The Ultimate Memphis Regionals Predictions Post

We are coming up on the largest Regional the Pokemon Trading Card Game has ever seen. If the actual attendance matches the registration, over 1000 Masters-division players will descend on Memphis, Tennesee, and only 32 of them - just over three percent - will advance to the second day of competition.

This Regional is going to be insane, and, realizing that, I've gone all-out with my preparation for it. From November 18th to December 10th, PokéStats collected its largest-ever sample of League Cup data: 56 Cups in total, with several thousand CP awarded in the top cuts of those Cups. I've also partnered with Some1sPC to provide a "weekly tier list" page -- our "Meta Snapshot" -- inspired by a concept from the Hearthstone community. Based on these two related but slightly-different efforts, I've decided that there are two tracks I can follow when predicting the Memphis metagame. These are:

1. What if the Memphis metagame is close to the metagame from our 4-week Cup post-EUIC Cup sample?

2. What if the Memphis metagame proceeds on trendlines from only the last week of our 4-week sample?
     - secondary question: what do these trendlines look like? Linear? Polynomial? Exponential?


To elaborate a little further on these: the first option is to assume that the Memphis metagame will not change much from the metagame of PokéStats's 56-Cup sample. There is merit to this assumption. No new sets or promo cards have been released since the EUIC, and the decks seeing success at Cups have been almost entirely the same decks that did well at the EUIC.

However, the second option involves looking at the sample not as one "block" of results, but as four individual weeks of data. This is best visualized with a line chart, which is what our Some1sPC Meta Snapshot does. With this model, we can examine the trends of each deck and try to extend their trendlines to see where they might end up for Memphis.

The second approach is riskier, but offers a more satisfying reward if I get it right. One of the first things you'll learn in any Statistics class is that you must be extremely careful when extrapolating from a line chart. (Anyone well-versed in Statistics, feel free to skip this next section.) Pokemon is constantly changing as the game's top players craft new strategies and bring unexpected decks to huge tournaments. Let's take a look at a Meta Progession chart from the 2016-17 season as an example. (Sorry it's so small! You might need to zoom in...)

To see what I mean about extrapolation being dangerous, take a look at the beginning of this Rainbow Road chart -- specifically the first four data points. The r-squared value, which can range from 0 to 1 and measures how well a linear model fits a set of points, is 0.9681 for those four events. This suggests that a linear model fits them almost perfectly. Yet if I extrapolated from that model when predicting the metagame of last season's EUIC, I would've predicted Rainbow Road to take 19.87% of the Day 2 metagame. My estimate would've been more than a five-fold increase over the actual Meta Share.


Keeping all this in mind, I've decided to use both methods and come up with two different sets of predictions. First, I'll look at the four-week Cup sample as a whole, and, with the assumption that the Memphis metagame will be similar to that of the entire sample, predict the most-played decks at Memphis. Then, I'll use the Some1sPC Meta Snapshot to predict how well each decks will do based on the trends observed over the sample.

I'm making a slight leap of faith here: I'm equivalating CP Meta Share (as a percentage from the Cup sample) with Placement Meta Share (measuring the number of Day 2 spots a deck will earn in Memphis). For example, ZoroPod took 16.79% of the CP in the Cup sample, so I'm then projecting it to take 16.79 of the Day 2 Placements (not CP) in Memphis. I think in the future I'll actually do some math on this to see whether my assumption is statistically backed, but we'll try it for tonight and see what happens.

Here are the decks I predict to make an appearance in Day 2 of Memphis.

Day 2 projected placements
Zoroark/Golisopod-GX - 5
Brokenvoir - 5
Buzzwole-GX/Lycanroc-GX - 4
Decidueye-GX/Zoroark - 3
Volcanion - 3
Gardevoir-GX/Sylveon-GX - 2
Greninja BREAK - 2
VikaBulu - 2
Drampa-GX/Espeon-GX/Garbodor - 1
Metagross-GX - 1
Silvally-GX/Metal - 1
Zoroark-GX/Lycanroc-GX - 1
other/rogues - 2


Next up, I'll use that Meta Snapshot and look at some trends that we've seen in the past couple weekends of Cups:
If we continue to follow all those lines, here's what we can expect for the performances of these decks relative to the metagame of the last four weekends (including EUIC):

Trending up: Zoroark/Golisopod-GX, Buzzwole-GX/Lycanroc-GX, Volcanion, Greninja BREAK, Zoroark/Lycanroc-GX, Metagross-GX, Decidueye-GX/Zoroark

Trending down: Gardevoir-GX (all variants), VikaBulu

It does look like Zoroark decks of all varieties are taking over Standard. Both Lycanroc-GX and Decidueye-GX are formidable partners for it. This is also reflected in the final part of this article, which is our expert panel's Deck Power Rankings.

PokéStats Deck Power Rankings - Memphis Regionals 2017

Best of luck to everyone competing in Memphis tomorrow, and for everyone watching from home, stay tuned to @PokeStats_TCG and this site for the usual coverage!

No comments:

Post a Comment