosing your captain? Is it truly just whoever has the highest projection? Are there different stats you’re looking for that these types of players are better fits? What are you looking for when choosing your captain?

Wil:
Yeah, so I do a lot of different things as captain. I’m usually pretty spread out. I think most often the two, I guess, under owned areas are usually the top person on an underdog. For instance, if there’s like a plus 150 or plus 200 dog on the slate, like today there was a game called Sinners versus Entropic, and Sinners was a plus 175 underdog. Their best player, Oscar, I think he came in at 11% captain ownership. If they win, like they won the first map, he scored nearly 40 points just on map one, which set him up really well. I think plays like that where it’s the top person on an underdog or alternatively, if you want to play the favorites, playing one of the lower projected players on the favorites that often get overlooked.

Wil:
Pricing is typically pretty weak on these and that leads to a lot of crowding around the top players. I took a look at tomorrow’s slate and you can fit in pretty much three of the top five players of Counter-Strike today in one lineup, which is most likely going to be the top performing lineup, but ultimately you can get quite a bit of EV in getting unique and taking a slightly worse player from one of those teams and fading the chalk.

Andy:
On that note, yeah, do you want to? I think we’re going the same direction, Jordan.

Jordan:
Yeah, I was just going to say, I mean, can you talk a little bit more about what your process looks like in terms of assessing what the chalk might be heading into a slate and also how well, you mentioned it a little bit, but how well the field does at finding the best plays and how much ownership condenses on CS:GO?

Wil:
Yeah, so I think there’s a typical, I think people are still underestimating the variance in CS:GO across the board. When I’m building, the way I do it is I simulate each contest for Counter-Strike. I’ll attempt to estimate the actual lineups that my opponents are using. When I’m doing that, I’m basically, instead of using my true skill odds that I have generated to assimilate the map veto or anything like that, I’ll replace those with the Vegas odds, because that’s what drives most of the projection systems that I’ve seen out there, because they have that strong reliance on Vegas. I’ll use those to generate a pool of potential lineups for the field.

Wil:
But from that point on, I assume that they’re going to play perfectly optimally. I assume that they’re operating on the same data that I am, which is probably overestimating the field at least a little bit. But I think it’s safer to assume that they’re going to play it perfectly than try to assume what mistakes they’re going to make. Because I do think that we see the field evolve pretty quickly to changes. I think the biggest thing that they miss is just understanding variance.

Andy:
When you say they don’t understand variance, is it as simple as they don’t think it has much, and that’s why like those plays are good or is there some component to what some of the distributions look like that you focus on more?

Wil:
Yeah. So I think, the biggest thing is that pricing, if people are just going off of a pure projection system, you’re typically going to get the top plays, the top players, like tomorrow there’s a guy named ZywOo who’s been three years running the top player in Counter-Strike. There’s a lot of times where they’ll just be the top productive player and then there’ll be owned 85% or 90% on a two game slate. There’s really only like a 75% chance that he’s going to get you the points that you need, which without looking at ownership, 75% chance of getting those points is fantastic. But he gets put in so many lineups that there’s quite a bit of EV, if you can build a solid lineup without him and just get that 25% dice roll that the field is typically missing quite a bit

Andy:
On a similar note, when it comes to unique lineups, that has to be a major component, especially for a lot of these are smaller slates and there’s just not, there’s not big teams, so there’s just not that many possible combinations. What are you doing to try to encourage unique lineups? How do you find the trade-off of, okay, this lineup is unique, but it’s also viable? What are you looking for there?

Wil:
Yeah, I think it gets a little bit tricky. I think that it goes back to what I mentioned on captain, where if you’re playing an underdog, playing the top projected players from the underdog, because it’s unlikely that a team is going to pull off an upset if their two best players are awful that day. It certainly can happen, but they’re the keys to success. I think the other flip of that coin is on favorites, the people that are lower projected, they typically have worse roles on the team. They’re typically just sent in to gather info or they just run out and die instantly. They definitely have their games where they go out there and they kill everybody before they die, and so they have those high ceilings and they can score more points than than their projection might lead most people to think just because their averages are so low.

Andy:
Okay, and what I’m trying to figure it out …

Max:
I …

Andy:
Yeah. Go for it.

Max:
Sorry. I’d love to button with the question, which is you know, and I’m not as familiar as CS:GO as I am with League of Legends, but I feel like it has the same dynamics, which is that correlated, like he said, the correlations between players on the same team, it’s really high, way higher than baseball, way higher than football, way higher than basically any sport that if you only play the big four that you’re going to be familiar with. This is true for League of Legends. I’m pretty sure it’s true for CS:GO.

Max:
So in terms of, if we’re playing a GPP, what is the thing that we’re heuristically trying to do? I heard you mention, these teams where they’re underdog and they end up winning. You can have some really high upside games. Is that something that you’re looking for where you’re just basically looking for the most under the radar team that could win and that win would propel you to first? How are you looking at that? How’s the variants work in real time? It’s League of Legends and I think CS:GO as well, the win-loss dynamic is really big, right? If a team wins, it can make a player one of the best players on the slate, whereas his projection might be very low, so could you explain that a little bit?

