The ‘Perfect’ Deck

I’ve spent quite a bit of time in the past working out which decks are stronger than others in this meta. I also had the character lineups from World’s for a precise strength comparison. However, I’ve never in the past attempted to write a programme to automatically build me a deck.

I recently challenged myself to write a programme to automatically create character lineups for Destiny. This worked, but it didn’t tell me anything about what cards would go in the decks.

I recently got access to the complete deck lists from the Las Vegas Open (120 decks). My thanks to Nicholas Nelson for providing this in a nicely formatted spreadsheet for the world to see, because it meant I could use it to analyse the exact card choices used in each decklist, and create the ‘perfect’ deck for each archetype using the results.

I’ve analysed deck lists for lineups with at least 3 decks in the tournament. This gave me the following decks (with number of each deck in brackets):

ReyLo (16), ePalp-eWatto (9), eVader-Greedo (7), Chopper Droids (5), eLeia-eYoda (5), eIden-eBeckett (4), eLor-eBail-Ani (4), eSnoke-MT-Mando (4), eMaul-Mando (3), eObi-eR2 (3), Ewoks (3).

I figured that a successful deck had good card choices, so I gave each deck a weight based on performance of that deck in the tournament. I then gave each card a score based on how often you see it combined with the deck weights, giving me a card ranking. Finally, I looked at the number of copies of each card in decks, and worked out the optimum number based again on deck weight.

Below is a graph showing the results for ReyLo. The higher the bar, the higher the score. Blue are cards which are two-ofs, Green are singletons:

This shows that there were 48 distinct cards in all the ReyLo decks. Considering you can have 2 copies of each card but only have 30 cards in your deck, this means there was actually quite a lot of variety between ReyLo decks.

I used these results to build a deck, starting with the best card (and taking 1 or 2 depending on the optimum number calculated) and working my way down the list in one’s and two’s until I reached 30 cards. This method of course is not perfect. It ignores playstyle and synergies. It also risks missing cards which are functionally interchangeable. But it gives me a good idea of what cards I should be looking at and I can also look at the next 10 cards to give me some flex options. I saved my results here if you want to look.

Shortcomings of this method are noticeable for the ReyLo deck, in which the ‘Perfect’ deck has two Luke’s Training, but only two abilities. This is because the only ability ReyLo decks could agree on was Niman Mastery. I’ve therefore dropped the Luke’s Training and pulled in a the next two cards from the Top 40: R2-D2 and Power From Pain.

Anyway, so this is the ‘perfect’ ReyLo deck: