All posts

August 8th 2026

Predicting the Pokémon TCG World Championships

Mathematical Modelling Java Data Analysis

Last updated September 19th, 2026 · what changed

With the 2026 Pokémon TCG World Championships coming up later this month, I decided it would be interesting to take a look at the top decks and which ones I think will have the greatest chance at success. Unfortunately I was even close to qualifying for Worlds this year to what makes me qualified to talk about this? Honestly other than me knowing some math and programming not much, just doing this for the vibes.

The decks in this prediction

At the time of writing this, only 1 major source, Tag Team Podcast on X, has made a prediction on the best decks and their respective metashares. Since their team consists of some of the best Pokémon players in the world I decided to take their prediction at face value.

Tier list grading Pokémon TCG decks from S to D. S: Dragapult, Dragapult Dusknoir, Dragapult Blaziken. A: Slowking, N's Zoroark. B: Festival Lead, Alakazam Dudunsparce, Basic Box, Crustle. C: Grimmsnarl Froslass, Mega Excadrill, Lucario Hariyama, Ogerpon Meganium Hydrapple. D: Cynthia's Garchomp, Dhelmise. Each deck is labelled with a number.
Deck tier list by @TagTeamPokemon on X.

How my prediction will be made

I spent the past few days creating a tournament simulator that will fill a tournament based on a given metagame and play out a full 3 phase tournament. The full code for which can be found here. A key assumption I had to make was that all players would be of equivalent skill level. This is obviously not true but I had no other way around this. Another thing not accounted for was ID's and resistance. Honestly I was just too lazy to implement them and do not believe it is going to be drastically important. The way it decides between a given matchup is a weighed guess between the 2 decks based on the matchup data from Trainer Hill. I decided to download and use matchup data from June 1st 2026 to August 8th 2026 with tournaments with at least 100 players. This is because I wanted to include at least 1 major tournament in the matchup data. It is important to note that Worlds will not be played in the same format as NAIC, however I do believe none of the decks have changed significantly enough with the release of Pitch Black, with the exception of Grimmsnarl Froslass, that excluding any games before then would be beneficial. Furthermore given that Grimmsnarl Froslass is already predicted to be an unpopular deck I figured it would not make much difference and any improvement would only be marginal. The tie rate was pretty much taken directly from NAIC and the player count was sum of the invitations posted from the championship website.

The only thing left to do now was run the simulation 10000 times.

10000 tournaments of 425 players, 13.5% tie rate
Deck Field Phase 2 Top 8 Wins
dragapult-ex19.82%24.78%27.73%29.39%
crustle-dri5.95%7.95%10.30%12.68%
dragapult-blaziken9.90%9.84%9.10%8.00%
basic-box-m4.94%6.51%7.48%8.20%
slowking-scr4.97%5.95%6.97%8.27%
dragapult-dusknoir9.88%8.44%6.86%4.99%
n-zoroark7.93%7.26%6.43%5.82%
festival-lead6.93%6.76%6.37%5.81%
mega-excadrill-ex4.95%5.57%6.24%7.20%
alakazam-dudunsparce4.96%4.57%3.81%3.28%
grimmsnarl-froslass2.97%2.83%2.46%2.10%
ogerpon-meganium-hydrapple1.97%1.71%1.45%1.20%
cynthia-garchomp-ex1.98%1.76%1.40%1.05%
dhelmise-pbl1.97%1.60%1.25%0.87%
other7.91%2.86%1.24%0.57%
lucario-hariyama2.97%1.61%0.92%0.57%
Players Counted4,250,000682,69880,00010,000

Phase 2 and Top 8 Charts

Bar chart of Phase 2 share by deck: Dragapult ex 24.78%, Dragapult Blaziken 9.84%, Dragapult Dusknoir 8.44%, Crustle 7.95%, N's Zoroark 7.26%, Festival Lead 6.76%, Basic/Slop Box 6.51%, Slowking 5.95%, Mega Excadrill ex 5.57%.
Phase 2 share by deck.
Bar chart of Top 8 share by deck: Dragapult ex 27.73%, Crustle 10.3%, Dragapult Blaziken 9.1%, Basic Box 7.48%, Slowking 6.97%, Dragapult Dusknoir 6.86%, N's Zoroark 6.43%, Festival Lead 6.37%, Mega Excadrill ex 6.24%.
Top 8 share by deck.

