Drafting in Forge

[Chris H.]

p.s.
I've been offline for around 15 days, I think its a new world record for me, ha.

`
It is nice to see you back. We tend to miss you when you are gone for any length of time.

[mtgrares]

Wise man says, "I'm not dead, just slow moving."

[juzamjedi]

How does the AI pick cards? One of the only things that I know. Each of the 7 AI players pick a two color combination. First they pick 16 creatures in those colors then they choose spells. It sounds simple but getting the AI to actually pick cards and create a deck on the fly is pretty hard.

OK. Would it be possible to have archetype pick orders instead of color pick orders? I think it would help the Forge AI player a lot.

For some color combos there would be little difference. Green and Red will generally be "good stuff" aggro colors and I could come up with a pick order and they will mostly work fine. Blue will usually be control and the emphasis is definitely spells over creatures. However, Black and White are a lot more deck-dependent. Black + Blue is usually control while Black + Red is usually aggro. White + Green is *super* aggro, White + Blue is *super* control.

[mtgrares]

Having archetype pick (AP) orders is probably better than just having color pick orders. It seems like the AI might be able to use an AP list but I'm not 100% sure. The stuff that I, or the programmer, needs to know is:

--How many archetypes (and what are there colors) are there for your draft files?
--What are the ideal number of creatures and spells for each archetype?

I have no idea would just having one pick order per archetype would be good enough or should there be at least 2 files per archetype: one with creatures and the other one non-creatures.

p.s.
Archetype is a weird word that doesn't come up in "normal" conversation.

[mtgrares]

I don't expect you to generate 5 archetype pick lists right now (because I know it is a ton of work), maybe just 2 or three would do.

[juzamjedi]

Just now checked this thread again and hopefully you don't mind an Excel spreadsheet it's just easier for me that way. For starters I can keep it simplest for the AI and go with the allied-color pairs and see how that works out. I can also give some guidance on number of creatures, spells, and lands for the archetypes too. Naturally this will take a little time to rank the cards so I'll begin that tonight after work.

It seems that the lurkers are mostly interested in the "best limited" cube with some number of broken vintage cards included. I will incorporate this into the next cube list because I needed to cut some of the rares anyway.

[juzamjedi]

OK I made some more changes to the cube. The enemy-color cards like Maelstrom Pulse are gone including the enemy dual lands and enemy fetchlands. I also changed a number of black creatures and removed some of the cards that weren't playing as well as hoped.

I did not get around to listing pick orders yet since most of the night was spent tweaking the cube list itself. Just to emphasize my point though let me share with you a deck that I drafted tonight.

Code: Select all

absurd_draft.deck
[general]
draft
[main]
4 Island
3 Plains
3 Swamp
1 Underground Sea
1 Tundra
1 Dimir Aqueduct
1 Azorius Chancery
1 Mox Sapphire
1 Mox Pearl
1 Mox Jet
2 Ancestral Recall
1 Vendilion Clique
1 Trinket Mage
1 Mystical Tutor
3 Mulldrifter
1 Impulse
1 Repulse
1 Cloud Djinn
1 Divination
1 Shadowmage Infiltrator
1 Enlightened Tutor
1 Swords to Plowshares
1 Path to Exile
1 Day of Judgment
1 Demonic Tutor
1 Phyrexian Rager
1 Diabolic Edict
1 Smother
1 Hypnotic Specter
1 Sphinx of the Steel Wind
[sideboard]
1 Dark Ritual
1 Disenchant
1 Deft Duelist
1 Pacifism
1 Oblivion Ring
2 Hymn to Tourach
1 Moment's Peace
1 Coiled Tinviper
2 Tendrils of Corruption
3 Duress
1 Darksteel Colossus
40 Mountain
37 Swamp
36 Island
40 Forest
37 Plains

[mtgrares]

The pick orders don't have to include every card, I don't want you to kill yourself. Maybe the pick order would include the top 10-20 cards for an archetype. (I don't really know about the specific numbers, they are just a guess.)

[juzamjedi]

Really it's not a problem. There are over 300 cards in the cube and just assuming 5 archetypes that means at least 60 cards need to be ranked per color combination. Besides it's a good exercise to do anyway. So far I have seen a couple of cards that I wasn't happy with in any of the archetypes so those got cut.

[mtgrares]

Well I'm glad that it isn't tedious. I would like to make the AI better by being able "understand" what it is drafting, like weenie or something but the AI is very basic, it can barely walk.

[juzamjedi]

I have updated the cube. There are a lot of cards that got cut and I added a few to fill in some holes (mostly I added a few creatures to fill some missing parts of the mana curve). I removed most of the "broken" cards although there are still a few like Skullclamp. I haven't updated pick orders yet.

Try it! I am liking the direction this "good for limited" draft set is moving.

[mtgrares]

I haven't updated pick orders yet.

No problem, I'm in no rust, I mean rush.

[juzamjedi]

I have made some changes to the cube. Download the attachment. First of all there are a lot of new cards and some of them were too good to leave out. Slivers are gone. Multi color cards are also gone. They were sweet, but the AI was clueless and kept putting Blightning in a R/G deck, or Qasali Pridemage in a W/U deck, etc.

Overall you will notice the red cards in this cube are MUCH more powerful than they were previously. This is mostly due to the number of awesome new red cards that have been coded since my last cube update. I am disappointed with how White plays right now. Blue was already excellent, Black and Green are fine, Red got a lot better this time around but White is sort of "meh".

I am pretty unhappy with White's "Soldiers" theme in particular. The controlling cards (Wrath of God, Path to Exile, etc.) are good but White Weenie feels underwhelming. Suggestions?