Transcript

Andy Baldacci:
All right. Hey, guys. What’s up? Thank you so much for joining us as we walk through everything that you need to know to beat daily fantasy League of Legends. My name is Andy Baldacci, I’m the CEO of SaberSim. And I’m joined by DFS pros, SaberSim partners, and twin brothers, Max and Danny Steinberg. How’s it going guys?

Max Steinberg:
Hey.

Danny Steinberg:
[crosstalk 00:00:26].

Max Steinberg:
So, this is what we’re going to cover today and we’re going to first go through just how League of Legends works at a high level, then break down the secrets of building winning lineups, walk through an actual build so you can see how we put this all into practice. And then wrap it all up and tell you what do you have to do next to get this going.

Max Steinberg:
So, jumping into how League of Legends works. This isn’t going to be a super deep-dive into everything there is to know about League of Legends, but I’m going to just try to give you a starting up understanding so you can confidently play the contest and actually know what’s going on. If you want something more in-depth, though, we actually did a pro Q&A that you can find on our YouTube channel where we broke down all the ins and outs, tons of intricacies about it, but for now let’s just cover the basics.

Danny Steinberg:
Yeah. Well I just want to jump in too and say, it’s just like any other daily fantasy sport. It’s the similar strategy so you don’t need that deep of understanding of the game to actually beat it, you can beat it without having a deep understanding of the game of League of Legends but having a good understanding of daily fantasy.

Andy Baldacci:
Yeah. And this is something we’re going to get into in a bit, but just by the nature of the game and how it works, having strong understanding of lineup construction and what goes into strong lineups will make you a winning player in the long-run here even if you’re not that familiar with the game, it’s understanding kind of the underlying fundamentals of DFS and how they apply here. But. Yeah. League of Legends is what’s called a MOBA which is a multiplayer online battle arena, and Max actually gave me this analogy that it’s similar to NBA where every game is made up of two teams, in this case it’s red and blue, and they each have five players, and each of those players has a different role.

Andy Baldacci:
Some of the roles have higher upside and projections in general, while others might be on the lower side. So, unlike the NBA, while both League of Legends and basketball have assists, League of Legends, the scoring primarily comes from kills. Each team has a base they need to guard while simultaneously attacking their opponent’s base, and out of each of the bases there’s what’s called three lanes, you have the Top lane, the Middle name or Mid, and the Bottom lane or Bot, and the space in between all those lanes is called the Jungle. The positions in the game are based on where in the map each player is focusing, so they’re pretty straightforward, Top is the Top lane, Mid is in the Middle lane, Jungle is in the Jungle.

Andy Baldacci:
In the Bottom lane, though, you have ADC, or Attack Damage Carry, and Support. The ADC is one of the primary damage dealers, as the name implies, and Support’s job is to support them. At the back of each of the bases, there is a building called a Nexus, and you win the game by destroying the enemy team’s Nexus. And again, if you want to cover more of the details to get a better understanding of what the actual gameplay looks like, you can check out that in-depth breakdown of our League of Legends pro Q&A on our YouTube channel, but this should be more than enough to get you started, at least just familiar with how things work.

Andy Baldacci:
The next thing to think about is just how the actual scoring works and fortunately for us DraftKings and FanDuel both have the same scoring, and it really comes down to, as I mentioned before, kills and assists. The only differences between the two sites are rules for how many players you can play from the same team and we’re going to get into that and some of the other variables to pay attention to next. And from here, Danny, why don’t you take over and just jump in and break down those keys to winning lineups and just what that looks like for League of Legends.

Danny Steinberg:
Okay. Cool. So here the secrets of winning lineups. So there’s really three parts to this, correlation, ownership, and variance. So let’s just define upside first. First, building high upside lineups is the key to winning in daily fantasy GVPs, and that just means building lineups that consistently perform better than the projections would lead you to believe they do, which means you are able to separate yourself from the rest of the field and able to win a huge contest where first is a gigantic amount of money. So the three elements that are contained in upside is correlation, ownership, and variance.

Danny Steinberg:
Correlation is just a measure of how players on the same team score together, so players with high positive correlations, when one has a good game the rest of them are going to have a good game probably and players with negative correlations if one person has a good game the other players likely have a bad game. And this is where the value of staking comes in. So with LoL or League of Legends, there’s just gigantic correlations, maybe the highest correlations of any sport I think I’ve seen, at least. So when you stack all your lineups or you use a lot of stacking your lineups are going to be more boom and bust, but it’s a trade-off worth making because of the structure of the tournaments we’re recommending you playing. With large GPPs is you want boom or bust, so you can win first or not win at all. Don’t really care to get in the top-

Andy Baldacci:
You’re not just trying to sneak into the money.

