Hey, the bot is playing really poorly. What can I do to optimize its algorithm? Note that all the things I am describing happened with the maximum number of own turn board (80K) and respectively the other similarly worded option (1K).

I observed the bot playing and this is something that legitimately happened and not only once:

Opponent is at 12 HP, he has a 7/4 on the board. I have 25 HP (so no danger of dying practically) and 3 Silver Hand Recruits. The bot chooses to (in this very order) attack face with the 3 Silver Hand Recruits, than play 'to arms men' (to those unaware: Quartermaster) and than uses hero power. This is absolutely infuriating! Why on earth is hero power being used after and not before playing Quartermaster? Why are Silver Hand Recruits attacking dealing net 3 damage instead of net 9 damage after being buffed?

During the next turn the computer attacked with all the Silver Hand Recruits for lethal when it could have attacked with 2 Silver Hand Recruits, a Quartermaster and weapon charge to cover it's butt in case of Ice Block (it's better leaving the oppo at 1 hp rather than 3 hp baring he has no Alexstarsza in hand). He ended up not having Ice Block up, but the computer could not and was not aware of that.

Another scenario playing paladin: computer plays Acolyte of Pain and ends its turn when it could have hero powered, because it had the mana to do so (clear misplay as there was no danger of MC Tech with only 1 or 2 minions on the board).

But wait, there's more: yesterday playing Hunter there was lethal achieved by 2 independent combination of cards.

Wolfrider + bare Kill Command + Hero Power (8 damage in total) or Ironbeak Owl + buffed Kill Command + Hero Power (7 damage in total). The opponent has precisely 7 Health Points, there was sufficient mana for both plays. He ended up not dying that turn. What the heck?


These are all very obvious and grave mistakes.

Sorry for sounding rather harsh, but I mean no ill will. I am interested in helping this software achieve a desirable efficiency.

Are you saying the developer shouldn't be interested in improving his software? There are obvious ways to improve the algorithm of the bot. Oh, and my MMR is so high that the bot has a 10% winrate. Your 'few' hours are 10+ hours in my case.

Sorry lad, I'll write Blizzard a mail asking them to reduce my MMR based on my skill as a real player, because the bot I am leaving on overnight can't handle my opponents due to obvious misplays. Yeah right :D

On a more serious note, there are seasons where I tend to play casual games instead of ranked, because I can't be arsed to climb for hours to finally face some decent opponents. Hence my casual MMR is pretty high (I only face legend players).

Are you saying that conceding for many hours straight would tank my MMR? I thought conceding in the mulligan phase of the game had no effect on someone's MMR, but I may be wrong.


By the way, I do intend to use the bot for climbing to rank 10 every season for reasons mentioned above, but in its current state that is impossible. However, I plan on changing this with some help from the developer. If the developer of course is at all willing to improve his software.

A good start would be optimizing the bot for a single deck and I think we can all agree that Face Hunter AKA 'Me Face, Me Hunter' being the easiest deck around and getting a lot of free wins from his opponents not having the right answers is the perfect candidate for this role. As you may or may not be aware, a sub 50 per cent win quote is enough to climb to legend given enough games. Due to the short duration of an average hunter game, a decently playing bot would be able to get pretty far in the ranked system I imagine.

Now, I have posted a thread (http://www.hearthranger.com/forum/yaf_postst4846_Reword-rules-I-formulated-into-conditions-for-DeckRules-Excel-sheet.aspx) devoted to making an Excel sheet that would serve as a kind of filter to the bot's default behaviour. Unfortunately from what I can see, some conditions haven't yet been created (such as the condition to sum up the attack of the enemies' minions to check for whether he has lethal if I my minions don't trade) and as such are impossible to implement.

Instead of creating decent deck rules, try to find a good bot deck and all with default settings is a much easier way to climb up to rank 10.

I'm sure hearthranger bot can climb up to rank 10 or higher very easy, already have user reported he climbed up to legendary with this bot.

Yes I'll add more conditions in the future, now if you want to control the bot with more details beyond current supported , you can try to program it with silverfish's source.