Wil:
Yeah so I think it definitely, it hits on a lot of those same beats. Pretty much the top three players of a winning team are going to be being viable. I think that’s a product of, right now countership contests, they’re typically 5K to first and sometime between 1,700 and 2,500 entrance. It’s a GPP with a nice big prize pool, but it’s not hitting all of the possible combinations that are out there. There’s typically 100 or 200 lineups that would beat the, what ended up being, the winning score that day just by virtue of different ways that you match them all up and people aren’t covering them.

Wil:
I think it’s, for me, I typically just trust my process with looking at who the top players are on those underdogs. I typically have my highest captain exposure is typically either the best [inaudible 00:17:12] and underdog or the fourth, third best player on a favorite if I happen to like both favorites more than Vegas. That’s typically times that I’m overweight on them. I think I’m typically just going for getting the wins and hoping that chalk busts in some manner. I think that’s the easiest path to victory and the one that most commonly comes up when I assume the slate. We’ll get something where it’s 2% probability that a lineup will win, which is pretty nutty considering a hundred thousand possible lineups. It’s just a factor of pricing being really weak, and all that sort of.

Andy:
One thing I just want to touch on there is that with your model, to be clear, you are not pulling in Vegas factors into the model directly.

Wil:
Mm-hmm (affirmative).

Andy:
Whereas, I would, I’m very confident saying that other projection sources for the, especially for the non-major sports are based on heavily on Vegas numbers, probably on salary as well. And so what I think you see naturally is that you’re going to get a lot of clustering, especially when pricing’s off and everyone is pointing towards Vegas, you can get a lot of clustering around those players and just by having a model that is independent of that, you’re going to get some uniqueness just from that alone. Is that fair to say?

Wil:
Yeah, absolutely. Yeah, the only time that I pull in Vegas is just for calculating ownership, so it doesn’t impact any of the projections or anything outside of the ownership pool of lineups that I look at. I do think, yeah, I can’t think of another model that is using something that’s not just directly pulling in Vegas for those win probabilities.

Andy:
One question we got in Slack that’s more direct where it’s saying you can have, I know you have a more thorough process for how you’re building lineups and all that, but if you were to give rules of thumbs, do you think that leaving salary on the table makes sense for CS:GO? Is that something you’d recommend doing? If so, how would you think about specifically what to leave?

Wil:
Yeah, so the specific number I think would be hard to come to, but pricing is just bad for CS:GO. I don’t think DraftKings has a lot of reliable data, especially on slates like today’s slate and yesterday slate where it’s really small teams, say it’s a tier two Russian tournament. And so, you know, DraftKings just doesn’t really know how good these players are, so they’re all like 6,500 to 8,000 salary. You can pretty much if you know, which players are good, so we have years of data on these players. We know what their role is, and we can also regress them towards the stats of their traditional role. There’s lots of different data that I don’t think DraftKings is using to make their salaries, and so I don’t think that you need a high salary floor at all. I mean, I’ll run mine ups like at 38,000 salary, because it’s almost not even just to be unique. It’s just because sometimes that’s a good lineup at DraftKings’ price hit.

Max:
Just to clarify, 38,000 would be leaving 7,000 on the table or leave 2,000 on the table?

Wil:
12,000.

Max:
12,000, okay. You feel like you can make optimal lineups that leave 10,000 on the table.

Wil:
Sometimes, yeah.

Max:
Okay.

Wil:
I think there’s, obviously, if they’re pricing the top player on a dog really well, like I think tomorrow there’s a plus 120 dog and their best player is at 10K or something like that. Obviously, that’s going to raise your salary cap quite a bit. But there’s definitely slates where they’ll just completely fluff pricing and you can leave a lot on the table.

Andy:
That’s reflected in our default sliders too or default settings in that case or minimum salary here, if you’re just going in to run something on defaults, it’s down at 20K for TK. It leaves you a ton of room to build some lower salary build.

Wil:
Yeah, and I think you won’t naturally get too many that low because they just won’t project very well. But it’s definitely not something where I think you really need to impose a salary floor of any kind.

Andy:
Right, it’s not going to be like a basketball where the market’s pretty efficient, pricing is usually solid, and unless there’s some big injury, you should be wary of leaving a ton of money on the table on a bigger slate. For this, it’s that they frankly don’t put a huge amount of effort into their pricing and even if they do just the way that if they were to, they’re just not, unless they hired, they poached you from us, I guess, on the one way to get better pricing. But it’s just not something that, unless they have someone who all they’re doing is Esports pricing, I don’t see how they can get that good at it. And so it’s just not their core competency, so you can almost throw it out the window. [crosstalk 00:22:14]

Wil:
Yeah, and I think they’re also obviously very reliant on the Vegas odds for it. Vegas itself is just super inefficient for Esports. I think that’s something that you’ve touched on in the past. There’s low limits. There’s not a lot of interests, so the lines are just set and move around in that area without, like they’re definitely not, I think, reaching the true mean.

Andy:
One other point I just want to touch on in this, is that in these contests, how are you thinking about how many entries to put in? Do you just go ham and max everything out?

Wil:
Yeah.

Andy:
Do you think that’s a viable strategy? How would you recommend it, is that something you’d recommend to everyone if the bankroll supports it and they think they’re plus EV to put in 3% of the entries, whatever DraftKings allows, or do you think there’s merit to taking a more conservative approach there?