Danny Steinberg:
Right. Exactly. You want to get that boom or bust. Ownership refers to how frequently a specific player is rostered in a contest. So, if one of the players you play has a really good game that’s obviously great, but if they’re super highly owned, that’s not going to be as great as if you played someone who was very low-owned.

Max Steinberg:
Yeah. And especially in League of Legends, there are going to be really high-owned players so those can be really important. Some players are going to get 50% ownership and some players are going to get 2% ownership, so the spread of ownerships can be really, really high.

Danny Steinberg:
The other element is variance, it just refers to how a player’s performance game to game varies. Two players with the same average projection may have way different floors and ceilings and a lot of that can depend on the role. Like often an ADC or a Mid can have a very high ceiling, but maybe a Top doesn’t necessarily have high ceiling all things considered, all their projections are the same. So there’s differences between the variance and different players’ performance and it’s important to be able to identify that. So here’s how you find upside in League of Legends. League of Legends is like baseball, except it’s on steroids in the sense that these players are just insanely correlated to each other. All-

Max Steinberg:
So, like 1990s baseball.

Danny Steinberg:
Right. Yeah. Because of all the home runs, you’re saying? Yeah. So all the players on the same team have gigantic positive correlations and all the players on opposing teams have gigantic negative correlations. So basically, always stack, you should always be stacking because the correlation considerations are going to dominate any projection considerations you [have 00:07:22]. Individual performances are mostly a function of how well the team performs, so if a team does really well, almost all the players on the team are going to do well. But if the team does badly then almost all the players in the team are going to do badly, there’s not really those one-off games, like in baseball, where one guy hits a home run and gets a lot of points but the rest of the guys get zeros, it’s really all the players on the same team really moved together. So, because stacking is so good, and because we want high upside, it’s really important to find low on stacks and the players who have the most variance.

Danny Steinberg:
So, in terms of correlations, basically, all players on the same team are highly correlated, but there are some that are more correlated than others. Normally the Top lane is not as positively correlated to the rest of the players on the same team as something like Support and ADC and Jungle and ADC are very correlated to each other, but all the correlations are really high where it’s like, I think we found in certain contests, or in certain game formats like best out of three and best out of five, you have a .9 correlation between Support and ADC and ADC and Jungle, which is just absolutely ginormous and with Top it’s .7 or .8 so it’s still very high but it’s just not as high as the ADC/Support correlation or the Jungle/ADC correlation. And with that same thing in mind correlation between players on opposite teams are very negative so you should almost never play players on opposite teams with maybe a very rare exception, which we’ll get into later. So because there’s super high correlations and high variance, the average projections are not actually that valuable, correlations dominate projections so you’re just stacking as much as possible just reiterating.

Danny Steinberg:
So, the role of ownership and uniqueness. So there are not a lot of possible different lineup combinations in League of Legends. So for the most part, because of only seven spots to choose from, you have six players and a team. That’s a lot less than in baseball or NBA where you have a nine or 10-person lineups. So you’re going to have a lot of duplicate lineups in this contest. Often you see an 100 way or 1,000-way tie for first in some of these contests, and if you have a duplicate lineup in a GVP really hurts your EP, so because negative correlations are so strong it makes a lot of sense to use a stack opposing a team that’s going to be high owned, or just play players that you think it’s a good stack, that’s going to be under-owned that still has a lot of upside.

Danny Steinberg:
So there’s different leagues, too, to consider when you’re playing daily fantasy League of Legends. There’s the American League, the European League, the Korean League and the Chinese League and the Korean and the Chinese leagues are normally bunched together in one slate, or sometimes are bunched together in one slate, and the European and American leagues are also sometimes bunched together in one slate. The Korean leagues and the Chinese league LCL and LCK tend to be the most aggressive and most entertaining leagues to watch and often have way higher scoring than the American or the European leagues, but the strategy is mostly the same, you’re mostly just trying to find the good low on stacks or good stacks to leverage off of.

Danny Steinberg:
The Leagues have different structured matches, so some of the leaves are best out of one, meaning they just play one game, but other leagues are best out of three and sometimes they do best out of five during playoffs, so it’s important to realize if the contest you’re playing in is going to be a best out of five, a best out of three or best out of one, we found basically that correlations, increase the more games they play. So, players on the same team in a best out of five matches are going to have much stronger positive correlations to each other than players on a team playing a best out of one match. So, in general in a best out of five matches or a best out of three you’re going to stack as much as possible, in a best out of one, it may make some sense to not stack but we probably still recommend it, but it could make sense to not go a full on 4/3 stack, and maybe go to a 4-2-1 to give yourself some leverage.

