The finals of the LEC 2020
happened this past weekend, with G2 being crowned winner for the second year in a row. The top three teams - one more than 2019 - will participate in Worlds this year, while the fourth-placed MAD Lions will be in the Play-ins.
G2's path to the play-offs was by no means simple or, in fact, easy to understand. In the Spring Split, they finished first in a double round robin tournament - that is, they played each team twice and came away with the highest amount of wins. The top six teams from the group stage went on to the play-offs, with G2 being the top seed. Despite dropping to the loser bracket, they managed to come out winning the Spring tournament and receiving 90 championship points.
In the Summer Season, G2 played in another round robin, this time finishing only third. This brought them only 70 championship points (top-placed Rogue got 140), but still ensured they would start in the winners' bracket in the play-offs. Losing to Fnatic in the Semi-Finals, they came back from the losers' bracket to play Fnatic once more in the finals - and win.
LEC Summer Season table and play-offs, courtesy of Gamepedia
Round robins, championship points awarded differently between seasons, seeds and brackets. Sounds complicated? That's because it truly is. It's also different to the way LEC was played in 2019, and different again to how it was played before that. League of Legends as an Esport has been around for a while, but it's still a baby compared to sports like basketball or soccer. It's understandable that Riot is still looking for the best format to put their teams into.
There are many reasons why a tournament might be set up in a certain way. Logistics of moving the team around, getting a certain amount of matches, hype build up for the finals are just some of them. Esports tournaments are great spectator events, even more so now that traditional sports have had to take a step back. Ultimately though, leagues and tournaments are also competitive sporting events. The goal of the LEC is to determine the best teams in Europe and to send them to Worlds. And as such, it has the same objective as any serious competition: A tournament set up should always try to maximize the chances of the best team winning.
So while the complexity of the LEC 2020 is a clear downside - it takes a long time for fans (and, frankly, us) to understand all the rules - the upside might in fact be that it is better at determining the right winner. This is a theory we did not want to leave untested. Instead, we decided to run a theoretical model of the LEC 2020 format to explore it.
In our experiment we ran Monte Carlo
simulations of LEC 2020 and LEC 2019. Since we want to compare them with more traditional formats, we also simulated a two-stage tournament. For this one, we had the teams play four round robin rounds (corresponding to the Spring and Summer round robin stages). The best eight teams then played a double elimination play-off stage, with seedings given by the round robin results. Finally, we also simulated a double round robin tournament without playoffs as a baseline. All simulations ran 10^7 iterations.
We used the same tie-break rules as LEC 2020 does. For the play-offs, we assumed that the first seed will always choose to play against the weakest team they can pick.
Ten fictitious teams participated in each tournament. In real life, it is not possible to know exactly how strong the participating teams are - that's why we have tournaments in the first place, and that's also why rating systems can become very complicated
. In our experiment, however, we could have perfect knowledge of our participants' skill and win probability for any possible match.
We used Elo
to express the team skill and ran two different setups:
Strong favourite: Elo values range from 1600 to 2400. The probability of the team ranked n-th to win against the team ranked n+1-th is 0.62. The probability of the top ranked team to win against the 10th ranked team is 0.99.
Weak favourite: Elo values range from 2200 to 2400. The probability of
the team ranked n-th to win against the team ranked n+1-th is 0.53. The
probability of the top ranked team to win against the 10th ranked team
Of course, our experiment setup is not perfect. In the real life, the
teams are not spaced out evenly on an interval. Some teams are very
closely matched while others are clearly dominant against one another.
This is why we see teams going back and forth, like G2 winning and
losing against Fnatic in the same tournament. We would argue that the "weak favourite" setup probably represents the reality better than the "strong favourite" for most professional esports events.
First we look at how often the best team actually ends up winning the tournament. LEC 2020 does not do well here - the strong favourite wins only 57.4% of the time. This is the lowest
win rate of all our simulated tournaments. It's marginally better for
the weak favourite, but LEC 2019 and round robin formats still do better
at determining the absolute champion. It's quite surprising that the strong favourite only ends up winning the tournament in roughly 2/3 of the simulations - at least until we remember that they are very likely face the second strongest team at some point in the playoffs, and only have a 62% chance of winning against them! This also explains why the round robin does much better in this metric. Without playoffs, the two strongest teams only play each other twice.
Odds of strongest team winning the tournament
Probability of the favourite going to Worlds
In order to have a spot at Worlds, the strongest team needs to
finish at least third (for simplicity, we are going to assume the same
for LEC 2019, where in fact a top 2 placement was needed). Here LEC 2020
can truly shine: With almost 75% rate for the weak favourite, it
outdoes all other tournament formats. For the strong favourite, only the
round robin does marginally better.
The fact that LEC 2019 performs
worse than LEC 2020 is not surprising if we remember the rules. The
teams played the same type of final as this year with the winner going
straight to Worlds. The other teams then played a gauntlet, meaning that
if our top team placed 3rd in the play-offs, there was still a decent
chance of them not making it to the final top 3.
Odds of strongest team qualifying for Worlds
Probability of an upset
We want to send the strongest teams to Worlds. We definitely don't want to send the weakest team in our pool to represent Europe. Luckily the odds of the weakest team winning the event are negligible for all tournament formats. But what about the team qualifying?
Odds of weakest team qualifying for Worlds
For the strong favourite, the qualifying rate still rounds down to zero. Not so for the weak one. The 1.3% qualifying rate means that if we run the tournament 100 times, the weakest team will qualify exactly once. The same team is almost four times as likely to qualify in a double elimination format - since the top 8 and not 6 teams join the play-offs, the chances of our bottom ranked team still placing go up a fair bit.
The odds of the 4th ranked team to qualify are, of course, significantly higher. LEC 2020 has the highest likelyhood of all to bring them to Worlds. Like everything we've seen before, this is inherent to the tournament format: The 4th ranked will almost always place 3rd or 4th in the summer split, meaning they will (in our simulation) almost always end up playing in the winners' bracket. They will need to win at most three (and sometimes only two) matches to get to the top three. In the 2019 format, with the same placement after the group stage, they will likely need to play 3-4 matches to end up in the top 3.
Odds of 4th ranked team qualifying for Worlds
Like all tournament formats, LEC 2020 has its benefits and drawbacks. Our statistical analysis shows that the main benefit of the format is that it will reliably send the strongest team in the pool to Worlds. The round robin format arguably achieves the same thing - but it lacks the excitement that comes with having play-offs. By having only six instead of the standard eight teams, the LEC format minimizes the chances of a weak team sneaking into the top 3 and creating an upset. Here the gauntlet format of the 2019 season did a better job of gatekeeping, making it harder for the 4th ranked team to play its way into the final three.
One particularity of our experiment is that the teams do not change their skill level during the experiment - if you were the strongest team at the start of the series, you are still the strongest team at the end. In real life, things get more complicated: Roster changes, patches, and motivation are just some factors that influence team performance, and the number 1 ranked team at the start of the season might not be the one deserving the spot at Worlds six months later. This is probably the reason why significantly more championship points were awarded for the Summer Split results than for the Spring Split.
Do you find the LEC ruleset easy to understand? Were our results surprising? Which other tournaments would you like us to simulate? Reach out to us on Twitter
and let us know!
About the author
Darina Goldin is the Director Data Science at Bayes Esports. She
started playing competitive Team Fortress 2 in grad school. While no
longer competing, she is still an avid Esports fan. At Bayes, she has
created numerous predictive models for Counter Strike, DotA2, and League
of Legends. When not crunching numbers, you can find her at the gym
training Brazillian Jiu Jitsu.