2019 Holiday Exchange!
A New and Exciting Beginning
The End of an Era
  • posted a message on Generating Magic cards using deep, recurrent neural networks
    Quote from Talcos »

    To be fair, you are expanding the network more than you think, because the network units at one layer are fully connected to the network units in the next layer, so if you have N nodes at each layer, that's N^2 new connections and N^2 new weights that we now have to take into consideration.

    And worse how, exactly?

    EDIT: Just a thought, have you tried using dropout to help train the network? like, if you do -dropout 0.5, then on each pass, we disable half of the network units. It's a very powerful regularization technique.

    Worse as in outside of those, tons more gibberish and just meaningless cards that drag off halfway through text.

    I'll try dropout in a future run - still playing with parameters, I didn't expect it would be as much of an expansion. Makes sense though. I'll get to that after I tweak the VB settings as:

    Quote from SirSpunky »
    Some performance tips for people running VirtualBox on Windows:

    Settings -> Storage -> Controller: SATA -> Check "Use Host I/O Cache". This alone made my generation twice as fast. However, I read that it will increase the risk of data corruption in your virtual OS, just so you know. I haven't had any problems so far though.

    Settings -> System -> Processor -> Processor(s) -> Increase to max (within green color). Requires 64-bit Linux installation.

    To be able to install 64-bit Linux in VirtualBox, you must enable virtualization in your normal BIOS. Using a 64-bit OS should also provide a speed boost.

    There are other settings that you can experiment with, like Base Memory (I put mine at 2048 MB), Video Memory and 3D Acceleration (I enabled it and set video memory to max 128 MB), and if you have your virtual HD on an SSD you should probably check "Solid-state Drive" in Storage settings for your virtual HD.

    On my 4 year old Intel i5-2500K, I'm getting around 0.3 sec/batch on default settings (128 rnn_size, 2 num_layers) on a virtual Ubuntu.

    And that's likely the issue - I jammed up the RAM for the VB, but it looks like I'm still on a single processor when there's 4 available. Virtualization and 64 bit is enabled as well (to ensure I was able to use all the memory I wanted to). I'll try that after this current run.

    Edit: Killed run and restarted, 3 layers went from 18s -> 12s per batch at size 600 with 4 CPUs and host cache.

    Not seeing any performacne increase with dropout enabled. I'm reducing rnn size and getting 3s/batch at 300 and 3 layers. I'll see how this goes for the next day or so.
    Posted in: Custom Card Creation
  • posted a message on Generating Magic cards using deep, recurrent neural networks
    I'm also having issues with speed once I jump from 2 layers to 3. For size of 600 it goes from 2.1 seconds at 2 layers, but 15-17 second at 3 layers.

    Also sad to see what I ran overnight seemed to have produced worse results than yesterday evening.

    Fezek Sfaster GG
    Whenever a player copies $THIS, it deals 3 damage to target creature or player.

    Stone Shaspelin W
    Destroy target nonbasic land.

    Neubreal Storm 3W
    Creature - Druin
    $THIS enteres the battlefield tapped.
    T: Add W or U to your mana pool. Draw a card.

    Cage, the Cruelters 6UU
    Legendary Artifact Creature - Dragon
    3W: $THIS deals X damage to target creature or player.

    Guardian Hornor 4GG
    Creature - Cat Monk
    Whwnever you draw a card, target player loses 5 life

    Posted in: Custom Card Creation
  • posted a message on Generating Magic cards using deep, recurrent neural networks
    Another batch from tonight before I go to bed:

    Firsh Exprout BB
    Whenever you cast a Fulm, you may rach onderif 6 (Attacking creatures you control have "RT,Shiff 6(3: Put a level counter on this. Each other artifacts who pope.)
    2W: Target land becomes a 2/2 Wurm.

    Row of Madgusion B
    Enchantment - Aura
    Enchant creature
    At the beginning of your upkeep, enchanted creature gets +1/+2 until end of turn.

    Healion Hoors 2BBB
    Legenday Creature - Human Advisor
    You may have 5.

    Karphlasterm 1B
    Creature - Human Druid Wizard
    Sacrifice $THIS: Activate this ability only during your upkeep.
    Posted in: Custom Card Creation
  • posted a message on Generating Magic cards using deep, recurrent neural networks
    Been playing around with this last night and today.

    Been gettign some good balanced results, but trying to figure out how to have them not violate the color pie so blatantly.

    I also want to split by types, but that gets a lot of unique hybrids...but most are awful.