Max Steinberg:
I also just want to add that best of five and best of threes, because of how bonus points work if a team sweeps another team or wins before the series is over, DraftKings awards these players bonus points. And so the projections are going to differ a lot more in a best of five or a best of three. And so especially with those, finding stacks that are under the radar but do actually have high upside are going to be really, really important that under the radar is usually going to mean that if you look on Pinnacle or some sports book, the team doesn’t look like a big favorite, but they actually have a decent chance to win.

Danny Steinberg:
Yeah. Good point, Max.

Andy Baldacci:
How are you adjusting your strategy based on the size of the slate itself?

Danny Steinberg:
Yeah. So, normally the slates in League of Legends are really small, sometimes you’ll only have two games. A big slate is five or six games. So basically bigger slates mean you don’t really have to worry about duplicate lineups and there’s going to be less concentrated ownership. So you really just want to maximize stacking and upside and picking the best guys. With a smaller slate, like a two game slate, there’s going to be really concentrated ownership and there’s going to be a lot of duplicate lineups, so it makes sense to maybe emphasize uniqueness more there. Maybe you fade ownership a little bit more than you would do in a bigger slate, and even just leaving some salary on the table to give yourself uniqueness in a small slate I think makes a lot of sense.

Andy Baldacci:
And this is something that Max will get into as he walks through a build for a two-game slate, but I really think is something that not enough players are thinking about or talking about is just how invaluable uniqueness is in these smaller slates. I mean, we’ll all kind of complain when we do have that 100 or even multi-hundred-way tie for first place but that’s where a lot of the conversations stop. I don’t think people really grasped how horrible that is for your EV, your expected value, and your ROI. Can you just touch, either Max or Danny, just briefly on why it matters so much, why having duplicate lineups is just such an ROI killer?

Danny Steinberg:
Okay. I think I have a good answer for this. So let’s just examine a toy daily fantasy game where you’re just picking one player, and it’s a 10-person contest. So let’s say it’s just winner take all, it’s $10 buy-in versus $100. Okay. And let’s say nine people have picked the exact same player and one person picked a different player. So there’s basically two outcomes, either the player the nine people packed does better than the player the other person picked and there’s a nine-way tie for first, which in that case everyone wins $2 apiece or something or $5 apiece. And then the other scenario is where the other player does better than the one that the other nine people picked, and that person who picked the unique player will win $100 outright. So you can see how not having a duplicate lineup greatly increases your EV in that situation, especially if that player is actually decently likely to have a better game than the player that everyone is owning. And this is just a micro-example of what’s going on with a seven-person lineup. So, if a bunch of people have the exact same seven-person lineup, then their upside is greatly decreased or the most money they can win is greatly decreased. While if you have a completely unique lineup, the amount of money you can win is equivalent to whatever the first place prize is. So I think that’s a good example [crosstalk 00:16:08] as well.

Max Steinberg:
No. I think that’s a great example. Yeah. And I’ll just add I mean, you can see it in practice when you actually play these contests. I think actually last night I played a contest where there was… I forget. In one contest, the high stakes contest even at a 10-way tie for first. And that turned a $15,000 first prize into about $1,000 first prize or something like that. I think there’s maybe a 10 or 15-way tie. And it’s really hard to get a Top lineup anyway and so if when you get the Top lineup, you’re sort of capping your upside to $1,000 in a $330 buy-in contest, you’re going to lose money really quickly. You need to be able to have a lineup that’s going to, when it does well, is going to win that 15,000 because even if it’s a lot more unlikely, that’s still such a huge disparity that it’s just worth it, you can tell just intuitively. And there’s a lot of math to back it up, obviously as well.

Danny Steinberg:
Yeah. Great point, Max.

Max Steinberg:
Yeah.

Andy Baldacci:
Yeah. I think that makes sense and Max is going to dig more into that as he walks through the build and just showing you how to try to kind of optimize for uniqueness. But one thing I wanted to talk about is, I had mentioned that the scoring is identical between DraftKings and FanDuel, but there are some differences in terms of lineup construction Can you talk a little bit about those, Danny?

Danny Steinberg:
Yeah. Okay. So the scoring is, I believe, exactly the same on DraftKings and FanDuel.

Max Steinberg:
Yes.

Andy Baldacci:
Yeah.

