What Happened to Flash Wolves?
By Colin Bowers
Filed under eSports
StatXP’s first big project upon launching is our ongoing coverage of the 2018 League of Legends World Championships built around our Elo model. It was very important that the model do well and so far, it has. We’ve correctly predicted 6 of 8 teams that advanced and 34 out of 51 matches overall. Several of the misses in that 67% success rate were 50/50 games where our model was not overly confident either way. Despite this success, we did have one huge miss in Flash Wolves and it is important to be upfront about your misses. FW was our #1 ranked team entering the tournament and we gave them a 24% chance of winning the whole thing. What happened?
If you’ve read StatXP for any length of time, you know that this is not the place for a nuanced discussion of LoL strategies and tactics. However, we are equipped to look at the likelihood of Flash Wolves losing in the manner that they did. FW started their Worlds run as expected, beating PVB and AF. After that, disaster struck and they lost 4 of their next 5 games, including a game versus PVB in which StatXP gave them a 92% chance of winning.
Looking at actual game performance, FW lost two games that they could have won based on the margin of victory. The StatXP model uses kill-to-death margin as a proxy for margin of victory in order to give the model some additional granularity. Kill-to-death margin is not a perfect proxy but in our database of games going back to 2016, teams that come out ahead on kill-to-death margin win 95% of their games (green means the team won and blue means the team lost).
As expected, the relationship between kill-to-death margin in LoL and winning follows a logistic relationship with a particularly steep drop-off (meaning that teams with a negative kill-to-death margin almost never win).
Revisiting the table from the top of this column, FW lost two winnable games, including one in which the margin of victory would suggest that they only had a 16% chance of losing!
Using kill-to-death margin, we’d expect a team with FW’s group stage performance to win 4.26 games rather than the 3 that they won. Had they won another game, they would have advanced over G2 Esports and kept the hopes of the StatXP Elo model alive.
While FW was definitely unlucky, its clear in hindsight that we also had them overrated, a possibility we noted at the beginning of our tournament coverage. While we mourn the loss of our fallen favorite, the StatXP model has moved on to bigger and better teams.
Check out our knockout round predictions and follow along with StatXP on Twitter (@StatXP) as we continue to model and forecast the 2018 LoL Worlds.