    Examples I've gotten:

    Russark Caba 1G
    Legendary Creature - Dwarf Elf
    When $THIS enteres the battlefield, put a 1/1 white Goblin Archer creature token onto the battlefield

    Iron Weader B
    Put a 1/1 white Soldier creature token onto the battlefield.

    North the Undisider
    T: Add G or W to your mana pool.
    Equip 2

    Silver Blood 1
    T: Put a 1/1 white Dragon creature token onto the battlefield.

    Lava Dead
    T: Add 1 to your mana pool.
    1GGT: Target player reveals his or her hand onto the battlefield face down as a 2/2 creature. $THIS deals 4 damage to you.

    As Blook 1G
    Enchantment - Aura
    Enchant numbed creature

    Eycondance Scarpice 1UU
    Enchantment - Aura
    Enchant creature
    Enchanted creature gets +1/+1.
    Flashback W (You may cast this spell any time you could cast an instant)
    Flashback (This creature can't be blocked except by two or more green or black creatures.)

    Razos Mony Ampharian 2WW
    Creature - Spirit
    Whenever a player attacks you, search your library for up to five cards, put them onto the battlefield, then shuffle your library.

    And then gibberish like this:

    Seash Panistmer 2R
    As an additional cost to cast $THIS, it deals damage equal to its power to target creature. Another target creature becomes a Monntefilk.
    Posted in: Custom Card Creation
  • posted a message on Large changes coming to PTQs/GPs/PTs
    Quote from asmallcat »
    Quote from Georg51 »
    This also makes it harder for vendors to bulk buy and sell singles. PTQ's provided an inexpensive chance for smaller vendors to get face-to-face time with customers. Now they can only do that at GP's, Invitationals, and Opens, where the cost and availability are significantly higher.

    Translation: singles may become less available and more expensive as a collateral effect from this change.

    While there are lots of issues with this change, I don't think this is one of them. The lion's share of all of these companies' sales were either online or through their brick and mortar store (if they had one), not at these events. I doubt this will have any appreciable impact on singles prices, except that people won't see the occasional great sales that vendors sometimes had at these events.

    At first , I saw this and didn't really feel concerned. I saw another vendor mention something like this as well. How common is external vending at PTQs? I've been to at least 20 in the last 18 months, and I think only 2 I've seen external vendors (as in, not the store). One was one where the store's buddy was vending (and vends at all those store events, basically is part of the store), the other had 4 vendors. Does this really hurt? I didn't know it was common for vendors to be at so many PTQs, I've only had the store running it vending most of them.
    Posted in: The Rumor Mill
  • posted a message on [Primer] Jund
    Quote from W0lf90 »
    I dislike 4 ravines as i find i draw them too often, i cut one to make space for a grove and have no looked back.

    I feel you often dont want to see 2 unless the game goes long and never 3. Having 3 ravines has worked great for me and reduced those double opening ravine hands that hurt so much. It was diffrent when we had treetops because they could be useful early and double activations were often possible as well as rarely treetop+ravine.

    You can (almost) never double activate ravine which means multiples are only useful for if the first gets answered.

    That said i dont think playing 4 is wrong.

    This is exactly how I feel as well. It's one of the worst cards in our deck in multiples, and because of that it's just a Shivan Oasis a good majority of the time. I've only attacked with two Ravines once in all the games I've played with this deck. I have also cut down to 3, going to a Copperline Gorge in place of one, may even replace it with a Stomping Ground just because of how taxing the CIP is sometimes around turn 4-5 when I really need to play two spells at once.

    As for affinity hate, I'm at 1 Rakdos Charm 2 Ancient Grudge now. I've moved off Shatterstorm just because it's too expensive and doesn't hit things like inkmoth when you need it to. I've had several games where I barely stabilize with a shatterstorm T4, and then T5-6 they pump inkmoth a bit and suddenly I'm dead. Once the affinity player knows its coming, they'll preemptively put the +1/+1 counters on it. Though since I added Slaughter Pact this hasn't been as bad.
    Posted in: Modern Archives - Proven
  • posted a message on [Primer] Jund
    The real other questions the lists from Huska and Edel yesterday answer seem to be that 24 lands, 2 Courser, and Anger of the Gods main are pretty good. I picked up Slaughter Pacts to work with yesterday, really a fan of them. The only card choice I'm not really on board with is the EEs in the board, and going to 3 Fulminators, though I can see their use. I've been struggling with Affinity, but I think maybe those EEs and MD Angers will help with the matchup.