Final prediction

No deck won more than 50% and we see that Dragapult and friends are a far favourite to win the whole thing, with straight Dragapult leading them, but still only taking 42.38% of the winning decks. Even though Dragapult and friends are the most likely to win it is still more likely it will lose, however other decks are even less likely to win.

Donut chart of predicted tournament win share by deck: Dragapult ex 29.4%, Crustle 12.7%, Slowking 8.3%, Basic/Slop Box 8.2%, Mega Excadrill ex 7.2%, N's Zoroark 5.8%, Festival Lead 5.8%, Dragapult Dusknoir 5%, Alakazam Dudunsparce 3.3%, and smaller shares for the remaining decks.
Predicted share of tournament wins by deck.

So what actually happened?

Written September 19th, 2026

Writing this a little late but I'm here now.

Metashare

So turns out theres this cool thing that calculates "how wrong" a prediction is called the total variation distance where the lower it is the closer the prediction was to being correct.

\[ \mathrm{TVD} \;=\; \tfrac{1}{2}\sum_i \bigl| p_i - q_i \bigr| \]

where \(p_i\) is the predicted share of deck \(i\) and \(q_i\) its actual share. For the metashare prediction it resulted in 16.06% meaning about one in six players brought a deck that was not expected. This result seems okay but I am by no means blamining the Tag Team Podcast since I'm quite confident I would have been much worse.

Interestingly enough the 3rd place finisher, Brent Tonisson, made a pre-worlds prediction which I have below (when I was copying it I forgot about Lucario Hariyama which should be 1). His prediction has a TVD of 17.36% which is slightly worse than the one I used. When he published this I did want to run another simulation but there was not enough tournament data to help support his prediction due to how many specific archetypes were mentioned.

Alternative pre-tournament tier list with a metashare weight on each deck. S: Dragapult 16, Dragapult Blaziken 14, Dragapult Dusknoir 10. A: N's Zoroark 7, Raging Bolt Ogerpon 3, Slowking 6, Festival Lead 7, Mega Excadrill 7. B: Basic Box 4, Alakazam Dudunsparce 4, Crustle 2, Ogerpon Meganium Hydrapple 3, Grimmsnarl Froslass 3. C: Dhelmise 2, Lopunny Dudunsparce 1, Lillie's Clefairy 1, Mega Lucario 1, Beedrill 1. D: Lucario Hariyama (left blank), Toucannon 1, Rocket's Honchkrow 1, Rocket's Mewtwo 1, Mega Greninja 1, Cynthia's Garchomp 1.
Pre-worlds tierlist by Brent Tonisson.

Going back to the original data here is a breakdown of the TVD:

