Right! I'm back!
Sorry I left off on a rant, I was driven somewhat mad and as was helpfully pointed out needed a break.
Well, not all of that is true, I'm not exactly back. Yet. Got some RL stuff coming up and going on, so I most likely won't be able to devote as much time to the whole thing as before. However, Korath did listen to plenty of feedback and this did open doors which were closed before. So I kinda owe it to him to put the stuff to good use.
Also in the meanwhile my dropbox folder got wiped and I'm not sure I have the very last iteration of the deckpack. Not to mention that it had glaring issues and untouched decks. I kinda left you guys in a bind and feel like I let you all down.
Part of that was because the environment that I was trying to do stuff in was nowhere near where it needed to be to even consider making decks in it. Fundamental stuff wouldn't work. When you have an AI which has a hard time operating underpriced vanilla creatrues, like
Leatherback Baloth in a mono-green deck, the problems go too deep. That's not even going into the then-current bugs, and the comes-into-play tapped lands handling. Many things of that sort going on - you get heaps of cards locked off as unusable in practice. I fiddled with new cards in the last couple of days and it was a bit disenheartening to see quite a few problems still there.
HOWEVER! Korath has fixed quite a number of bugs and made quite a few things work. Ha also added a metric ton of cards including a lot of cards which if I had at my disposal and if the AI was capable of using them the way god intended would've made life much less difficult. Or, at least, I could make the enemy decks more casual and less murderous, I think.
Or I just got options that I didn't have before. Fight cards in green like
Prey Upon have the potential to change how that color plays entirely, and when the deathtouch fix rolls around next update there's some really screwy shennanigans to be had. Not to mention that just having "working" deathtouch is an immense change for green on it's own.
Some stuff, however, is a "conversation opener". Sometimes this is an annoyance, sometimes this is a really good thing.
For annoyance - cantrips and scry cards are effectively unusable by the AI. They were before for sure, but preliminary testing seems to indicate that they still are. This isn't a problem just for someone trying to use a card which happens to have "draw a card" attached to it, it indicates that the AI is using logic that the cards weren't designed for. It shows as nonsense, leads to bad plays, messess up the designers attempts to make sensible decks etc. I'm not gonna badger about it, but whatever is causing cantrip missplays is wrong on multiple levels and no matter how big it is, has to eventually work in a different way.
---
Then there's the "Wow, omg, does it really work?" kind of thing like mulligans. I've noticed in infini-duel that the AI tends to mull down to 5 rather often.
It would be very useful to know howevermuch there is to know about this to prevent screwups. I see two "knobs", one which makes the AI less likely to mulligan ("small numbers go a long way" doesn't tell me enough, unfortunately), and another I don't fully understand either.
---
BUT, here's something that opens up a conversation with "PRAISE JEEBUS S**T FINALLY WORKS LIKE INTENDED!" and puts the "I" into the AI:
- Big Ole Pile of Suddenly Functional Stuff | Open
- .3528 4 Onslaught
.720 4 Mana Drain
.1604 4 Fervor
.7200 4 Earthshaker
.7221 3 Guardian of Solitude
.7265 4 Kami of Fire's Roar
.7380 3 Sire of the Storm
.7749 2 Skyfire Kirin
.7402 4 Teller of Tales
.266 4 Volcanic Island
.5432 4 Shivan Reef
.164 8 Mountain
.126 8 Island
.16153 4 Mausoleum Wanderer
That's not an actual deck. That's a pile of cards in blue and red, off the top of my head, which turned from "unusable by the AI in ways which make you want to throw your compuer out the window" into "apparently works as intended" by the .csv override Korath put in.
In Shandalar.ini you need to find
CreaturesBeforeCombat
and adjust it to 1. Try playing against it first, while it's on 0. Open up the console, look at it hand, see what it does, do 10 or so games. Let it "goldfish" you, don't try to beat it. Then turn the CreaturesBeforeCombat switch to 1, and see what happens then.
The difference is like playing against someone who can't interpret the cards at all, and someone who can. This was where a lot of my exasperation with the AI comes from, not nearly all of it, but plenty of it.
And this is no small deal or something that applies to just those cards by any means. Take Mana Drain - in order for Mana Drain to work, at least with creatures , every creature in any deck Mana Drain is in has to be playable in the first main phase. This still doesn't make Mana Drain "fully operational" as it can still bugger you if you plan on making a deck which intends to use Mana Drain to cast a non-creature artifact (which is generally what it's used for irl). But it just went from "fails almost always and in order to not have it fail you have to know how cards are coded to be able to tell which things it might work with and which it might not" to "works as intended with pretty much all creatures". Big deal? Huge deal!