    After yesterday, the real question is - do we have a better answer against Scapeshift? The lists yesterday show that they were very strong against creature based decks and combos, but seem a bit weaker to Storm or Scapeshift outside of our discard suite.
    Posted in: Modern Archives - Proven
  • posted a message on [Primer] G/R Aggro/Midrange
    I'm assuming mentioning here:


    I'm not a fan of the 5th scry land, but I still think I'd rather it be red. I'm definitely disliking the flesh//blood in place of the fourth mortars main. Agree with the chandra #2 main though, it's so much better to make a dude unable to block instead of a 2/2 from xenagos (worst case).

    As for sideboarding, Ketter's board is pretty easy to figure out what goes in, but what comes out in the various matchups?
    Posted in: Colossal Gruul
  • posted a message on [Primer] G/R Aggro/Midrange
    Looking at the list that got T4 at SCG Nashville today, mainly I'm trying to see what the use of the Arbor Colossus and Bow in the SB are for:

    Also, Chandra main? 4 Mortars, 0 Flesh//Blood main? I understand the Unravels over Destructive, but the other choices seem...off to me?
    Posted in: Colossal Gruul
  • posted a message on How much is a misprint Boros Charm?
    Quote from jaytothen
    Gatecrash had a lot of 'fuzzy' printing issues from what I've seen and read from others. I wouldn't put a premium on them in trade personally.

    This. There were whole boxes that were printed like this. I wouldn't treat it any differently than a normal Boros Charm.
    Posted in: Market Street Café
  • posted a message on [PRIMER] Esper Humans / Midrange
    I agree as well. Our G1 against UW/Esper is just miserable. Maybe 30-70 at best? Post board is pretty much the opposite, assuming we hit what we board in. But post board we just have so many threats that can win games alone, we can usually force them to run out of counters/removal. I think my record game 1 against UW is like 1-7, and I've ended up winning all but 2 of those matches. We can easily adjust our game plan around what they have, this deck is that powerful and flexible. I can't really give a good rule of thumb, but the key to the post board is just taking the removal that hits what is in your hand (a pretty good rule for thoughtseize in general).

    Also, the mirror is a miserable game 1 to play. It usually ends with each side having a Whip, Obzedat, and Solider of the Pantheon and you just chump things and have life jump up and down 2-3 pts at a time.
    Posted in: Standard Archives
  • posted a message on SCG Open Indianapolis
    Did he also just play Mutavault, attack, then Nykthos after combat? I know he got a mutavault from Jace, looked like it went down immediately, yet he played a Nykthos without question later that turn.
    Posted in: Standard Archives
  • posted a message on Esper Aggro?
    I think it's a very strong deck, powerful against the UW decks and the devotion decks that are based on 1 for 1 trades. I don't like how it matches up against the red decks though, the hyper aggressive red is a terrible matchup for it, and red devotion decks are slightly unfavorable.
    Posted in: Standard Archives
  • posted a message on Unable to start game, localized to single account
    I got DC'ed earlier today during a daily, and since then I have been unable to join any game.

    When the game would begin to load, the chat pops up for it, but the game screen never starts. For tournaments (such as the daily), after 10 minutes I get dropped. For other matches (such as in practice rooms), this happens after 1 minute.

    In the practice rooms, I start (or join) a game, and after one of the players hit start, the status of the game says "Connecting" as opposed to "Dueling" or "Completed", and then after a single minute, it switches to completed, with out the game ever starting.

    What the main issue is, is that this is restricted to a single account. I've tried multiple other computers and internet connections, and this issue persists, through both V3 and Wide Beta. If I use a second account, that account can play just fine.

    Anyone seen this issue before and/or know a fix for it?
    Posted in: Other Formats
  • posted a message on [Event Discussion -] Grand Prix Dallas/Fort Worth
    Quote from thememan
    Anybody else aggravated a bit by the commentary on the Red-Devotion vs UW control match-up? Jensen didn't do anything particularly amazing. This is not knocking Jensen in the least, and there were plenty of beautiful plays on his part, but he didn't dig out of an unwinnable hole like they are seemingly implying. Against a slower creature-based deck, U/W is standing pretty as they know what to watch out for with huge fanatics, and Verdicts just killed the board position that was gained before there was ever much of a problem. Granted, if Domri didn't brick so much it would have been a bit more drawn out, but I honestly don't think it mattered much at all by the time the second verdict went off. And the revs certainly helped. It's how the deck works, and it's well suited for that red match-up.

    Noticed that slightly, same again with this game - they're making it sound like Huey is super lucky to have staved off the attack, yet 10 turns ago they said "Oh, Huey has like a 90% shot of this game) which makes no sense.
    Posted in: Standard Archives
  • To post a comment, please or register a new account.