VA Regionals predictions!

He's going owl the way?
It's that time again... Virginia (Blue Ridge or Roanoke or whatever you wanna call it) Regionals are tomorrow and I've been sitting here doing some math for the better part of the afternoon trying to figure out what will do well.

I posted recently about my model's failure to predict the downfall of Quad Lapras in Brazil and how I planned some tweaks to my formulas and methods to prevent this kind of stuff from happening again. I have made those changes, but the predictions for Roanoke should still be viewed as a test of the formula, and as such, this post should probably not be the deciding factor in what you play tomorrow, if any VA competitors are reading this. :)

(In the future, I think the Rankings will be accurate enough to help with choosing the best deck to play... just not this weekend.)

So anyway. Here's some boring math, before I jump into the Top 10 Meta Share predictions!

A Note on the Math
After inputting data into my model for Utah Regionals and Brazil Intercontinentals, I noticed something very strange: an outlier that caused the sides of my quadratic to curve upwards. I checked and the outlier turned out to be Dark Dragons, which saw a decrease from 6.39% of the Utah Day 2 meta to 0% of the Brazil Day 2 meta. What the outlier was saying was that Dark Dragons should have been a spectacular play for Brazil - which I obviously questioned, since not a single one made Day 2, and I don't think it would have been so overlooked had it really been a top competitor.

The formula with this outlier omitted jumped from y = 0.0574x^2 - 1.3354x + 8.8976 to y = 0.12x^2 - 2.09x + 10.51, a significant change. Likewise, R-squared (a nerdy stats thing that tells how accurately the model covers the data) changed from 0.2854 to 0.2671, significant given the amount of data in my model. I figured it was best to remove the outlier, much to the collective relief of everyone who didn't play Dark Dragons in Brazil, I assume.

My new predictions honestly haven't changed much from Brazil, and I doubt that's surprising. PRC-SM is a pretty stale format right now and I'm very excited for Guardians Rising to make some serious changes to the metagame.

Virginia Meta Predictions
Here are the Top 10 predicted decks for this weekend, per PokeStats's Power Rankings:
  1. DeciPlume
  2. M Rayquaza/Fire (possibly with Gumshoos-GX)
  3. Turbo Darkrai-EX
  4. DeciPlume Rainbow (Jolteon-EX/Regice)
  5. M Mewtwo
  6. Quad Lapras
  7. Volcanion
  8. Vespiquen/Zoroark
  9. Gyarados
  10. Yveltal-EX/Garbodor
Umbreon-GX/Wobbuffet made a big jump in the Rankings but is still outside the Top 10. Two new additions to the list are Zygarde-EX/Carbink and Yveltal-EX/Zoroark, although neither is predicted to get much more than a single Top 32 placement, if anything. 

We'll see how this goes...

No comments:

Post a Comment