Then there's Onslaught. You can make any number of decks with Onslaught. Before this change, you could only make exactly one deck with Onslaught - a deck where all your dudes have haste (or otherwise you know which creatures got routinely set a certain way while being coded). Now, again, this opens a conversation - Onslaught targeting is silly and the AI chooses to tap down a 1/1 vanilla instead of a 4/4 vanilla and such. But that's something you wouldn't really be able to even tell, or care about. You can now actually make decks, using any creatures, with a card like Onslaught!
Then there's the Spirits. Vast majority of them with triggers make very little to 0 sense if triggered post-combat. With the vast majority of creatures limited to post-combat and the arcane spells not being there at all, if this isn't turned on there's almost no mechanical point to the spirits. And plenty of them are servicable build-arounds.
And it doesn't stop there, these are just random things I wanted to see working as intended in an official Shandalar release to know that there's a point to trying to put togather various decks. MOST buildarounds work like that.
Fervor and simmilar, which are listed as the main reason for this switch to be added to the file, are just the obvious ones. It's important to notice that fixing
Fervor or
Concordant Crossroads or
Fires of Yavimaya individually isn't going to change the situation.
But we can finally maybe see this stuff in action, and see where they might be bugged or missplayed individually. The big problem with buildarounds isn't just that stuff is/was getting played post-combat. That makes them unusable to the point that you can't tell if there's anything wrong with them rather than the AI. Once you plug in the AI's brain, so to speak, what I see coming up as a problem is sequencing, which is to say that the AI doesn't understand which things in it's deck are buildarounds.
So what might happen is that it values a random vanilla creature or hyper-misvalues a cantrip instead of laying down an enchantment or artifact which is meant to set it's deck up. Like, say, skipping
Fervor in order to play a Grey Ogre. It's been madning and it cropped up in many places with many cards or deck concepts for me allready. There's no way to contextualize a particular card within a particular deck. But just seeing that pile of junk up there going through the correct motions suddenly lit up hundreds of cards as usable and god knows how many decks at the very least worth attempting.
---
Also, the "increase agressiveness" and "increase defensiveness" switches - I can't say I understand the instructions or the way they scale, or how they actually work, but boy do they sound appealing. Seeing the AI unwiling to trade his one dude for 3 enemy ones and just sit there sadly with an army because one might die if it attacks just made me want to cry.
The problem here might be that if the status is global, fiddling with it might mess some colors/decks up. Various Green decks want to madly attack because their cards were designed to trade stupidly well in combat, but what's default mode for Green probably isn't for, say, Blue.
Still, the "won't attack if any one guy could die, but will madly gang block anything with anything" is pretty much the opposite of how MtG is generally played (to the point that "Nobody ever blocks" is kind of a semi-true meme). The only way to explain the insane hyper-defensiveness of Shandalar is the game warping lives format.
If I managed to fiddle with knobs and get the AI to behave more like a (decent) human player, or at least more like the Forge AI, in terms of attacking and blocking, I'd be able to see if and to what degree I'm right about that. Or just get it to play more sensibly, maybe.
V. interesting.
---
Anywho, it's been a cornucopia of additions, bugfixes, workarounds, making hundreds of various things be actually playable, etc. etc.
Made me wonder if that whole "AI can randomly pick a deck from a bunch" thing is still possible to do. It's been soo bloody difficult to get the working decks working last time around, it's still hard to work up steam to sit there and try out different, idk, druids if you know you've got one that works somewhat.
The infinite duels thing picks random decks out of a box, it would be really swell, and very useful, if that could be done with enemies in-game.
I'd make me a Bear themed Druid with
Bearscape and
Mulch, and an Insect themed druid with +1/+1 counter shennanigans, and an Elf themed Druid, and a Beast themed Druid, and a Spider themed Druid, and a Treefolk themed Druid, a druid with a dash of this, and a dash of that - if I went to make a deck I wouldn't have to worry that much about how it compares to all the other concepts and whether it's good enough to replace the current one.
It's also a big deal when it comes to flavor. Green Kamigawa Spirits are in fact great semi-vanilla+ creatures for the Druid. But their flavor is so specific, eastern and abstract that making every druid you meet be a kami-druid would feel weird. If there was 3-4 different potential druids to run into - great, works like a charm, mixes thigs up, a bunch of them are into weird spirits.
Well, that's a feature I'd love if it's something that isn't that complicated but has slipped Koraths mind. When I asked for infini-duels he seemingly whipped them up out of thin air. So no harm in asking, I suppose.
Ok, sorry about the long post, I'll see if I can use what got better and more plentiful since the last time to fix up stuff that I left unfixed, at a relaxed pace and I won't spam.
Posted: 22 Dec 2015, 17:01
I've got another Goblin Lord rework that's not in there yet, too.
moments which are my fault:
That should be easy to fix though. There's a bunch of other potential seers I tried - even a mono-blue 
I do hate the "4 mana 4 toughness reach" creature type, but the AI is the AI and red needs all the help it can get.