World Deck Development thread

[Xyx]

Here it is. The usual "it's still under construction" caveats apply.

You'll need to add this line to worlds.txt:

Code: Select all

Name:Block 30: Theros|Dir:30 Theros|Sets:THS, BNG, JOU

Why "Block 30"? Well, unless I've miscounted, Theros is the 30th set that'll make for a unique quest world. I've also been working on worlds 00 (Beta) through 07 (Ice Age), and I thought I'd leave some room for the future.

Feedback welcome!

[Xyx]

be sure to only include decks that the AI can play well. Netdecks and precons will need at least a few tweaks.

OK, I'll try, but I may need you to be a bit more specific.

1. Is it not sufficient to downgrade the difficulty to reflect the AI's performance?
2. "Well" as in "like a pro"? Or just "not completely boneheaded"?
3. What extent of tweaks are we looking at? No cards marked "AI"? No cards marked "?" (what's the difference anyway?) Some of those aren't completely disastrous.

[Joril]

It's probably easiest if I just show you what I've been doing. I built six worlds, one each for Beta, Arabian Nights, Antiquities, Legends, The Dark and Fallen Empires.
...
I plan to work on either Ice Age (with Alliances and I guess Cold Snap) or Homelands next. Ice Age is first in line but Homelands is really interesting; full of flavor and bad cards.

Thoughts? Anyone care to try a quest?

I absolutely LOVE this idea I began playing Magic with the Third edition and stopped with the Seventh, so I'm a bit uncomfortable with all the mechanics that were introduced afterwards... Shandalar is a lot of fun but having other quest worlds from "that time" would be awesome

[Xyx]

Glad to hear it! If you're interested, here's my world folder.

The Ice Age world is still very rudimentary but the Homelands world is pretty much done. Quirky little world. If you're looking for an extra challenge, try to build a deck without 4-drops.

I'll be getting back to Ice Age after I finish Theros.

[Marek14]

Xyx: Is that on purpose that ARN The Queltosh has deck full of Desert Nomads, but no actual red mana sources to ever cast them?

Also, you seem to have a typo in title of LEG UB 3.dck -- should be "Sivitri Scarzam", not "Sivitry"
DRK Big Band 3.dck uses Name and Title from Firewater 3.
DRK Scarwood 2 should be probably named DRK Scarwood 3.
DRK Wormwood 2 should be probably named DRK Wormwood 3.
THS Mechanic: Herioc: Black - nuff said.
The Heroes vs. Monsters decks in Theros might use the actual duel deck pictures instead of regular.

[Sloth]

be sure to only include decks that the AI can play well. Netdecks and precons will need at least a few tweaks.

OK, I'll try, but I may need you to be a bit more specific.

1. Is it not sufficient to downgrade the difficulty to reflect the AI's performance?
2. "Well" as in "like a pro"? Or just "not completely boneheaded"?
3. What extent of tweaks are we looking at? No cards marked "AI"? No cards marked "?" (what's the difference anyway?) Some of those aren't completely disastrous.

Not completely boneheaded is the goal, but that's hard enough sometimes.

1. Cards marked as "AI" have to go, unless the AI can't do anything wrong with them (like a tutor with only one target).
2. The AI will ignore certain synergies and some cards can get wasted in combination with other cards (like casting Terror on a creature that has already been enchanted with Guard Duty). Testing will show cases like these.
3. A lot of stellar cards in pro decks are trash in the hand of the AI (albeit not useless), like Brainstorm or Thoughtseize.

[Xyx]

Xyx: Is that on purpose that ARN The Queltosh has deck full of Desert Nomads, but no actual red mana sources to ever cast them?

Totally.

you seem to have a typo in title of LEG UB 3.dck -- should be "Sivitri Scarzam", not "Sivitry"
DRK Big Band 3.dck uses Name and Title from Firewater 3.
DRK Scarwood 2 should be probably named DRK Scarwood 3.
DRK Wormwood 2 should be probably named DRK Wormwood 3.
THS Mechanic: Herioc: Black - nuff said.

All fixed. Thanks! Very helpful.

The Heroes vs. Monsters decks in Theros might use the actual duel deck pictures instead of regular.

Hmm... I can't find them. In fact, I can't find any Duel Decks pics in the pics\precons folder. Am I looking the wrong spot?