Danny Steinberg:
There’s two real main differences one is on DraftKings they allow you to use a 4/3 stack which means you stack four people on the same team and three people on another team. So you’re only using two teams total on your lineup. On FanDuel you have to use three different teams in your lineup, so the most you can stack is a what’s called a 4-2-1 where you’re doing four people on the same team, two people on another team, and then one person on a different team. The other difference is that in DraftKings, when choosing the captain, they increase the salary of that player than if you were to choose him in a different position, and on FanDuel all the salaries are the same, no matter what you choose.

Danny Steinberg:
So, on FanDuel you’re basically just putting your best player in the captain that’s in your best data pack, no matter what, there’s no reason to not do that because you’re not losing any salary by doing that. With DraftKings there’s a lot more economization to take into consideration, where maybe it makes sense to punt a little bit in the captain spot to give yourself more salary to use in the other spots. So those are really the two main differences,

Andy Baldacci:
That was great, Danny. Thanks for all of that and I think people really are going to now kind of get a much better understanding of just the things to pay attention to at a high level of what goes into those winnings lineups. And now Max, why don’t you take over and just show how to actually put that into practice and build the real lineups. Do you mind taking over here?

Max Steinberg:
Yeah. So let me just explain this model. So I actually created the League of Legends model for SaberSim and I think it’s really great and it’s very sophisticated and it’s based on a lot of data, we’ve actually gotten a ton of League of Legends big data going back to even 2016, and the machine learning model takes a lot of this historical data and creates protections from it. And so it’ll use simple stats like moving averages of kills and assists, it’ll also take into account matchup stats, which I might get into a little later, gold earned, damage, things like that. And I think the thing that I’m most proud of what this model is I actually created my own Elo model to handicap match up. So I actually have my own proprietary-

Andy Baldacci:
Can you explain what Elo is?

Max Steinberg:
Yeah. I mean, it’s something that you’ll see on something like FiveThirtyEight for NBA, it’s basically a ranking system. It’s every time a team plays, based on the rating, whoever wins or loses or whoever has a better game score, based on how you quantify what a game score is, their rating will change. So if a team in League of Legends wins and they win really fast, they’re going to get a really good game score and if someone loses really quickly, they’re going to get a bad game score. And if they lose a close game that’s not going to affect their rating as much as they lose a really terrible game and get crushed. So basically, this system also goes back to 2016 and basically every time a team plays each other the Elo rating changes and now we have a pretty sophisticated grading system and it actually differs from what you might see at a sportsbook.

Max Steinberg:
Sometimes, it has just a different perspective on the teams and. And given that the limits are pretty low of League of Legends it actually makes sense that a model like this can actually quantify things better than a sportsbook. So we use that Elo model when trying to get expected win, and we also use what’s from the sportsbooks and what that does is it uses that sportsbook data that a lot of good projection systems you’ll find out there use, and it also uses this Elo model and so it’s going to have some differentiation from what you might see in other projections. Some things that my model might be missing is… one thing that’s really hard is champion selection and this is something if you know League of Legends really well is players on a team choose their champion before the game and that’s really hard to predict and I actually don’t know how to do it. But if you actually have a good grasp on it, there’s some champions that are going to be better than others and especially with Support and ADC and Mid, the roles within those can change given what champion they choose. So if you have something to predict that, that’s going to be something you can use to actually adjust the projection of the model.

Max Steinberg:
Also, looking at line movement, we just take lines at face value and we take my Elo model but you know if you’re analyzing line movement you’re saying, “Oh. Invictus started out as an underdog and now their favorite. Okay. Maybe they’re stronger than I think,” we’re not going to take an account, we’re just going to take whatever the line is and put it into the model. So, let’s just get into this a little bit with, in general, what you’re going to do when you’re going to start a build. So one thing with League of Legends is some teams have many players, some teams just have the five players they use, some teams have more than that, and so they’re going to have a starting lineup which they announce before the game. And so how we handle that is basically, if a player started the last game, they are not going to get a questionable tag, if they didn’t start, they are going to get a questionable tag and so you’re going to just want to straight up remove these players, which I’ll do a little more in a bit, from consideration because we do not want to use them.

