[Theory] Hearthstone decks for botting

[JD]

Hey guys,

I wanted to quickly go through my thoughts on HS decks for botting and see what you guys think. I've been playing this game for a while and I've made played quite a few decks so let the brainfart commence!

First things first
Building a Hearthstone deck in general isn't always the same and definitely won't always bring the same results. Some decks are made for rushing down your opponent, some are made for control and some are made to just trade correctly. Really any deck can be used to rank relatively successfully. I believe that trading correctly is worth much more than top-decking Leeroy just when you need it (sorry, I have a personal hatred against people who do that against my token druid deck). If we look at a few of the most played decks you can very easily see that they're all based on the same principle. Zoo is absolutely, completely and purely based on trading your low value cards for your opponent's high value cards. Handlock is about losing just enough life while grabbing enough cards so you can throw your big guys out and destroy your opponent. At this point you're going to be trading your cards/life for making sure you don't get killed before any of that happens. [lexicon]Ramp[/lexicon] druid is about getting to the late game as soon as you can and trading so your opponent doesn't kill you before that happens etc. etc. Trading is extremely important. Good players can (almost) always make the correct trades so even with the shittiest decks they'll be able to do quite a bit of damage.

Bots vs Humans
Theoretically bots would be the perfect traders. They can always almost instantly calculate the value of their own cards vs their opponent's cards. Most humans don't have this liberty. People like Trump obviously get pretty close to making perfect trades all the time but even they make mistakes and mess up on making judgement calls. When it comes to trading correctly bots definitely get a point. Unfortunately for us humans have something bots don't have (yet). We understand how our opponent might act. We understand their psychology and we're better at judging how they'll react to our moves. Another thing humans have is the ability to take a loss so they can come back the next turn and destroy our enemies. I can't tell you the amount of times Hearthcrawler or any other bot I've ever used has made a move that pure mathmatically was the perfect move but that would lose me the game in the long run. Bots aren't good at strategy unless we tell them exactly how to run that strategy. A bot won't think for itself, essentially making it dumb (however much better it is at playing HS than me).

Using bot powers
So. We've stated that bots are good at maths but bad at being clever. How do we play into that? Simple. We make decks that are based on trading well. Zoolock is one of those decks. If you trade perfectly with a zoo deck you'll find that you can very easily destroy most opponents who aren't playing anti-zoo decks. We shouldn't play decks that require us to play fancy tricks (Miracle rogue) or take a huge loss in order to win the game (Handlock). I'll assume you know how to generally build decks and which cards are good for which play style so I won't go into the very basics but I'll explain what to keep in mind when building a bot deck.

Good:

High strength cards that cost little mana.
- This will allow you to trade effectively against higher mana cost cards of the same strength
- This will also allow you to [lexicon]wipe[/lexicon] a board of low value cards

Low value cards that trade well
- These cards don't need to trade against big guys but they're useful for removing small guys
- The bot will automatically try to get as much value out of a card as it can so one of these cards could suddenly trade for twice the amount of mana

Bad:

Combo cards
- Any cards that require you to very smartly play other cards that buff it, give you more cards, damage the opponent based on luck or anything like that

Luck-based cards
- Any cards that require you to be lucky when it comes to using their value. Hell, Hearthcrawler even silences Rag when it can.

Examples

With all this being said I can imagine you'd like some examples so here we go, some non-hero specific cards that the bot could use very well:

Abusive Sergeant
- This card allows the bot to get some very high-quality removal. Use this in combination with a low-mana cost creature and trade that creature for a high value creature.

Argent Squire
- This card allows the bot to get some early game kills without necessarily losing the card itself. It's 1 mana so even if you kill the 2 mana Pally token twice you've got quite a bit of value out of it.

Voodoo Doctor
- This card allows the bot to heal another creature so that creature can kill something without dying.

Crazed Alchemist (IN THEORY)
- I'm not sure if Hearthcrawler can do this but this could theoretically allow the bot to get a high-value kill with a low-value card

Alright, hopefully you understand what to look for in a card now. If you have any questions or if you think I'm absolutely wrong, please let me know. I'd love to discuss this with the lot o' ya!

Cheers,
JD

[zoxxski]

I feel that future bots should use databases for possible decks when facing an opponent. Matching the cards on board with the database (connected to hearthpwn or similar site). Not only then would the math be perfect, but predictions will be easier to make. For an example, not put a handlock below 13hp at round 5 and so on. And for arena bots, there's plenty of good draft tier lists that are updated with every patch that the bots could use to build most optimal decks.

Edit: If there's any bot coder reading my post, I've got loads of solutions and suggestions. But I don't know how to code for 5 cents. Would like to share everything to improve the general botting scene.