Welcome! Previously we looked at what wins on 2 game slates, today we’ll be looking at 3 game slates.
Before I jump in, let me outline my criteria. I counted all slates from the end of the summer player break (August 26th) until the last Tier One tournament ended in December. The data includes any tournaments in Europe (including CIS) and North America, and does not include tournaments that were Asia/Oceanic or South America only, as there are often huge favorites and/or weird pricing in those and I didn’t want to skew the data (when SA teams are part of a bigger slate they were included). When I say SA I mean teams that currently play there, Furia and MiBR and classified as NA teams since that’s where they are now based out of.
I used tournaments that had total prize pools of $10,000+ and a buy-in of under $20 (the only tournaments over $10k on a given day were the big $10/12 and then sometimes the $222 or $180, the high dollar value tournaments are not included since they have so few participants).
What I looked at was what lineup constructions won slates/finished in the top 1% of slates compared to how often they were used overall. That latter piece is important because without it you may be led to believe that a certain lineup construction was better than another when in reality it wasn’t. To illustrate what I mean, if I told you (these are not actual numbers) that on 2 game slates a 3/3 lineup construction and a 3/2/1 lineup construction each won 50% of tournaments, which one would you say is better or are they equal? It’s a trick question, because you need to know how many total lineups there are of each. If 90% of the field is running 3/3 and 10% is running 3/2/1 (again, not actual numbers, just to be clear) but they each win half the time, the 3/2/1 construction would clearly be superior. Hopefully that makes sense.
The constructions I looked at are 3/3, 3/2/1, and 2/2. What 3/3 means is 3 players from 2 different teams. By default this means they are from different games and therefore cannot be a gamestack. 3/2/1 means 3 players from Team A, 2 from Team B, and 1 from Team C. There can be game stacking here, Team A and B can be from the same game, Team A and C can be from the same game, or Team B and C can be from the same game. 2/2 means 2 players from Team A, 2 from Team B (Team A and B could theoretically be opponents), and then either 2 from team C or 1 from Team C and 1 from Team D. Of course there are other lineup constructions possible, particularly on larger slates, but they’re not as common so I’ve grouped them all together as “Other”. Note that on 2 game slates, 2/2 lineups are also considered other because they require either a gamestack or a 2/1 from each game.
There’s a quick TLDR at the end again if you don’t want to sift through all the data and just want to see the takeaways. Let’s jump in.
3 Game Slate Lineup Construction
We have a 22 slate sample size, although one slate different lineups tied for 1st, so you may notice there are 23 lineups considered winners and that the percents add up to slightly more than 100% for the winners. Across those 22 slates the breakdown of lineup constructions looked like:
Unlike 2 game slates where 3/2/1 and 3/3 where fairly close in usage, on 2 game slates 3/2/1 is clearly the most popular construction. 2/2 is actually 2nd most popular. Note that 2/2 includes both 2/2/2 lineups and 2/2/1/1 lineups. The 2/2/1/1 lineups by default have at least 1 set of opponents on a 3 game slate. When we add winning lineups and the difference between usage and first place finishes we see:
|1st Place Finishes||9.09%||77.27%||18.18%||0.00%|
|Wins Minus Used||-1.06%||26.84%||-11.31%||-9.87%|
Again, it’s 3/2/1 lineup constructions dominating the 1st place finishes, winning over 75% of the time while being used by roughly 50% of players. However, the Top 1% data tells a bit of a different story. Here it is:
|Top 1% of Lineups||10.41%||53.09%||31.22%||5.13%|
|Wins Minus Used||0.26%||2.66%||1.73%||-4.74%|
Here we see a much more balanced distribution relative to the percent the lineups are used. There are 2 ways (that come to mind for me at least) to interpret this data. One is that with a relatively small sample size we have a lot of noise in what actually ends up winning, and over time the winning stats will look more like the top 1% stats. The other is that because going from a 3/2/1 to a 2/2/2 or 2/2/1/1 opens up a ton more possibilities, it’s harder to nail that “perfect” 2/2/2, and thus a lot of lineups get close but don’t actually win with that construction. My belief is that it’s probably a combination of the two. Over time I would expect the 3/2/1 winners to dip a little bit, after all 3 out of the 17 times a 3/2/1 lineup won 3/2/1 lineups made up under 50% of the Top 1% overall (although one out of the 5 slates a 3/2/1 didn’t win they did make up over 50% of the Top 1%). Over time, I’d expect the 3/2/1 win percentage to fall to somewhere in the mid-60% range, while the 2/2 lineup construction win rate should rise to somewhere in the mid-20% range.