Wil:
Yeah, so for now, I see a lot of EV in Counter-Strike. I think they’re really soft contests, I think, especially with my model, so I max it every day and I’ve done so for the past couple of months. I think that building your process, you shouldn’t do that. I wouldn’t just say if somebody hasn’t played Counter-Strike before, but has the bankroll for it to just come in, max the entries and go hit the build button.

Max:
Don’t do that, please.

Wil:
Yeah, don’t do that. I think it would be good to throw in 10 or 20 entries and play around with the stack rolls, play around with what the model looks at versus Vegas, and that sort of stuff to figure out what their process would be for it. But I definitely think that it’s really soft and there is a lot of room to go hit it. Because I mean, in the 180 today, I think there’s two lineups that are duped to combine like 120 times or something like that. They’re not even that good of lineups. I think they’re just some projection systems optimal, so there’s plenty of EV to go around, in my opinion.

Andy:
That makes sense. For listeners as well, Wil is playing the bigger contests with, I think, it’s like $10 buying is simply the biggest prize [inaudible 00:24:23] contest. They’ll usually have $150, two entry max contests, but they also have $1 contest. They have the quarter jukebox, which is 25 cent and dime time, which is a dime. These are contests that are going to be much softer and are going to be really good ones to get your feet wet, so I would recommend doing that before moving up the stakes, even if your bankroll supports it.

Andy:
That’s something that’s really important as you’re experimenting with other sports is that you want to just make sure that you really have a grasp on the concepts before going all in to supplying your standard bankroll management, because bankroll management doesn’t matter if you’re a losing player. You’re going to go broke, and so you want to just make sure that you’re able to get in a good enough sample to be confident in it before moving up. I guess this might be my last question, but if anyone else has anything, jump in, but Wil, what has, not the CS:GO variance within the game, but as a player, what does variance look like? I know that you are at the top of that, and so your swings are going to be different than the average player. What has swings look like when you’re maxing out all these contests?

Wil:
Yeah, so, I mean, I had, I think, Andy, you remember my run in March. It was like I banked, I think eight times in the two and a half weeks or something like that.

Andy:
Yeah, it was [crosstalk 00:25:51].

Wil:
Yeah, that’s not a normal [crosstalk 00:25:53]. Then I think I banked last week, but before that I had a two or three week run where I was maxing every day. I was getting third or fourth or something like that. That definitely brought my bankroll down a little bit, because there is that variance of it. There are still quite a few swings in there, but I think overall it’s obviously been really kind to me.

Andy:
Yeah, and one last question. I’ll just comment on this one. Ben asked on the subject of Vegas odds for Esports in general, how do you think the Vegas lines are? And so one important thing to point out is that for the major sports, DraftKings and in the US-based books have really gone all in on props. And so they are setting props on pretty much every player in the big three, so baseball, basketball, football. Because of that and because of getting a lot of action on that, they’re getting constant feedback on how these individual players will perform, and they’re investing in getting accurate lines up there and that makes their pricing for DFS sharper. Then the market helps get their props sharp as well.

Andy:
For Esports, maybe some of the European books might have some props, but you see a lot of weird props of like map one kills and just things that don’t relate to individual players, but you don’t really have props on individual players, and so that helps the salary be softer, because there’s not a point of reference there. But in general, I’ve done a good amount of CS:GO betting based on Wil’s model. The limits are usually 300 bucks maybe. Sometimes you can get a thousand down at a single time. If I say something’s that plus 150 and you got $300 on it for the MoneyLine, it will very likely go to like, plus 141 or something like that. It’s going to move pretty quick on not that much action. That’s makes it not that sharp of a line, but also means you can’t profit a ton from it. That being said, where I think you can get edge that’s meaningful is all the different bets that they have out there. The main ones are the Moneyline who’s going to win this game. What is actual term for the overall set of three maps? Is it a match?

Wil:
Yeah, I always get it confused. I think there’s, because they also call each game … There’s game, match, set and series, or that might be tennis too. It’s way too confusing.

Andy:
All right, so the Moneyline is who is going to win this set of games. Then you have a spread, which is usually 0.5, so it means, does this team win at least one bad play? Or is it just 1.5? Right, so did they win at least one map? Then you have, so are you saying, are they going to get swept or not or is the other team going to sweep them? And you also have the total, which is over 2.5, and like, what is the chance that this goes to all three games.

Andy:
This is something where the lines aren’t sharp to begin with and they don’t move in tandem with one another. I think you can get a lot of great betting opportunities where just some of the bets just haven’t moved in a while, even if the MoneyLine come down 40 cents, the other ones haven’t moved at all. Because I think casual betters are going to bet the Moneyline. A lot of these other numbers just get stale. This is long-wined answer to say there is edge there for sure. It’s hard to make a ton from it, but if you have access to a good number of books, you can do well with it. League of Legends, I am assuming just by judging from the limits that I’ve seen, it’s pretty similar where, yeah, it just can’t be that sharp. Yeah, that’s how I view it. Max, have you done any League of Legends betting?

