So far, I have around a 72% win rate with it! I mainly use hunter and zoolock to keep it simple for the AI. what other decks are good with it/ what is everyone's win rates? I tried a few arena runs with it, drafting the decks myself and it averages about 5 wins! Some bugs: Defender of argus is usually played on the far right, and will usually taunt up one minion. Example: it will use argus on a singular minion with 3 minions on board, even if it has a nerubian egg in hand. What it should do: play nerubian egg and taunt both targets. Polymorph is a tough card for it. it only morphs something last minute, or with atk of 7+ if it has no taunts. it should be removing threats that can get that big, IE anything that buffs itself or will be buffed (undertaker in hunter????) Flamestrike is a hard card for the AI. it will go all in turn 6 without any idea that flamestrike will come next. OR will keep minions on board when they will inevitably die next turn. not sure if this is hard to program, but it typically has no idea that flamestrike is waiting for it when making plays in the mid game. enhance=o=mechanico is a very hard card for it to understand. it will not get full value of it by playing minions before dropping it, if it has the mana to do so. Example: turn 6 with mechanico and nerubian egg in hand, it will drop mechanico BEFORE nerubian. I will keep my eye open for anymore observations!
To add another observation to your post, I have seen the arena draft being pretty bad. For example: Bot picking up mistress of pain instead of doomguard. The AI is not perfect atm but its not bad. the bot in arena needs some work
From what I can tell just by looking at the .xml file, each card only has a single value. So basically, given 3 cards, it would choose the same card no matter what, even if you were playing a different class. This is most likely the key issue with arena at the moment. Simply taking card values from an existing Arena Tier List would be enough to improve the drafting significantly. Then adding in some point additions or deductions based on mana curve would probably make the bot very competitive.
Cool, thanks for that link. That's a relevant observation. OP wasn't asking for support. He was making observations.
Np. He mentioned some bugs. Instead of each person making a new post, its better to collect it in a single post. And since there were some placement bugs, they should be posted as Silverfish errors.
Have you noticed the bot actually worry about placement? While he called it a bug, it's a feature that doesn't exist in the bot. I'm sorry that you've had experiences where you've been told to submit a log, but this doesn't make you a mod. Let them do their job. If they think a log is needed, they'll ask.
Hey guys! been out for a few days, just got back to the post. let's avoid arguing haha. I did not really mean for it to be a bug? I should have been clear in stating that it seems the bot cannot understand the value of certain cards, which may be out of the bots programming entirely. i'll copy and paste my observations in the GvG thread, but i do not really see the point of submitting a log if it has already been mentioned? I made this thread to more or less single out some observations since the beta thread is already full with multiple things.
