Last time, we looked at the set of Continental Championships focusing on what IDs were represented during the three weekends. This time, just in time to hopefully throw a spanner into the plans of those competing in Intercontinentals tomorrow, we will try to answer what decks did well.
The two bar charts below show the win rates for the most popular Runners and Corps across the three tournaments. Note that these are more volatile than the frequency charts from the last blog post. We need to keep in mind the number of players representing that ID, for example, Adam and CtM both have win rates above 80% at Asia Pacifics but this is because that there were few people on these IDs. We will address this issue more carefully later but for now, if we are careful though, then there are still some interesting stories to find.
The most interesting stories appear in the Corp statistics although these are invariantly connected to the Runner side of the meta. Let’s start with Titan, the winner of the first Continental. At Asia Pacifics it had a win rate over 70% (and nearly 80% during the Swiss), which we wrote about immediately after that tournament, posturing what would happen in following weeks. Now, we can see that the attempt to counter the fast advance lists had an effect, with win rates dropping to around 60% the following weekends.
The other winningest Corp, Asa, showed good numbers throughout the weekends. After winning Europeans/Africans, I was a little shocked to see the number of players drop off slightly, but the win rate remained high. With only nine players in Americans, they all did rather well. After my last post, I heard that a number of players were scared of the Freedom threat, but this did not appear to materialise.
Comparing Win Rates
As mentioned, the biggest issue with these bar charts is that we haven’t taken into account the number of players. I blogged in the past about placing error bars on win probability estimates but that has been using the entirety of the meta to get good confidence intervals. With these smaller tournaments, we are much less certain so I’m going to pass the buck there to https://knowthemeta.com to address that issue. Instead, we can test whether there is strong evidence that the win rate has changed between tournaments without being too sure of the exact win rate.
We shall look at the changes between Europeans/Africans and Americans as these have the most data to play with. Without going into too much detail, for each ID we calculate the probability of the results in the two tournaments in two different ways. Firstly, we assume that there is one probability of winning shared by the two tournaments and compare this to a model where each tournament has its own probability. Depending on this ratio, we can say if we need two probabilities to explain the data or just one.
The following two plots compare the Runner and Corp win rates across the two tournaments. The dashed diagonal line shows the line of equal probabilities. Below this line, the ID performed better at Europeans/Africans, above the line, better at Americans. The colour represents our confidence in whether the probability changed. If the point is dark blue, then we have no reason to believe the win rate changed over time. These are usually close to the diagonal line or we don’t have enough data. Dark red, often away from the diagonal, have good reason to suspect a change. Colours in-between red and blue are less certain, for example, green shows middling evidence of a change, yellow slightly more evidence, orange more still until we get to red.
On the Runner side we have strong evidence that Leela was worse in the Americans Continentals. Note that while Sunny is further away from the diagonal we don’t have enough evidence to explain away what I’m dubbing the “Cpt Nice” effect. Of the more frequent IDs, Hayley had some noticeable improvement in Americans.
On the Corp side, we see Replication Perfection having a sizeable improvement between the two events. I found this interesting as it appears asset spam improved with Asa and CtM having some evidence of improvement. Perhaps with faster decks, like Titan, being targeted, smart meta players spotting the opening there and took advantage. MirrorMorph I cannot explain.
Anyway, I’ll be very interested to see how all of this is rendered irrelevant in the event starting tomorrow.
Find The Truth 