Max:
Yeah, I’ve done some betting. I think the limits are probably a little bigger than CS:GO. Sometimes if I do see line movements, I will use that to adjust my, when I’m making lineups, so I think they are they’re valuable information, but yeah, it’s not like these NFL games, you’d bet $100 thousand on them, so they’re huge markets. It’s completely different.

Andy:
Were there any other questions we wanted to cover on CS:GO? I think this was a great crash course in it, but Jordan, was there anything that you saw?

Jordan:
Well, I had one more for myself that was just a clarifying question. When you’re looking at the app on CS:GO, you’ll see the odds for each team. Are we pulling those from Vegas or are those your calculated, I think you said true skill odds, Wil?

Wil:
Those are pulled in from Vegas right now. I know we’re planning on putting up the betting page with the actual outputs for it. But right now that’s just what Vegas has the game at. I think it’ll also be pretty clear from the projections who’s favorite and who’s not, just because there is such a stark contrast in winning versus losing the game that the projections usually just reflect the win probability of the team.

Andy:
Perfect, and on some of these too, I mean, you’ll see some numbers that, like a team will be massively favorite compared to some of these other sports. How do you handle games like that, broadly speaking? I’m looking at today’s and there’s a minus -6638 favorite.

Wil:
I think they’re pulling in the live odds, I think that’s …

Andy:
Okay.

Wil:
Yeah.

Andy:
Gotcha.

Wil:
I think that they’re definitely, especially on some of these smaller slates, they’ve run the past couple of days, you’ll see like a minus 2000 favorite. There’s not really too much you can do about that. The best players from that minus 2000 favorite, they’re going to win and they’re going to score really well, so that’s chalk that’s not really fadable unless like you get the world’s craziest upset.

Andy:
Well, and sometimes also they just don’t play.

Wil:
Yeah, yeah. There is also withdrawal risk that happens sometimes. I know this morning, the Young Ninjas team had two players that just sat down. They brought in one of their old players that they just transferred away and their 37 year old coach to sit in and play for them. There’s definitely, that happens on the smaller tournaments. I think that’s also why the prize pools are smaller for them. Tomorrow is the big Intel Extreme Masters Tournament. They brought back the 5K to first. There’s much lower risk of cancellation or anything like that.

Andy:
Yeah, and that is an important thing for people to keep in mind in all of Esports is pay attention to the league it’s for. The tournament series, I guess, is more what it is in CS:GO. Some of the random ones just get these are third tier teams. There’s very little data on it. A lot, the prizes in DraftKings are bigger than they’re playing for in the actual event. It’s not that there’s match [inaudible 00:33:25] or anything. It’s just, I think there’s not as much motivation. They’ll just want to try new things, swap people out. Those ones you just want to be cautious of, but if you focus on those bigger contests that’s where I think things will be most stable. Just always keep that in mind.

Andy:
Yeah, I think from here, we’ll jump over to League of Legends, but Wil’s going to stick around, so if anyone does have CS:GO questions, feel free to dump them into the chat. I guess, just quickly, someone in the Slack asked, do you use the true skill rating for predicting team strength? It’s basically like ELO, if you’re familiar with chess. It’s a rating system. I think it’s from Microsoft. They developed some of it. That’s basically what it does, is it’s to predict a team strength, right?

Wil:
Yeah, so it’s very similar to ELO except with ELO., they have a standard variance. It’s basically every game is assumed to have the same variance. A true skill model has adapted variance, so it’s based on recent form, time with the team, et cetera. In my opinion, it’s just a little bit more comprehensive. I think it better accounts for the variance that’s in Counter-Strike.

Andy:
All right, and yeah, people want to sweat along with Wil for any of his inevitable banks. Yeah, you’re reformed racer on DraftKings. Yeah, League of Legends, kicking it over to Max. We’ve done some videos on this before. I recommend that people check those out. We’ve had a lot of success over the year as we’ve been building out the model there. Max has led all of that and learned a lot along the way. And so, Max, I guess, just starting from the beginning, can you give just, League of Legends is different than CS:GO. I mean, they’re both computer games. I guess that’s the similarity. They both have a lot of correlation, and so you’re going to get stacks, but League of Legends has positions. They do have a draft with the champions, but it’s not really modifiable. I guess, just in contrast to what we’ve talked about with CS:GO, how would you even think about League of Legends for someone who’s not familiar with it?

Wil:
Yeah, so I would say CS:GO and League of Legends have actually a lot of similarities. You can think about them in similar ways. I think, one thing I’d emphasize is that, and as we said before, that the line of builder that we’re using for these are not, we’re not building it based on simulation data. So having your risk sticks and having strategies around getting players into particular spots in your lineup can be really important. With League of Legends, it’s like a captain style showdown-like thing, except it’s usually a best of three or sometimes the best of one.

Wil:
You’re going to focus on a couple of things. One is getting that captain spot with players who actually have the ability to have really high upside games. A lot of times I’m going to filter out, actually pick my pool of players for the captain spot. Specifically, you would be best served taking out a position like support or team from your captain pool. Those players have just such a low probability of being the top score. It’s usually a really good move to just take them out.