Max Steinberg:
But if you see it starting lineup change, that’s going to allow us to actually put that player back in and maybe remove the player that we thought was going to start and so we just wanted to give you that flexibility. There’s also some certain situations where you’re playing a best of three or a best of five and luckily in this slate it’s just a best of one so whoever starts is going to start for that game and you’re fine and you’re going to know that before the game. But there’s going to be some situations when you’re playing a slate where this it’s a best of three or best of five series where some teams actually like subbing players in and out during the series, and that’s where you’re going to want to look at Twitter, you’re going to want to look at our Slack channel, especially, to just see what players are in danger of that happening, and just remove them from consideration entirely. So, in terms of building, what we’re going to do is we’re just going to remove all the questionable players first and this will take about two seconds and we’re just going to do this. And-

Andy Baldacci:
Actually, one thing that I want to add is just when it comes to the starting lineups. I mean, practically all eSports are similar to this where there isn’t kind of a clearinghouse of these starting lineups where there’s not going to be an official source you can go to and to say, “Okay. Who is going to be starting today with 100% certainty,” things like that. So that’s why we have our system in place for marking certain players as questionable, and why we’re not able to do something more automatic. So what we do recommend doing is double checking on Twitter, to see if the teams have put out who’s going to be playing, check in Slack, places like that, just to get information to make sure that nothing was missed there.

Max Steinberg:
Yeah. Yeah. Absolutely. So actually, let me just go over the projections for a second so I can sort of show you the machine learning model in action because if you go onto another DFS site that has projections, you’re going to look at the lines of this game and we see that JT Gaming is a decent favorite over Gen G and DragonX is a closer match with Invictus, and you’re probably going to see that for most players JDG players are just going to be the best players because a lot of what goes into a player’s projection is just how likely they are to win in the game. There’s a huge correlation between a team winning, and how high their projection is and how many fantasy points they score as we talked about earlier.

Max Steinberg:
But you can see in my model, actually, DragonX, the ADC actually, has a higher projection than JDG, and that’s for a couple reasons. One is because my Elo model actually likes DRX a little better than IG, whereas Pinnacle actually likes IG better than DRX, and that means that DRX is going to be projected a little higher than you might find other places. IG is also a pretty good matchup in terms of matchup stats, so I have my own match of stats there’s a lot of people look at this website Oracle’s Elixir for matchup stats as well, and it’s in terms of combined kills per minute… and this is basically a pay sort of engage rate stat. So Invictus Gaming has a really high combined kills per minute, which means that they’re sort of aggressive, they try to engage a lot.

Max Steinberg:
And because of that they’re a good matchup and so for a team like DRX, i they actually are a little better than IG and have this good matchup, their players are actually pretty good players. And so inherently just by if my model is correct, which I’m confident that it is and does a good job quantifying this, you’re going to have an inherent edge in general because DRX probably is not going to be that high owned of a sack, but actually our model is saying that a lot of their players are really the best players to use. So I think, just by having projections that use an Elo model and use matchup stats and use good stats historical data, you’re going to already have an edge on the field which I think is really great.

Danny Steinberg:
So I think League of Legends sports betting has been around for two months or something. Basically, the market is not mature, there’s not a lot of betting on it. I don’t think there’s really any professional sports bettors, this really drastically differs from almost every other sport. So I think with League of Legends specifically, I think there’s no real good reason why we should expect the lines to be really good. So I think the fact that Max has added this Elo model is going to be a really nice edge for our projections in general.

Andy Baldacci:
Yeah. And just expanding on that, there’s a couple things I want to touch on. The first is that when we talk about Vegas, we talk about odds, it’s kind of assumed that those are the gold standard and for the big sports, they really are. You need a very good reason to significantly diverge from those for baseball. We’re very common in our model and we’re confident in diverging from the market, because we have put so much time into that. But there are still spots where we’re going to defer to Vegas and then for the other sports, it’s just really hard to beat them. And that’s shown by the limits that the books give you on how much you’re able to bet on these different games. While I might be able to theoretically bet 50 or even $100,000 on a football spread on Sunday morning, on League of Legends, before the game starts, I think the most I’ve seen is maybe $500?

Max Steinberg:
It’s more like 200. Yeah.

Andy Baldacci:
It’s between 100 and 250 throughout the day and then sometimes I’ve seen again it little bit higher, right before the match starts, but what that means is that the books aren’t as confident in their models, but also that there isn’t as much action being put down to kind of let the market decide what the right odds are because that’s really where the strength of the books come from. And so because of that, it’s not as though we’re saying, “Ignore Vegas. Ignore Vegas,” but we are saying listen to them, but make sure to listen to other things as well and that’s what this gets.

Andy Baldacci:
And then the second point just building on that is that we’ve talked about how important uniqueness is and so having a model that differs from kind of the market, that differs from what the majority of people are using, is going to just automatically get you some uniqueness. You’re not just feeding the same projections into the same optimizer that everyone else is using and getting the same lineups, you’re getting kind of, not even randomness, you’re getting variance from the market projections in a smart way so there’s just a lot of power in this model that Max has built, but I’ll let him take back over from here.

