Mechanics, themes and archetypes that need support to work

[lujo]

And also because of "Mind Control", but generalist just doesn't and can't work and only leads to people trying to make decks to have the odd nervous breakdown and be unable to actually do anything meaningful. It didn't even work for the original pool, too many decks had just one thing that they can use in it and everything else was unusable filler.

Shandalar needs to have scripted opponents to work, obviously leaning on a global base of common things, but it just can't work any other way.

[BlueTemplar]

But then you would have to make a script for each AI deck... and possibly change it each time you modify the deck... and then it would tend to break at the first unconventional situation, so you end up with a very rigid AI that you spent a lot of effort on; see this for an example (Sparky) :
http://arstechnica.com/gaming/2011/01/s ... etition/1/

Actually, what might save effort is that if these scripts were automatically generated, this explains the (relative) success of neural networks and genetic algorithms in recent years :
http://games.slashdot.org/story/10/11/0 ... algorithms

(P.S.: In a way, what you're doing by making decks for the AI, is already a form of "scripting".)

[BlueTemplar]

I'm sorry that you're getting frustrated and burned out. If it's any consolation, consider that (from what I've seen), Shandalar's AI is actually above average as video games go.

And also consider that you have the luxury of adapting the decks to the AI only. There are plenty of games where you would be forced to make decks (or their equivalent) that are both fun enough FOR the player to play as well as against.
And then also games where the same decks would have to be fun and competitive in multiplayer. (Most of developers seem to have dropped the ball on the competitive part, or are building their games around the AI instead of the opposite.)

Maybe you just need to chill out a bit, take a break? That Centaur deck it the very last one you haven't tweaked? Maybe don't push yourself too hard on that one, for instance what kind of evasion could be centaur-themed? How does the AI deal with having cards with Menace for instance?

[lujo]

Eh, all deck building IS scripting, that's the problem. What I currently have to do is lousy, time-inefficient scripting which doesn't allow for so many things to be scripted that it's astounding. Ofc I would have to script each deck, and rescript when I change it - that would be paradise compared to now and I could churn out hundreds of decks without burning out. Making a deck work for the player, be enjoyable to play and play against, work in multiplayer, any of it - that's easy it's just that these programs are generally made by folks who're much better at programming than they are at MtG. I'm not, I'm just really good at MtG*. I made and sold cubes for fun before there was a widely known concept of "cube" (and full fledged custom draft sets people used to regularly have largeish lan-party tournaments with because regular drafting is expensive), making a hundred-odd multi-purpose multi-tier constructed decks is a joke provided the bloody AI could be made to play each of them correctly. And I'm not even talking about the complicated ones, either.

And it wouldn't need to be scripted bottom up - all it really needs is ability to customize priorities for phases on a per-deck basis, and ability to base certain decisions on the presence of individual cards, from what I've seen. It would turn making any deck from a several-hour thing (with low-to-limited chance of actual sucess) to a half-hour thing at most - 5 minutes to make a deck (you could wake me up in the middle of the night and ask me to rattle off a decklist for any concept, even with special parameters, and provided the cards are in the pool and work I think I'd be on a 100 out of 100 success with just some playtesting. I could probably also rattle off a list of options for scaling it, too, so playtesting would be just a matter of settling on an option out of several for whatever slot), 10 minutes to script necessary sequencing priority and add expected targeting checks for things to play on it's own stuff, 10 minutes to go through phases to make sure stuff is allocated to be played in the correct phases if it matters. I can't see it being more complicated than that, if it is, something's not right. With a better deckbuilder (or just the whole pool made to be a set in Forge) to double check my instinctive choices for other options, make it 10 minutes to make a deck. 0 Frustration whatesoever. Some actual power/curve playtesting to make sure I didn't arse up and the deck flows properly.

I do need a break, deffinitely, but the all the decks need retweaks, the wizards haven't been done, and making the centaur work is simple - I just make it exactly the same as every other deck. Making a deck brutally efficient and capable of beating you is easy, but making a deck playable against, with this AI, isn't.

6 more decks to go, I'll do them, then retweak them, then I'm off sorting the lists at a leisurely pace. Before mulligans, some degree of scripting assistance for individual decks and ante tweaks there's not a whole lot that can be done. I'm satisfied with some of the decks, but not with most of them at all, and if people are, well, cool, but the evaluation standards are really low and most of these decks are way more powerful than they should be for reasons that were out of my hands. I could easily be on the receiving end of a very scathing, disparaging and sarcastic review of the pool and I'd have no way to defend myself but to point a finger at the AI and engine in general. But it would be hard to find someone who'd understand how wrong these decks are who'd also believe that someone seriously attempted to bother making decks for it before improving/changing the AI significantly.

*Which makes sense, because if you spend enough time on MtG to get good enough to be qualified for it it's a wonder if you have barely any skills to feed yourself with otherwise. And if you've spent enough time on getting good enough at anything else to feed yourself, like programming, chances are good I know/understand more about magic than you (you might even be a good player, too, but unaware that there's even things to be aware off beyond a certain front-end sphere - wotc and top pros generally keep quiet about the inner workings of stuff) It's not something I'm proud of, or that's very useful in general, and I highly envy anyone who's put anough time and effort in their life towards acquiring an actually worthwhile skill, but when I'm not being annoying and inadvertedly offensive I'm actually quite handy to have around if you're doing something like this.