Wil:
Then there’s certain strategies with what positions you want to keep, right? Do you want to, some people only use ADC and mid and more recently jungle. Some people will use top, which these are all positions in League of Legends for the captain spot. It’s sort of like center point guard, shooting guard, small forward. Not a flawless analogy, but there’s positions. Different positions have different roles. That means they’re going to have different upside. ADC is going to be a position that’s going to get a lot of kills, and so that’s going to be a popular captain [inaudible 00:37:23]. But basically, focusing on getting that captain pick right is going to be really important.

Wil:
Then focusing on getting stacks, as Wil recommend in CS:GO. In League of Legends, you can do a four stack. I highly recommend you force a four stack. I would say, if you’re new to League of Legends, even forcing it, doing something where you’re forcing a full four, three stack is probably going to be pretty valuable for you until you got the hang of it, and even when you get the hang of it, some people just only do that. I think it’s a good strategy. Correlations are really high. You’re not going to see the correlations between players can be 0.7, something like that. This is higher than any other sport. If your team wins, you’re probably going to do well, especially as a [inaudible 00:38:11], so that’s something that’s really important as well.

Andy:
In … Yeah, go for it, Jordan.

Jordan:
Well, I was just going to say, so Wil had mentioned on the CS:GO side that the game is more variant than people think, and that ownership condenses and there’s some opportunities to find leverage against the field, fading some of the chalk. On the League of Legends side, do you think that the field does a good job of assessing strong plays and ownership condenses around the best plays similar to something like basketball? Or is this more of a sport like baseball where you can find a lot of leverage against the field fading, some of the highest owned teams in place?

Max:
Yeah, I think League of Legends is the perfect sport for someone who likes to fade, because I have found, doing my own analysis, that teams that are underdogs actually have a little more upside when they win than teams that are favorites. And so when you’re fading that favorite team, sometimes that team is going to, the team that you’re like underdog that you’re playing, is actually going to, when they actually, when I have higher upside than that team, that might be the big favorite. I mean, this also depends on how aggressive the teams are. Some teams are very aggressive. They will risk dying or getting a kill a lot faster, try to engage the other team a lot. Some teams are very conservative.

Max:
That is a lot of upside, but there also is a dynamic where those underdog teams do actually have a little more upside. And so it is a great spot where you’re looking for that diamond in the rough team that isn’t such an underdog that they’re never going to win, but is underdog enough where they’re going to be overlooked and you can use one of their best players in the captain and get a 2%, 3% on captain. That can be a slight winning play.

Andy:
Why is it that you think that the underdogs have higher upside potential?

Max:
The short answer is, I don’t know. My guess is a lot of the League of Legends scoring happens as the game goes on later in the game, because what ends up happening is early games, teams and players stay in their lanes. There’s three lanes on the map. Then as the game goes on, the team’s group up. A lot of the scoring and League of Legends is based on getting the assist, so basically if you damage a player that gets killed, then you get points. And so if the teams are grouped up, you’re going to score a lot more points, because all the team members are going to be getting those assists since they’re together. And so I think that has to do with why. Usually, if you’re watching, what you’re rooting for is for the game to go on as long as possible for your team to be winning, because they’re going to get a lot more extra kills and assists because of that.

Andy:
Okay, and yeah, Alex in the YouTube chat had a question about the draft phase. I want to just talk a bit more about that. Can you, I guess, actually, can you just talk a bit more about that? How does the draft actually work and just what opportunities are there around that, if any?

Max:
The draft is similar to what Wil was talking about with CS:GO with map picking is there’s different characters that players can play. Those characters have different abilities and are better for different stats, and so picking a character can be very, very important. I, sadly, do not have a model that’s as sophisticated as Wil where I’m simulating what champions they’re going to pick. I try to project the type of …

Andy:
I don’t know if you could.

Max:
You probably could actually.

Andy:
You think so?

Max:
That’s something that I’d be looking to do in the future, but for the time being, I just basically assume what kind of champions or not assume, I look at the data, what type of champions players like to play and use that and put that into my model. I think people who are really into League of Legends try to predict champion pick, because it actually is very important. If there’s certain characters that, for example, don’t get these extra points that have to do with killing minions, which I’m not going to get into, but there’s different characters that have a widely different array of skills and aggressiveness. Usually, if you’re watching, there are certain characters where if a player picks, you are going to be very unhappy or very happy. It is very important. I think it’s pretty sophisticated and tough to predict. You don’t need to predict it to win at League of Legends. I don’t do it, but it’s an edge you can have if you get really into it.

Andy:
Okay, and on that note, how good do you think the teams actually are at the draft? Is that something where they have their own version of Moneyball and they’ve really invested in figuring out the optimal draft? Or does it seem like they’re just not doing a great job there? This is just a random question I had, but I was curious.

Max:
I don’t have an opinion on that. I assume they’re pretty smart about it, but I don’t think it would affect your DFS play at all to think about the strategy, unless you’re just really into it and you want to focus on that. Go at it. But I don’t have a strong opinion.

Andy:
Interesting. Yeah, Matt, as a good point as well. He said the issue with predicting champions is that the meta changes so often it completely shifts which champions are viable. Can you talk a little bit about what the meta is and how that impacts things?