Max Steinberg:
Yeah. And so, as always, you can adjust these projections if you want, if you have some sort of data you’re looking at like analyzing line movement, analyzing champions selection. You can adjust these, I actually rarely do, and then we don’t have ownership projections but you can add your own, and I think it’s valuable actually. I mean, even if you do something like giving the biggest favorite all 40% ownership and the person who’s second biggest 30 and just going down the line like that, doing something is going to add value to your build. But for now I’m just not going to do anything with ownership.

Max Steinberg:
So, another key thing that we want to look at is captain. So I’ve looked at the data in best of one games, just individual games, and about 75% of the time in individual games, ADC or Mid has the highest fantasy point out button in that game. And so in general with the captain, you kind of just want to favor having an ADC or Mid, and you can make some exceptions if Jungle, Top or Support is pretty close. So we’re just going to look at each team and we’re going to say, “Okay. For Invictus, Mid and ADC are clearly higher than Jungle and Top.” So I’m just going to remove Jungle, Top and Support and especially Team which is the lowest projection from being considered in the captain. We’re going to go to JDG.

Max Steinberg:
So you can see here actually Mid and Jungle are pretty close. So, again, we’re talking about lineup uniqueness, you might want to uncheck Mid and keep the Jungle here and just consider two players for that. So you look at JG, so you have Roller, who’s clear favorite and then Mid BDD who also is a huge favorite, so we’re just going to consider ADC and Mid. And I think in general if you just considered ADC and Mid and you didn’t even look at this, you’re going to be fine, but if you want to get more granular you can actually look at this. So again, huge difference here so we’re just going to say okay fine, let’s uncheck all of this.

Max Steinberg:
So that’s all you really have to consider. You can adjust projection, you can adjust ownership. But in terms of reviewing the projections, this is really all you need to do. So now the Build Settings so this is what I love about SaberSim is usually you’re going to go to a lineup optimizer and you’re going to add these stacking rules and we allow you to do this, obviously. But given that we have this correlation data, and we can include ownership fade and you can do something like smart randomness to consider upside, I’m not going to use stacking rules at all and I think it’s actually something you can learn by actually looking at the lineups is what SaberSim lineup builder is actually doing. And so I’ll get into that just a second but like all sports you’re going to put correlation to very high, and I’m just going to put smart randomness to a whatever setting, because I actually want to just make sure that my projection differences are considered as much as possible.

Andy Baldacci:
Yeah. So the one thing I want to add on smart randomness is this is something newer that we’ve added to sports, we don’t have full simulation data for it. When we have simulation data we are able to use we call smart diversity, which samples different parts of a player’s possible outcomes from our 1,000 simulations to get a better range of possibilities in there for your lineups whereas with smart randomness, rather than just using what other optimizers do, which is true randomness and just randomly adjusting projections for you to get more diversity in your lineups, what we do is we apply randomness following what’s called a normal distribution, to make sure that you’re just not getting crazy numbers in there.

Max Steinberg:
Yeah. And so, with smart diversity, I almost always put it all the way up, but with smart randomness, which is a more simplified version, I’m not as keen as using it as highly but correlation I’m going to use very high. Two other quick considerations, max salary, we talked about having unique lineup. The chance of you having a unique lineup goes up as the max salary you use goes down. So if you leave $500 on the table, there’s going to be a lot higher of a chance that you have a unique lineup, then if you go to the full 50,000. So, what I usually do is leave at least 500 on the table, sometimes I do more, but we’re going to leave 500 on the table.

Max Steinberg:
And the last thing is, we have this box about allowing players on opposing teams. So, in general, you don’t really want to use players on opposing teams, especially on slates that are three, four, or five, or six games where you really don’t need to. However, on a two-game slate, this is something that you might actually want to check and so I would suggest most of the time that you uncheck it but I’m going to check it just to show you how the SaberSim builder actually will end up building the lineups in a way where they actually put players on opposing teams and just look at what it’s doing.

Andy Baldacci:
And on FanDuel you actually have to check that because to have a valid lineup, you need to have players from three teams.

Danny Steinberg:
Good point.

Max Steinberg:
Absolutely. Yeah. So on FanDuel that’s just going to happen. Okay. So, we have our lineups right now and of course SaberSim has a really good visualization of not only of your lineups that you’re actually getting, you have the exposure percentages, you see what teams are stacking and you can see DRX is a really popular one. And you can see the players you’re using. And so one thing that’s different in baseball, football, hockey, you can actually adjust these exposures but with League of Legends, in order to actually get new lineups or adjust your exposures, you’re going to have to adjust projections and do an entirely new belt.