One thing this data is quietly telling us, is that as slates get larger, playing opponents in the same lineup gets worse and worse. Unfortunately, I don’t have an easy way to track when the 3-1 or 2-1 in a 3/2/1 lineup is from the same game, or when a 2/2 lineup uses opponents. However, on a 3 game slate “Other” lineups by default use opponents, as only a 3/3, 3/2/1, or 2/2/2 can be built without using opponents. Both charts show that Other lineups are pretty substantially underperforming their usage frequency. Playing opponents again massively expands the amount of possible lineups, and creates a narrow outcome where you need negatively correlated players to both (or all if it’s more than a 1-1 from the same game) have big games. Until we have NFL sized tournaments, or at the very least NBA/MLB sized, it doesn’t seem necessary to build such low percentage lineups, as you simply don’t need to in order to win. What I mean by that is, if you build a lineup that’s going to win 1 out of every 10,000 a slate is run in a contest that has 100,000 entries, you have a massively +EV lineup. However, if you have a 1/10,000 chance to win in a contest that has 4,000 entries, suddenly you have a pretty -EV lineup. Playing opponents definitely decreases your chances of hitting a “perfect” lineup, in my opinion too much to do in these several thousand entry tournaments, with the possible exception of when a S1mple and/or Zywoo type are involved in the slate (are particularly if their team is an underdog).
Lineup Building Takeaways
Again it seems like 3/2/1 builds are underutilized relative to how often they are winning, even if you assume that over time the numbers will dip down a bit from the insane 77% win rate over this relatively small sample size. By that same token, I wouldn’t read too too much into the apparent overuse of 2/2 lineups just from looking at the win rate, as that will like normalize upwards to be within a few percentage points of how often they are used. Plus, it’s pretty likely that a good amount of those 2/2 lineups are running some sort of opponents in the lineup, which as I mentioned decreases the chances of hitting a perfect lineup. I would go so far as to venture to guess that 2/2/2 lineups without any game stacks are actually slightly underused, but unfortunately I don’t have an easy way to prove that.
I personally already build almost exclusively 3/2/1’s on 3 game slates, so while I recognize there is probably some confirmation bias here, nothing I’ve seen in the data has convinced me I should be going out of my way to build differently.
Again, let’s take a look at 3/2/1 lineups to see if there’s any more data we can glean on how to make our lineups better than the field. We’ll also look a little bit at 2/2 constructions, because I do think there are some edges we can gain there as well.
Investigating 3/2/1’s (and a little bit of 2/2’s)
The first place we started on 2 game slates was investigating game stacking, where a little bit more success was found than I anticipated. That was definitely not the case on 3 game slates. Only 0.79% of lineups used the 3/2 from the same game, and a grand total of 1 of those lineups finished in the Top 1%, good for a 0.15% rate. Don’t run a 3/2 gamestack on 3 game slates. While I didn’t have an easy way to look at all of the people who gamestacked in 2/2 lineup constructions (here I consider either a 2/2 from the same game or a 2/1 to be a gamestack), what I can say is none of the winner who used 2/2 constructions used either a 2/2 from the same game, or a 2/1 from the same game. In fact, the only time the winner was a 2/2 using opponents was when Navi and Vitality played each other and the winning lineup had, you guessed it, both Zywoo and S1mple as the 1/1 in a 2/2/1/1 lineup. Of the 17 slates where a 3/2/1 lineup won, the 3, 2, and 1 were never opponents, save for 1 time when the 3-1 was from the same game. The “1” in that 3-1? None other than S1mple once again.
Another thing we can look at is captains, and specifically whether or not winners stacked them with teammates. On 2 game slates, we noticed that when 3/2/1 lineups won the “1” was almost never the captain, which made sense because you had at least 2 of his opponents in the lineup by default. Looking at the 20 slates where a 3/2/1 or 2/2 lineup construction won (3/3 wins have 2 teammates of a captain by default), we saw 7 winner where the captain was a part of the 3 stack, 9 winners where the captain was part of the 2 stack, and 5 winners where the captain was the one off. Don’t forget we had a tie for 1st where a 3/2/1 and 2/2/2 are both counted as winners, which is why those numbers add up to 21. This distribution makes a lot of sense when you consider than even on a winning team, there’s a finite amount of points that can be had. We typically want our captain to be the highest scorer on the slate, which means taking a large chunk of those available points, taking them away from his teammates. There is of course some trade off in that the more convincingly a team wins the more total points they have available to them, so it’s not like the captain can’t be a part of a 3 stack but it definitely doesn’t need to be either. Even the one off as captain as viable, and that construction is one of my personal favorites, as it’s usually very unique.
3/2/1’s are still outperforming how often they’re used, by a huge margin if you just look at win rates, and by a much slimmer margin if looking at overall top 1% finishes. As more slates are played, I expect the number to land somewhere in the middle, meaning 3/2/1’s are again underutilized.
Playing opponents has not worked well on 3 game slates, with only 2/23 winning lineups containing players on opposite sides of the same game, so run those 3/2/1’s with 1 team from each game. Lastly, your captain can be in any part of the 3/2/1, as all 3 options have proved viable. One off captains in particular are an interesting way to get unique.
Hopefully you’re enjoying these articles, I’ll be back with data on 4 game slates and 5+ game slates soon.