Regardless... I don't think I want to include Heroes vs. Monsters. They only contain three Theros cards each. Not exactly representative.

Not completely boneheaded is the goal, but that's hard enough sometimes.

Will try!

[Marek14]

The Heroes vs. Monsters decks in Theros might use the actual duel deck pictures instead of regular.

Hmm... I can't find them. In fact, I can't find any Duel Decks pics in the pics\precons folder. Am I looking the wrong spot?

They should be in pics/cards, among other sets.
The Sorin vs. Tibalt decks in normal quests might be also changed...

Also, I wonder about the theme decks. I have this nasty suspicion they were created based on one weird idea I posted here many years ago, but I wonder when they were last updated.

[Xyx]

They should be in pics/cards, among other sets.

I've searched the following folders for "koth" and "tibalt" and whatnot and couldn't find a hint of Duel Deck pics.:

• Forge\res\
• AppData\Local\Forge\
• AppData\Roaming\Forge\

Could you point me to a specific file?

[Marek14]

Well, I have them in Local\Forge\Cache\pics\DDK. I'm not sure whether these pictures can be downloaded, though, I simply use the HQ releases.

[Xyx]

If I'm not mistaken, DDK is short for Duel Deck K, which is Sorin vs. Tibalt. My folder contains 80 card images and nothing else. No deck images.

[friarsol]

Also, I wonder about the theme decks. I have this nasty suspicion they were created based on one weird idea I posted here many years ago, but I wonder when they were last updated.

I spearheaded the theme deck thing as a way to create semi-unique decks with a theme that wouldn't play exactly the same each time. We had a run on them where hundreds were created in that first month or two, but I'm not sure many of them have been updated (or new ones created) since then. And that had to be like 5-6000 cards ago.

[Marek14]

Also, I wonder about the theme decks. I have this nasty suspicion they were created based on one weird idea I posted here many years ago, but I wonder when they were last updated.

I spearheaded the theme deck thing as a way to create semi-unique decks with a theme that wouldn't play exactly the same each time. We had a run on them where hundreds were created in that first month or two, but I'm not sure many of them have been updated (or new ones created) since then. And that had to be like 5-6000 cards ago.

I see...

[chops]

Someone ran an AI test in which all the Ravnica decks played against eachother, it would be very nice if such a test is still possible so I can better set the power levels of the decks.

Sorry this took so long. I took a break from Python for a bit. I've no automated the process now, so I can run these for anyone else who wants one. Here it is:
30.6 Slayer 3
28.8 RiGa 3
28.0 Kamsa 2
28.0 Nissa 2
27.8 Ulamog 2
27.7 Tuktuk 3
27.7 Emrakul 4
27.7 LSV 3
27.6 Shrike 3
27.0 Marego 3
26.9 Allies 2
26.6 NDE 1
25.9 Drana 2
25.5 JC 3
25.2 LSV 2
24.3 Borg 1
24.0 Miffy 1
23.8 Lemmings 2
23.2 Robocop 1
22.9 Kozilek 1
22.9 Wallsrus 2
22.5 Lambik 1
18.3 Kai Buddha 2

The numbers to the left of each deck title are TrueSkill's best guess of the deck's "level" on a 50-point scale, with a margin of error of ±1. That means that the system is 95% certain that the numbers are within 2 levels of the correct level. That's still a fair bit of uncertainty, but simulating more games should help.

These numbers come from running every combination of match-ups in for nine AI-vs-AI games. I put the result of each game through TrueSkill in a random order. I'll run another batch of 9 games each tomorrow and see how much that lowers the margin of error.

P.S.: I removed "Axis 1" from the decks because it tended to crash. Attached is a sample log from a mirror match. Maybe somebody can figure out why it crashes.

[Xyx]

I've no automated the process now, so I can run these for anyone else who wants one.

If you could, I would very much like to know the relative AI strengths of the PTJOU Block Constructed decks in their meta. They're available for download in this topic. Once I know which decks the AI plays best, I can eliminate the redundancy by weeding out the weaker versions. That would give me a good and varied selection of decks for the Theros quest world that I'm working on. It would also help me gauge the difficulty levels of the other decks.

Is this a lot of work? Is there anything I could do to assist, such as preparing lists or something?