Max Steinberg:
But so let’s just look at these lineups for a second because this is how I in general with League of Legends we’re going to do quality control is, first of all, if you look you can see, “Okay. A lot of these stacks are 4/3 stacks. Like we talked about, it makes sense. We’re four stacking with the captain on JDG, and then doing three-stack with DRX and you’d see a lot of design apps are going to be 4/3 stacks and again, players who are not on opposing teams at all. However, sometimes, you are going to get a little bit of different things. And you can see here is, there are some 4/2 stacks and so 4/2 stacks on a two game slate are going to inherently have a player from an opposing team. So you can see here we have a four stack with JDG, we have two players from DRX, and then we actually use the Support on IG.

Max Steinberg:
But if you look at Saber Score, which is the way we quantify how good a lineup is, there’s actually a pretty big drop off once we get to the points. So Saber Score 460 very high and all of these are pretty close, and then suddenly there’s a pretty big drop off and these are still 4/3 stacks but these are going to be less optimal lineups according to our builder. And then once you get to 4/2 stacks, there is actually a big drop off, it’s now Saber Score is 340, 349, 366, you go even lower it’s about 300. However, the question is, “Is this worth it?” On a two-game slate, where uniqueness is really important, it might make sense to actually use one of these 4/2 stacks, but you really want to analyze where the drop off is because once it starts getting into these 290, 250 range, that’s when I don’t really want to use this lineup.

Max Steinberg:
So I think in general, it’s better to actually do a build before you actually decide how many lineups you’re going to put into one of these contests like the Shock Blast, because you could just say, “Oh. Well, I can put 150 lineups in this contest. Let’s put 150,” but if she gave me League of Legends plate that’s a really bad idea because you’re eventually going to put in some bad lineups because there’s just not enough options. And so at this what I usually do is look at the Saber Score and just cut off when there seems to be a big drop off, and even so you can actually cut off where it says 4/2. So you could only say, “I’m just going to do 12 lineups or make sure that my Saber Score is above 400 and cut it off even more.”

Max Steinberg:
But that’s something you’re going to want to analyze because in general there’s not enough of an edge especially a two-game slate to put in just infinite lineups, you really have to look at where the lineups are really starting to fall off and I think that’s why Saber Score is such a good thing to look at is it’s not just taking you consider projections which are changing but it also is considering that correlation factor and that upside factor. And once those factors start getting really low, that’s when you sort of want to start cutting yourself off and entering your lineups.

Andy Baldacci:
And one thing I want to add just around Saber Score is that it’s important to kind of understand what it should and shouldn’t be used for. And, as Max said, this is our way of trying to quantify upside the upside potential of a lineup, beyond just average projections. But the way Saber Score works is that this isn’t necessarily an absolute value, you don’t want to set a hard limit for all of your bills that I need Saver Scores above this. The way Max is using it, though, is great because you want to kind of use that as a measuring stick within your builds but not necessarily across a bunch of different builds.

Andy Baldacci:
And that’s just something to keep in mind where, again, it’s not as though these exact numbers that Max is highlighting are what your target should be, but it’s looking within the builds that you’re doing and just trying to see, is there a cliff that these lineups fall off, whatever that number may be, and then asking yourself, “Do I think that there’s enough edge there for me to put them in,” and maybe that they have things like a lower salary that maybe will make them more likely to be unique and this and that, that will put you in the spot where you’re comfortable entering those lineups but, again, this is something that we see it all the time, especially amongst top pros, where they just default to entering the maximum in every single contest, without thinking about the other factors involved. And I mean, when the rake is often 15%, or more, when you’re playing contests that might be smaller, when you’re playing smaller slate sizes, is not a given that you can put in that many profitable lineups So, just follow Max’s advice here to use a little bit of caution on figuring out how much to enter in a contest.

Max Steinberg:
Yeah. And I just want to add more thing to salary is when we’re building lineups for you, we’re going to order by Saber Score but we do know there’s another factor that’s important for unique lineup which is salary on the table. So when you’re looking through this and if you want extra control in terms of picking what your top lineup is, look at the salary number because in general, I think that the more salary leave on the table, the better for your uniqueness and so in general when you see a Top lineup spread where it’s just a couple of points of Saber Score between these two things, I probably actually am going to favor the lineup that leaves some salary on the table, more salary on the table, because it just has that much greater a chance of being unique.

Andy Baldacci:
Awesome. And so from here, are there any other steps you take or are these basically going to be the 12 that you want to put in?

Max Steinberg:
These are going to be the 12 I’m going to put in and I’m going to download my lineups, upload them into DraftKings, and then I’ll probably actually enter some of the bigger stakes contests probably with the second one, instead of this first one because of that salary left on the table, but this first one obviously is a great lineup, too, so I can do that one as well. It just depends on how many entries are in the contest and how much I’m going to value a unique lineup.

