Casual Affinity Designed by Artificial Intelligence (Casual)

This deck was designed entirely with an automated algorithm. The source code is open source and you can look at it, improve it, use it at:

https://github.com/Darelbi/MTG-Mana-Curve-Simulation

The algorithm runs a simulation of the deck for 50.000 times using the real rules of the game, then it swaps a card an re-run the simulation and check which of the two decks is better. And use that deck as basis for another swap, as long as an improvement is found.

By running the simulation millions of times on the above cards the optimal deck was found. Originally the deck was winning on average at turn 5,07. But after I made a improvement in the mulligan strategy I came with this second deck that is capable of an astonishing average turn of victory of 4,90. Over 3.5% of improvement just by rethinking the mulligan.

If you want to perform great with this deck you have to use the same mulligan strategy:

Do mulligan down to 5 cards, if at least one of the following conditions is not met:

• Steel overseer, plus 2 creatures, plus 2 mana
• Master of etherium, plus 1 creature, plus 3 mana (1 blue)
• Cranial plating, plus 2 creatures, plus 2 mana
• Urza's Saga, plus 1 creature, plus another 2 mana

The simulation achieves AWT (average winning turn) of 4,90 with that kind of mulligan.

I tried to force in many ways to reduce number of lands and the results were always worse, this deck lives high with high number of lands. Math does not lie, of course it rely on the correct mulligan, if you mulligan bad you may get a penalty of over 20% on AWT. Mulligan avoids you mana flood. The first 10 cards you draw are the cards that leads you to victory.

SIDEBOARD:
The sideboard is a compromise, each card of this deck you give up, makes it perform worse (higher AWT). But anyway we need side cards, so I runned the simulation to find the cards that (in order) gives the least increase to AWT when removed, here are the cards (follow strictly this order):

• Glimmervoid (AWT: 4,948)
• Thought Monitor (AWT: 4,984)
• Island (AWR: 5,030)
• Steel overseer (AWT: 5,076)
• Glimmervoid (AWT: 5,134)
• Steel overseer (AWT: 5,186)
• Frogmite (AWT: 5,256)
• Glimmervoid (AWT: 5,324)
• Ornithopter (AWT: 5,395)
• Frogmite (AWT: 5,497)

I limited to 10 cards. As you see the further you go on, the worse is the worsening. The first card add 0,04 AWT, the last card add 0,1 AWT, almost 3 times more.

I do take commissions on custom decks, to find out optimal mana curve and optimal AWT. That won't be cheap since every card requires custom code. But I do accept cards as payment. The above simulation would be around 800€.

Before requesting a commission, try this deck in a real game to see how good it performs (Don't forget the mulligan).

NO COMMANDER decks, you just cant afford to make custom code for 100 different cards.