Max:
Sure, so essentially League of Legends, will update the characteristics of some champions or players that you’d pick and they’ll add new ones. Because of that, the dynamics, if they change it enough, the dynamics of the entire game can change of which player can be more aggressive and which one’s more conservative. It used to be, a couple of years ago, that ADC and mid-laners were the most high upside positions. They were going to be the players that are going to get the most kills. Those are going to be the players that you exclusively will use in your captain. Last year, jungle became something where a lot of, because of the changes of the skills of players, jungle became a viable position that actually have a lot of upside. You could get a lot of kills. This year, I’ve heard from one of our users, that top has become a more viable position. I haven’t looked into it that much, but it’s going to be reflected in the projections just simply because it’s based on historical data, so if some player or positions is getting more kills, that’s going to be updated in the model.

Max:
From season to season, that can change pretty drastically. It’s something you want to look out for, I think. I usually, like last year, I probably wouldn’t play any top position players in my captain. This year, that might be different. It just depends on how the game is changing.

Andy:
Okay, Jordan, do you want to …

Max:
One thing that you can … Yeah, sorry. I was just going to add one thing you can feel pretty confident about is ADC and mid are always going to be pretty viable captains.

Andy:
Okay, cool. Yeah, Jordan, do you mind jumping in. I’m lagging a bit, and so I just want to make sure we can keep things going.

Jordan:
Yeah, no problem. There is a question from Ben here in Slack that I think is interesting. He said with sports being highly correlated, with fantasy points being highly correlated to chance to win and the projections reflecting that, do you find yourself forcing certain stack combinations or adding a lot of additional noise to the projections to force some of those higher variants underdogs into your lineups? Or can you talk a little bit more about just actually what your lineup building process on SaberSIM looks like?

Max:
Yeah, so I’m just going to assume we’re talking about a four or five games slate, because I think with two games slates, it’s a little different. We can talk about that if you want. But essentially I think there’s a few things. One is I will manipulate projections in order to get more exposures to teams I want. I also might do that in the post build process. Sometimes I might straight up exclude a team, because I think they’re going to be too heavily owned. I’m just going to take a really risky position and just take them out of my pool.

Max:
Then also, I mean, we have a lot of tools in the post build process that’s going to allow you to balance your stacks, getting particular stacks. I definitely using stacking rules. I’m going to always force a fourth stack and sometimes I’ll do something depending on how much time I have is really taking advantage or filtering, right, is if you have the time, it doesn’t take that long to filter out some lineups that might just be, you’re like, “Okay, I have the favorite too much, I just don’t want that much exposure to them.” Then filter in lineups that have more underdogs. I think in general, yeah, I think I’m trying to manipulate the builder a lot.

Jordan:
Gotcha, and for looking at some of these underdog teams, maybe say that you’re building league lineups for the first time and you want to find some of those higher upside underdog teams, is there a particular place that you use to start some of that research and finding some of those opportunities or maybe some advanced statistics? We talked about a minute’s upside in basketball the other week. I know, I mean, we look at targets for wide receivers in football, things that maybe provide an opportunity to tell a story that doesn’t necessarily come out in the projections about upside.

Max:
Yes, so I have, I mean SaberSIM provides me with all the data I need, so I can look at it myself. There’s a popular website called Oracle’s Elixir. I think one of the most important stacks you can look at is something that’s called kill deaths per minute. It’s basically an aggressiveness stat, right? It’s, for a team, how many kills or deaths or combined are they getting a minute? That’s going to show you how often they’re engaging with the other team, because if they’re a bad team, they might die a lot, but that’s just as valuable to us. Or if they’re a very good team, there might get kills a lot. Both of those are very valuable, because it just shows how much they’re engaging with the other team. And so a lot of these teams, they’re going to be a very good team that can be very aggressive and gets a lot of kills, but when they lose, it means they die it a lot.

Max:
I’m usually looking at, I mean, I think the key, in a very simplistic sense, is you’re looking for a team that’s in a really aggressive game that might be a little bit of an underdog or might fly under the radar because those aggressive games, when those underdogs win, they can have a really, really high upside. I would say Oracle’s Elixir or there’s other sites, I would just look for League of Legends team stats like that. That can be very valuable information. That being said, do not go overboard looking at these stats, especially in a small sample because they can be noisy. So yeah. Well, that’s all I’ll say.

Jordan:
Gotcha, and similar question we asked to Wil before. In terms of max entering these kinds of contests, do you find on a typical, maybe four or five games slate that you can build 150 profitable lineups? Do you find yourself maxing contests out or more taking a few shots on just a handful of lineups? Or what is your contest selection mix look like for these slates?

Max:
I usually don’t build more than 20 lineups. I think if the tournament’s big enough, and you can get them were it’s a 5,000 entry tournament, I might do something like 50. But I usually like to take the approach that I want to make sure, especially since again, our builder is not, unless I’m just like forcing three stacks and again, just ensure, okay, all of these lineups are going to be four three stacks. I’m comfortable with that, whatever. It’s fine. I just want to make sure that the lineups are all just go through that quality control process for me. If there’s 20, I can manage that really well. If there’s more, it’s a lot harder.

Jordan:
For you, it’s less of a concern around the theoretical edge and more just because of the amount of manual oversight you need to have in the lineup, you want to pick a number that’s manageable for you to review in depth, right?

Max:
Yeah, absolutely. I mean, I’m a more risk averse player, so I don’t like doing something where I’m risking putting in an unprofitable lineup. And so I know if I do 20, I’m not going to be risking that at all, because I can go over all those lineups. Obviously, we have tools that are going to allow you to have good oversight of what your lineups are, but I just try, I think 20 is a good number to manage.

Andy:
Nice, and looking over at the questions, has anything come in on YouTube or Slack that we’ve missed, Jordan?

Jordan:
No, nothing that we missed. There’s been some really strong conversation going on here alongside our conversation. Ben and Maverick, both mentioned too, that in talking about why there is that extra upside on underdogs, that it seems like that’s when the underdog wins, the win condition is a longer drawn out game where there’s a lot of team fighting, the teams are scrapping and racking up those kills. Whereas, the favorite maybe has a win condition that’s more of a steam roll type win where there’s not really as much of an opportunity for upside. I thought that was kind of a strong point made there. But I think we’re pretty caught up with questions here. I know Maverick had mentioned some questions earlier in the day here that were some more complicated meta type questions. I think maybe if there’s nothing else coming in, that’s a good time to talk about maybe some of the more advanced concepts about how you handle the meta shifts and things like that.

Andy:
Before we get to that, there was just one basic question I forget is for these sports, there’s no late swap for either for CS:GO or for League of Legends. It’s like once the site starts, that’s it. Okay.

Max:
Yeah, absolutely. And again, I mean …

Andy:
Yes.

Max:
Yeah, sorry.

Andy:
No, I think it was just in some ways it’s actually more beneficial for the average user, because the only, you would, it’s not like MBA where everyone essentially has access to this information on who’s getting scratched. You might not have much time, but you don’t need to be super deep into the Twitterverse to know what’s going on. Whereas with these ones, you’ve got to really know what’s going on. I think that helps make it more approachable for more casual players, which is ultimately what you need for these niche sports. But yeah, what were your thoughts on it, Max?

Max:
Well, I was just going to say, I feel like we should mention this as well is with a lot of the League of Legends leagues, you can find who the starters are on Twitter. I don’t know, Andy, you might know more sources as well. Usually, you can just look at the box scores from the game before and unless they have announced something differently, you can be pretty confident that the starters from the game before, especially if it’s not the first game of the season or something, are going to be the starters. That’s something to keep in mind is you really want to be aware of if there’s a change in starters, if there’s a player who sometimes gets subbed out on these best of three things. The Korean League specifically is really notorious for it. They don’t really announce starters, so sometimes you have to be a little more careful with those players. Sometimes you can play them, and then at 3AM, it turns out one of those players is not playing. That’s just something to keep an eye on. There’s a good information out there. You just have to look for it.

Andy:
Yeah, before we get to Max’ meta discussion, that’s another good point where the start time of these slates, CS:GO, what is the primary market for CS:GO?

Wil:
Europe, mostly. Primarily slates will start, like the minor tier two, like the tournaments that are on this week were pretty much just small Russian tournaments, so they started at 3AM or 8AM. But big things like the Intel Extreme Masters are focused in Europe, but they’re aware of their New York and US customer base, so they’re like 10:30, 11, 11AM Eastern [inaudible 00:55:30] that go until 6PM.

Andy:
Mm-hmm (affirmative), and then what is, it looks a bit different for League of Legends, right Max? I think there’s a strong presence to be mentioned [crosstalk 00:55:40].

Max:
It depends on the league. Right, so usually a League of Legends every night there’s either the Chinese league, which is LPL or combined Chinese and Korean League slate, which is going to be four games. That’s going to be late at night. LEC, The European League is always in the morning. LCS is always in the afternoon. They’re only on weekends, on more three or four day weekend. I would say actually probably my favorite slates to play are LCS or LEC, because they’re only best of one, so there’s a lot more variants and they’re during the day. You don’t have to worry about that stuff. That’s funner, easier one to get into, because you can really take advantage of that variance from that only one game series that all these games are playing.

Andy:
When are you typically building lineups?

Max:
Typically building lineups just a couple hours before or maybe an hour before. You can do it pretty far in advance if you want. We project these teams pretty, pretty far out. I do it seven days in advance and update it every day. That’s something where you can build lineups early, especially for LEC or LCS, and you don’t have to worry about something changing. Usually, it’s going to be the same players.

Andy:
Yeah, and one thing to point out is that RotoWire recently started posting, not that recently, but it wasn’t up when we first started getting into Esports, but they now post some confirmed lineups for League of Legends. They don’t, it’s not perfect. They’re not updating all the time, but when they can get it, they will update it. That’s a good free resource for people to check out. But yeah, what does that look like?

Max:
Yeah, great resource.

Andy:
Yeah, and then I’m trying to think, so, yeah. Jordan, do you want to try to present the question for, Max?

Jordan:
Yeah, I think there’s some very specific questions about the current meta here that we could dive into. But I think one thing that would be most interesting just in terms of thinking about longer-term League of Legends strategy is just how can we adjust and take advantage of playing DFS in a game where the game itself is potentially changing every two weeks, right? That’s what we’re really talking about here with meta shifts is a patch will come out and it’ll actually shift some of the core items or champions or otherwise things that impact the way the game plays. I would just be interested, I know this is a really general question, but how that impacts your process. I mean, Do you look at patch notes? I mean, what comes into your process.