Andy Baldacci:
And then one extra thing that just came to mind is that this is a 3:00 AM slate. Obviously we’re still a ways away from that, it’s still daylight hours out so I’m guessing that you’re going to be up for a bit before going to bed and putting in these lineups, but when are you typically putting your lineups in?

Max Steinberg:
Yeah. I mean, usually starting lineups are announced several hours before the contest, so as long as, what we talked about before, you’re not playing any players that have a chance of being subbed, and I recommend going on our Slack channel just to make sure that’s the case, we have some people who have really up to date information on that. And those starting lineups are announced you can feel pretty confident for locking in lineups four, five hours before and it’s totally fine. I’ve done that very often, I’ve yet to once have a situation where someone had a zero or wasn’t playing. It’s not like these guys are getting injured so you can really put them in several hours before and be pretty confident.

Andy Baldacci:
Yeah. And that’s one thing that is different here and it also kind of goes back to what Max was saying about excluding the questionable players where, if you’re doing it far in advance, you want to be as safe as you can and excluding them is a safe way of doing that. As you get closer to it, and the information comes out, you can get a good idea of who’s actually playing and because the sports aren’t held outside because it’s less of a physical activity, to say the least, there aren’t injury concerns and this and that so the lineups aren’t changing a ton. So you can put them in earlier than you could, say, an NBA, or even a KBO lineup.

Andy Baldacci:
But with all that said, I think you did a great job of walking through this process and just giving people an idea of how to put together all the aspects that Danny talked about into actually winning lineups. But one thing we always want to stress is that while this is a great starting point, the more you can evolve your process and the more you can add to what you do to whether it’s research, whether it’s quality control, whatever it may be, the better you’re going to do, but I think this is enough to give people a really strong starting point to start taking advantage of these contests because, I won’t speak for the other eSports I think some of those are questionable in terms of profitability, but these League of Legends contests, because there is so much involved into proper lineup construction, these have been really profitable. And so I’m just going to recommend you guys check them out and I think this has given you enough to get…

Andy Baldacci:
All right. To kind of wrap up what we covered today, I’m breaking it down into the eight keys to beating League of Legends. The first one is just the most important by far and that’s stacking. If you take nothing else away from this, if you don’t use projections, don’t do anything else, if you just want to make good lineups stack. Stack as much as the sites will allow you by default, and you’re going to be in a pretty good spot. And I’m not saying that’s always what you should do, but if that is kind of the one thing you could do, you’d be okay. The second thing is, again, avoid players on opposing teams unless there is a very strong reason not to do so or the site forces you to use, like FanDuel, three teams when there’s only two games going on.

Andy Baldacci:
Variance is super high. So that makes the value of average projections go down and the importance of lineup construction go up to make sure that, one, you have more upside but two you can focus on uniqueness, it’s playing opponents of highly owned teams. The fifth key is paying attention to the match structure and by that I just mean if it’s a single game match or if it’s best of three or best of five because for each of those different structures, there are dramatically different adjustments that you need to make and that we covered earlier in this video. And six is focusing on uniqueness in the small slates especially, but really focusing on it across the board. It’s easier to find unique lineups the larger the slate is, but that doesn’t mean you should just set the min and max salary to 50K and just throw everything that comes up directly into the contest.

Andy Baldacci:
Seventh is always putting the best player in captain on FanDuel, that’s really the only adjustment you need to think about between the two sites, other than the three-team minimum. And last is just use the right tools. We’ve covered a lot here and we’ve really also shown that when using SaberSim, this doesn’t have to be that hard to do. You don’t have to spend hours setting up all kinds of rules and groups and getting your exposures just perfect in order to get lineups that have any chance of winning. You can do this almost out of the box, because we have software that makes it really easy to use the models we have under it, like Max has put together, and also just to get those correlations, to get the ownership, to get those other aspects of upside into your lineups.

Andy Baldacci:
And if you want to use SaberSim until one of the major sports returns, being baseball, basketball, football or hockey, SaberSim is 100% free, you can get access to our projections, our optimizer, everything that we have on the site. And we’ve got a lot on there now, we have League of Legends, CS:GO, Rocket League, KBO, NASCAR, MMA, golf, soccer, and more is coming out. So to get signed up for free, you don’t even need to put your credit card in anything at all, just head to SaberSim.comm and get started. But thank you so much for joining us today. Thank you, Max. Thank you, Danny, for sharing everything. And we really appreciate it and hope to see you around in SaberSim.