Full comparison table — 18 decks
Predicted metashare against the actual shares from Limitless. Rounding each row to two decimals leaves the predicted column at 100.01% rather than 100%; the actual column is exact across all 797 players.
Deck Predicted Actual Difference
dragapult-ex19.80%22.33%+2.53
crustle-dri5.95%2.51%−3.44
dragapult-blaziken9.91%9.79%−0.12
festival-lead6.93%2.89%−4.04
slowking-scr4.94%5.90%+0.96
basic-box-m14.95%10.66%+5.71
mega-excadrill-ex4.95%4.02%−0.93
dragapult-dusknoir9.90%10.41%+0.51
n-zoroark7.93%7.53%−0.40
alakazam-dudunsparce4.96%6.78%+1.82
grimmsnarl-froslass2.96%0.88%−2.08
cynthia-garchomp-ex1.98%0.50%−1.48
ogerpon-meganium-hydrapple21.98%2.26%+0.28
dhelmise-pbl1.98%1.00%−0.98
raging-bolt-ogerpon30.00%3.14%+3.14
lillies-clefairy40.00%0.50%+0.50
other7.93%8.53%+0.60
lucario-hariyama2.96%0.38%−2.58
  1. Combines Limitless’s Basic Box and Clefairy Ogerpon, which the prediction treated as one archetype.
  2. Combines Limitless’s Ogerpon Meganium Hydrapple and Ogerpon Meganium Arboliva.
  3. Not accounted for in the original prediction, so its predicted share is 0%.
  4. Also not accounted for in the original prediction, so its predicted share is 0%.

Phase 2

Full comparison table — 18 decks
Predicted Phase 2 shares against the actual Day 2 shares from Limitless. Rounding leaves the predicted column at 100.02% rather than 100%; the actual column is exact across all 143 day 2 players. Same grouping rules as table above.
Deck Predicted Actual Difference
dragapult-ex23.94%15.38%−8.56
crustle-dri7.93%5.59%−2.34
dragapult-blaziken9.72%7.69%−2.03
festival-lead7.84%3.50%−4.34
slowking-scr6.14%7.69%+1.55
basic-box-m6.50%14.69%+8.19
mega-excadrill-ex5.54%5.59%+0.05
dragapult-dusknoir8.43%5.59%−2.84
n-zoroark7.16%8.39%+1.23
alakazam-dudunsparce4.39%9.79%+5.40
grimmsnarl-froslass2.88%0.00%−2.88
cynthia-garchomp-ex1.80%0.00%−1.80
ogerpon-meganium-hydrapple1.64%0.70%−0.94
dhelmise-pbl1.58%1.40%−0.18
raging-bolt-ogerpon0.00%2.10%+2.10
lillies-clefairy0.00%0.70%+0.70
other2.89%9.79%+6.90
lucario-hariyama1.64%1.40%−0.24

This went significantly worse at 26.13%. This could have been due to a variety of factors such as errors propogating or innacurate matchup data due to the new format.

Top 8 and Winner

Broadcast graphic of the Worlds 2026 Top 8 decks: Brent Tonisson, Dragapult ex; Henry Chao, Dragapult ex / Dusknoir; Andrew Hedrick, Dragapult ex / Dudunsparce; Diego Cassiraga, Alakazam / Dudunsparce; Rune Heiremans, Dragapult ex; Satoshi Matsui, Alakazam / Dusknoir; Mateusz Łaszkiewicz, listed as Mega Kangaskhan ex Tool Box; Nathan Spry, Teal Mask Ogerpon ex / Lillie's Clefairy ex.
The Top 8. Official Pokémon World Championships 2026 broadcast — note that there are a few broadcast mistakes most notably it lists Mateusz Łaszkiewicz as playing Mega Kangaskhan ex Tool Box. He actually played Alakazam Dusknoir.
The Top 8 against the model expected, counted in players.
Deck Actual Expected Difference
dragapult-ex22.09−0.09
crustle-dri20.84+1.16
basic-box-m10.61+0.39
dragapult-dusknoir10.50+0.50
alakazam-dudunsparce10.29+0.71
other10.08+0.92

Overall the model was pretty much spot on for Dragapults in top 8. Although it did not predict both Crustles in top 8, Crustle was a deck that was largely written off by the community and it was not belived to have been able to make it to top 8 so still cool to see the model give the deck a lot more credit than the community. As for Alakazam Dusknoir (other) this was completely unexpected but would have been hard to predict since revolutionary new decks often come from testing groups offline which would not be included in any online data. Former world champion, Diego Cassiraga, managed to pilot Alakazam Dudunsparce to finals which was an unexpected deck for both the model and the community. As for the other 2 decks in top 8, since there can't be "0.5" a player anything with more than 6.25% would not have been terribly unexpected.

