Collectible Card Game Headquarters

Hey all,

Love Forge! Been using it alot for the last few weeks.

I especially like draft, but it seems I am getting the same darn cards over and over.

Lots of alphay myr and phrexian dredgers...oh a few blastoderms thrown in...but if I look at the possible cards to choose from...it seems it 95% of the same !

I saw a post for an update that had "draft every possible card"...but I can't seem to find it.

It would be great if I could spice up the draft card pool..its getting stale.

Any suggestions/help ?

Josiphos wrote:Hey all,

Love Forge! Been using it alot for the last few weeks.

I especially like draft, but it seems I am getting the same darn cards over and over.

Lots of alphay myr and phrexian dredgers...oh a few blastoderms thrown in...but if I look at the possible cards to choose from...it seems it 95% of the same !

I saw a post for an update that had "draft every possible card"...but I can't seem to find it.

It would be great if I could spice up the draft card pool..its getting stale.

Any suggestions/help ?

`
The draft mode is using a cube set to determine which cards are available in the draft. Look in the forge/res/draft/ folder and you will see the three rarity files. You can load them into a text editor.

This topic has some links to various files including an editor which will allow you to assign rarity values to all or some of the cards: Helper Files - tokens, cards pictures, set editor

To use a "flat rarity" so that every card has the same chance of appearing, download the attached file, and put the files common.txt, uncommon.txt, rare.txt in the directory /res/draft/

Likewise, I stole the rarity files from the res folder and overwrote the rarity files in the draft folder. That way you maintain the rarities, you just get to draft with a lot more cards (most of them crappier, I found).

Yeah, most of them are crappier.

I'm sure it's been suggested before, but it would be neat to have a utility that could select and create those files easily from the current universe.

Depending on what attributes are maintained, we could generate a pool by set or block. Or on a simpler level, just take the current pool and have a utility/macro that automatically deleted the cards you didn't want.

Perhaps an alternative could be a set of files that ONLY contained cards you don't want, that would also be loaded at startup, like: xcommon.txt, etc.

Just brainstorming here...

Being able to select a block for either draft or sealed mode would be a very cool feature. If it could be done in card packs as it is now the case when you win them in quest mode, it would be great.

So, if I wanted to conduct a draft with the Mirage block, we'd get packs from Mirage, Visions & Weatherlight to "pass around the table". As for sealed, we could get 1 pack from each for 45 cards to go & build a deck while getting whatever basic land we needed.
Or we could do the old Tournament pack of 75 cards that included basic lands and a few extra packs from said block to give players & AI more of a pool to choose from.

There is/was an editor the mtgrares put together, which only works with the rarity text files. We have since then, switched to using rarity as part of the cards.txt for all other functions.

The draft rarity is pulled from text files only becuase it would have been too difficult to switch it as the last version was released.

With some work, you could limit a draft to one set by only including the cards from that set. However, it is highly unlikely that we support enough cards from any one set to make drafting that way worth it. Maybe enough for a block, so if it's not critical that it be a full pack of any one set, then that my be possible. Keep in mind that some cards are not playable by the AI, so leave those out of the draft too.

That's why the draft mode is based on a cube-draft concept.

We have a gameplan for handling set-specification, at least the data portion.. actually making use of it will present a number of challenges.

Packs containing cards from all 3 sets of a block would be great too.

Maybe cards from a main set like M11 or from what was the current release at the time of whatever block we select like 4th Edition for example, could be added into the pool of cards to help fill out the packs we got for either sealed/draft play.

Your point about the cards that the AI can't play with gets me to ask how the decks created for the AI are made for sealed/draft play and is there a way for us to know what cards the AI has trouble playing with?
Cards like Necropotence are fairly obvious as troublesome for it, but is there a list of cards that point them all out?

There's an "SVar:RemAIDeck:True" in cards.txt for each card that is removed from sealed and random decks.

After a quick look at some of the different cards in that file, I can see quite a few already with the "SVar:RemAIDeck:True" code.

I wonder though, when I see a card like Nature's Ruin why that card is one marked for non AI playability. It seems like a simple enough card that targets just one colour of card. It gets me thinking how difficult it is to create a really good & competent AI opponent when playing a human.

Ever since the days of the Microprose game, all variations of the game that have had an AI component to it have resulted in a flawed system, doing things like attacking with all hands when down to less then 3 life for example where a human would always hold back a creature or two to block for the opponent's next turn.
The one thing that has never been able to be improved on in our case with ManaLink is in improving the way the AI plays the game. I'm not a programmer, but I've heard plenty through the years that it would need a total re write and new code to do such a thing. Even if that was done, I wonder how much better it could even be made to be?

IMO, I think that Forge plays better in some ways then does our old game, and is one of the better implementations of the AI that I've seen. Is it due somewhat to an advancement in coding from the early 90's when the Microprose game was written as opposed to todays' level of coding?
It goes to show that programming the AI to play MTG at a high level is much more difficult then doing so for an AI playing chess.
There are just so many variables that have to be accounted & calculated for that the complexity for a human like AI opponent seems very unlikely at this stage.

The new Duels of the Planeswalkers that came out last year is pretty good AI wise, given the limited 260 or so card that it came with, but still does some predictably bad moves that no human would likely do. Throw in some of the new cards that are being added by the mod's and you see the AI struggle even more.

Anyway, I find all this fascinating and I marvel at what you programming guys can accomplish and work around.

Thanks for the peek behind the curtain!

The "SVar:RemAIDeck:True" is somewhat miss-named. This SVar removes the card from the list of cards that can show up in a randomly generated deck. There are a number of reasons why we would do this. Nature's Ruin has the SVar because it only kills green creatures.

The SVar is removing badly broken/buggy cards, cards the AI can't use, cards that are only useful in limited situations, etc.

Forge's AI improves slowly and in very small steps. There are a number of good reasons for this but they are all fairly technical and do not make good conversation outside of the developers forum.

Some people are very competitive and it is unlikely that forge will satisfy their needs. They might be happier playing against human beings. People like myself settle for playing against a weaker computer AI and we still have fun, then again I am not very competitive.

Duels of the Planeswalkers has exactly the advantage of a VERY limited cardpool. (and a budget) They can spend the extra time on better AI for those cards, better combat AI, even letting the AI respond at instant speed.

The architecture we're working with right now is just too limiting for more advanced AI. The situational AI approach for playing particular spells is actually pretty good. Combat decisions are always going to be tough.

A chess AI is simple as all game state information is known, and all possible moves can easily be tested and discarded.

Anti-color cards tend to not work well for random decks, because they're random. Either you or the computer will sit there with Nature's Ruin in hand the entire game if there aren't any green creatures to be destroyed. However the AI should be able use it effectively given the right circumstance. We want the card to be available, because in quest mode you can build a color-hoser or strategy-hoser deck against the various AI decks. It would be good against King Kong and his squirrel army, for example.

Thanks again to both of you for expounding on some of the reasons & methods of what makes this engine run. It confirms what I have said, that all of the people involved with Forge are very talented folks and we're lucky as people who all love MTG to have you here!

To answer the original poster's question: you can thank me for the current state of drafting since I picked the cards in the draft set

If you want to roll your own draft rarity files go for it! The text files are easy to edit manually or you can copy / paste from a Spreadsheet (this is how I do it).

I highly suggest you will want to keep the draft structured with some sort of rarity though. There are a *lot* of bad cards in the history of Magic. By using the rarity files you get to limit the number of bad cards and shape the draft format. You can select the best aggro creatures, or make a draft set that supports a lot of comobos, or you could put in all of the most broken cards like Black Lotus, Ancestral Recall, etc. just to see what happens.