Andy:
Because there’s nothing really like traditional sports. It’s like, we can talk about juicing the ball or like making some things that have an impact and maybe teams start doing more of a shift, but it’s like baseball’s the closest and it’s still not nearly the impact that it seems like some of these changes make in League. Is that fair to say?

Max:
Yeah, so I’ll say this. The first thing I’ll say is I think that this type of nuance is something that you don’t totally have to worry about if you’re a new player, right? The top position could become a viable captain this season, but if you don’t use the top position to draft them and you’re building strong lineups and focusing on getting the right stacks and finding those under the radar players, you’re going to do well, right? We’re talking about how can you take advantage of these changes in the game? That’s going to be something that if you’re a more advanced player can be fun to look at, right?

Max:
I personally, actually have the opposite view of this is where when there’s something that could be viable, unless you’re really early on it, I think that it can make it. The bold staples are actually better plays, right? When suddenly top, jungle, mid and ADC are all viable captains, suddenly you’re going to see, okay, this really good mid is getting 2%, 3% ownership in the captain or this good ADC is the same thing. It’s going to depress the ownership of some of these guys are just safely going to be the highest upside plays. And so I think there’s sometimes where there’ll be something like this happens, and if you’re late to the party, you can actually just get an advantage by just not over adjusting to it and doing everything that you want.

Max:
I think one thing that’s something that’s a dynamic with League of Legends is the positions have varying salaries. ADCs usually have the highest salaries. Top usually has some of the lower salaries. This can be an advantage early when people are looking for any excuse to put a lower salary play into captain, because if you can do that, it means you can make two stacks of two of the biggest favorites and your lineup looks really good and feel fairly good, right?

Max:
But what we want to do is we want to use those highest upside players in the captain, like the best ADC, the best mid, the best jungle on a particular team. The try to get add value by just playing underdogs. Because when you do that, then the ADC is going to [inaudible 01:01:17] cost last, and you’re going to have a lineup where you get the highest upside play in the captain and when their team blends, you’re going to do really well. I think trying to do something where suddenly top becomes a little more viable and has a little more outside, I think can sometimes backfire as the other players adjust to it, because then that becomes a really awesome play. It becomes a really popular thing to do is use top or jungle because you’re going to save money. Then that ends up just not actually being a good play in theory, when the ownerships [inaudible 01:01:50] too much.

Andy:
Ultimately to me, for these more niche DFS sports, it seems, I think you can go super deep on them and a lot of players do, but you don’t need to over-complicate it. Because there’s frankly, not a lot of great information out there on how to play these. There’s not as much good data out there. Vegas isn’t as sharp. The average person coming in just frankly, is going to be bad. With that in mind, these really are sports where rules of thumb can get you pretty far and you don’t have to complicate it. I think as you start moving up in stakes, as you start putting more entries in, there is more value on staying on top of the nuance and just really digging in deep to these. But I don’t want listeners to think that you have to stay on top of absolutely everything. I think there’s a ton of edge in these sports still. Play within your bankroll. Follow these rules of thumbs. Then as you start getting some traction, start working on developing your processes and seeing how you can differentiate, where you can find spots to add some more value to the process beyond the rules of thumb. But I think you can get started with these without making it too complicated. Is that reasonable?

Max:
Yeah. Oh, absolutely. I think if there was three things that I’d love for someone who’s new to League of Legends to take away from this video is focus on just these things. One is getting your captain player full, correct, eliminating the players that don’t have high upside. You can get rid of all the supports, get rid of all the teams. You can get rid of tops in some cases. Get your captain pool where you are forcing the highest upside players into your captain spot. I’d say the second thing is make sure to take advantage of our stacking rules, right? Force a four stack. That’s a very, very important, at least force a three stack. If you’re really new, force qa four stack with a three stack or whatever, just make sure that your lineups are stacking, make sure that you don’t make any lineup combination mistakes in that way.

Max:
Then focus on the post build process, right? Get it is use our tools that we have in part three of that build process where you’re looking at your lineups, use filtering, use min and max exposures, make sure that you’re getting a nice balance, try, maybe try to raise exposure of a team that you really like that is maybe that underdog that’s in a very aggressive game. Use all those things and you’re going to do pretty well. I think you can make, even if you don’t understand League of Legends very well, you can get a lot of edge in just constructing solid lineups. That’s something that rings true for any form of daily fantasy.

Andy:
Awesome. Well, this has been a lot of fun. I hope people got a ton out of this. We are going to be back next Thursday with another strategy session. If anyone has any suggestions for topics you want us to cover, just let us know. We’ve been really enjoying these, enjoying the interaction and just really having a platform to go deep on some of the subjects that I think aren’t talked about enough. Thanks for tuning in. If you haven’t tried SaberSIM already, you can go to sabersim.com and sign up for a free one week trial. As always, if you have any questions at all, you can always reach out to us at [email protected] or shooting us a message in the Slack group we’ve got going. Yeah, Jordan, Max, and Wil, wherever you went, thanks. Thanks for helping out guys.

Max:
Thanks guys.

Jordan:
Yeah.

Andy:
Yeah.