I did make the prediction that Dragapult in all its forms would lose since it did not posses more than a 50% likelyhood of winning but this was unfortunatey proven wrong with Andrew Hedrick getting a decisive 2-0 victory in finals.

Improvements for the future

I still do believe a Monte Carlo simulation is still the best way to make unbiased predictions about tournament results but will probably still do worse than intuition from top players. To help with inaccurate matchup data Rielly Cooper created a system on his Metagross-EV tool which I won't get into but requires a minimum amount of games played between decks and will fill in missing games to create a more accurate win rate.

Updates & corrections

August 11th, 2026

Festival Lead + Grimmsnarl Matchup Data

It was pointed out to me that Festival Lead gained Gladion's Final Battle in Pitch Black which was a major deck improvement that I had overlooked. To fix this I decided to manually update the matchup data for Festival Lead only using tournaments after July 16th. I decided to do the same for Grimmsnarl. I also did not account for the full amount of competitors since I did not include Asia Pacific however this was easily fixed by increasing the amount of players. Below are the updated graphs and tables.

Updated results — 10000 tournaments of 750 players, 13.5% tie rate
Deck Field Phase 2 Top 8 Wins
dragapult-ex19.80%23.94%26.18%26.39%
crustle-dri5.95%7.93%10.49%12.53%
dragapult-blaziken9.91%9.72%8.67%7.63%
festival-lead6.93%7.84%8.49%8.97%
slowking-scr4.94%6.14%8.06%9.38%
basic-box-m4.95%6.50%7.60%8.03%
mega-excadrill-ex4.95%5.54%6.37%7.40%
dragapult-dusknoir9.90%8.43%6.28%5.26%
n-zoroark7.93%7.16%6.02%5.43%
alakazam-dudunsparce4.96%4.39%3.59%3.04%
grimmsnarl-froslass2.96%2.88%2.57%2.13%
cynthia-garchomp-ex1.98%1.80%1.40%1.25%
ogerpon-meganium-hydrapple1.98%1.64%1.20%0.90%
dhelmise-pbl1.98%1.58%1.16%0.73%
other7.93%2.89%1.01%0.42%
lucario-hariyama2.96%1.64%0.91%0.51%
Players Counted7,500,0001,201,64980,00010,000
Updated bar chart of Phase 2 share by deck: dragapult-ex 23.94%, dragapult-blaziken 9.72%, dragapult-dusknoir 8.43%, crustle-dri 7.93%, festival-lead 7.84%, n-zoroark 7.16%, basic-box-m 6.50%, slowking-scr 6.14%, mega-excadrill-ex 5.54%.
Updated Phase 2 share by deck.
Updated bar chart of Top 8 share by deck: dragapult-ex 26.18%, crustle-dri 10.49%, dragapult-blaziken 8.67%, festival-lead 8.49%, slowking-scr 8.06%, basic-box-m 7.60%, mega-excadrill-ex 6.37%, dragapult-dusknoir 6.28%, n-zoroark 6.02%.
Updated Top 8 share by deck.
Updated pie chart of win share across all sixteen decks: dragapult-ex 26.39%, crustle-dri 12.53%, slowking-scr 9.38%, festival-lead 8.97%, basic-box-m 8.03%, dragapult-blaziken 7.63%, mega-excadrill-ex 7.40%, n-zoroark 5.43%, dragapult-dusknoir 5.26%, and the remaining seven decks at 3.04% or below.
Updated win share

Although Dragapult and friends still massively lead, Festival Lead is now the 4th most likely deck to win as a result of its better matchup into Dragapult with the addition of Gladion's Final Battle. Furthermore this also pushed N's Zoroark lower down likely due to its grass weakness. Interestingly Dragapult Blaziken was able to move up, just above Festival Lead, however I am unsure why at the moment.