All decks no matter how bad or strong should be around 50% - Math Theory

[Zipnut]

If you're deck is good or bad, you will drop down to the ELO that your deck belongs, or move up. If you move up, you will eventually plateau and stay at where the deck belongs and will continue to win 50% of games. The same happens with a bad deck. You will drop down enough until you hit the level which your deck belongs and you will then stay around 50%.

In theory, it does not matter what deck you have when botting... Just sit back and enjoy the ride.

 

[zuabros]

it does happen. same thing at world of warcraft arena. That won't only be true at the both extremes: #1 player for example will have 100% wins, and the worst player of all will have 100% lose...

 

[Zipnut]

it does happen. same thing at world of warcraft arena. That won't only be true at the both extremes: #1 player for example will have 100% wins, and the worst player of all will have 100% lose...

We are talking about a bot there, along with a card game that mathematically will even out statistically over a long period of time. The bot will never lose 100% of the games. You will eventually get down to such a BAD elo that you will beat people worst than you are. If you are the worst, then you are talking 1 outta 1,000,000.

 

[scrampc]

then, theorically, bot should surrend 1 match after wining

 

[Diuuude]

then, theorically, bot should surrend 1 match after wining

You should mix a lot of alcohol with drugs, add some other toxic products to it and then drink the biggest possible amount of this mix, please.

You still can loose 10 in a row, and then win 10 in a row. If your fucking stupid theory is applied, then you loose 10 in a row and loose half of your wins just because you can't read a shit.

This game is also based on luck. You can face #1 player and just get a perfect card draw every turn, while he will keep drawing high cost cards. I defeated a top 100 player while I was at rank 3, 90% of this win has been luck at card draw. Every turn I had the best possible card I could draw.

 

[greenxeyezz]

Theoretically this is false.

If this were true, there would be no rank climbing. In order to rank climb, you must have a greater than 50% win rate. Why?
Simple, no bonus stars at higher rank, so lets assume no bonus stars at rank 20.

Rank 20, 0/3 stars. Win 1, 1 star. Rank 20 1/3, Win 2, 1 Star, Rank 20 2/3, Win 3 Rank 20 3/3 stars. Win 4, Rank 19 0/3 Stars. Win 5, Rank 19 1/3 Stars. Lose 1, Rank 19 0/3 Stars. Lose 2, Rank 20 3/3 stars, Lose 3, Rank 20 2/3 Stars. Lose 4, Rank 20 1/3 Stars, Lose 5 Rank 20 0/3 stars.

50% win rate.

But what you are trying to say that your deck is going to get a 50% win rate just at different levels.
As in, I start at rank 20, win 20 games get to rank 15, and then at rank 15 I end up with a 50% winrate if i have a "good" deck, therefore I would stay at rank 15. Or move to rank 14 fall to rank 15 over and over based on a 50% win rate.

And then you are saying that if my deck sucks, I would drop to rank 23 (if possible) and stay there beating decks crappier than mine and losing to decks better than mine.

So your saying the bot is only as good as my deck, and my deck determines my rank.

That would be correct if EVERYONE was a bot and played with the same logic. Statistically this is WRONG, but I wont divulge.

Yes, having a 50% win rate is a GOOD rate for a bot, but its not mathematically proven that every deck will eventually fall to a 50% win rate.

 

[vinfinite]

This is so wrong, common sense dictates that a bot playing a deck vs. a real player using the same deck would result in the player winning. The bot's logic is extremely primitive and a lot (if not all) spells are used very inefficiently. You can easily prove that a bot with ANY kind of deck will not achieve a 50% or 40% or even a 30% win.

The bot is built with a specific logic right now, and YOU as a player have to build a deck that will take advantage of this logic. Therefore if you build ones with complicated spells or creatures with specific abilities, the bot will be more handicapped than if you had just put in a Chillwind Yeti, Tazdingo, etc.

Also, if you just browse the forum a TINY bit, you'll notice that there are a bunch of threads asking for deck builds because they can only achieve a 10/20/30% win rate. THEREFORE, not all decks are equal and you will not achieve a 50% win rate unless you build your deck with the consideration of how the bot's logic works.

 

[Diuuude]

This is so wrong, common sense dictates that a bot playing a deck vs. a real player using the same deck would result in the player winning. The bot's logic is extremely primitive and a lot (if not all) spells are used very inefficiently. You can easily prove that a bot with ANY kind of deck will not achieve a 50% or 40% or even a 30% win.

The bot is built with a specific logic right now, and YOU as a player have to build a deck that will take advantage of this logic. Therefore if you build ones with complicated spells or creatures with specific abilities, the bot will be more handicapped than if you had just put in a Chillwind Yeti, Tazdingo, etc.

Also, if you just browse the forum a TINY bit, you'll notice that there are a bunch of threads asking for deck builds because they can only achieve a 10/20/30% win rate. THEREFORE, not all decks are equal and you will not achieve a 50% win rate unless you build your deck with the consideration of how the bot's logic works.

You are more or less right. The only problem here is that your bot can't just downrank enough to find opponents as bad as the deck you've built. After an Eleven hours botting session, my pally deck is at 43% win rate. Not that bad and not that far from the 50% the OP is talking about. I can't go lower than rank 20, maybe this deck belongs to some kind of rank 22 or 23 to achieve its 50% win rate.

A player will not "always" win to a bot. If the bot's logic is good enough, it can go very far in the ladder with no problem. Just look at the chess bots. They can easilly defeat almost any decent chess player, and believe me, not all Hearthstone players are as smart as Kasparov is.

 

[vinfinite]

You are more or less right. The only problem here is that your bot can't just downrank enough to find opponents as bad as the deck you've built. After an Eleven hours botting session, my pally deck is at 43% win rate. Not that bad and not that far from the 50% the OP is talking about. I can't go lower than rank 20, maybe this deck belongs to some kind of rank 22 or 23 to achieve its 50% win rate.

A player will not "always" win to a bot. If the bot's logic is good enough, it can go very far in the ladder with no problem. Just look at the chess bots. They can easily defeat almost any decent chess player, and believe me, not all Hearthstone players are as smart as Kasparov is.

Obviously an AI can be smarter than humans, and are you honestly comparing hearthbuddy's logic to that of a chess bot? That is built with a TEAM of mathematical geniuses? Why don't we just say that the hearthbuddy team will build a bot smarter than Watson, who will go out and win Jeopardy. I find it ridiculous that you'd even make a comparison between this bot and those that are sponsored by corporations.

Here we are talking about realistic goals for this current bot, which REALISTICALLY, will never reach the level of a chess bot, or anywhere near it. So let's compare apples to apples please.

As to your other points, yes, if we can rank even lower than 20, then perhaps we can achieve the coveted 50% on every single deck. But obviously we can't, so deck makeup will ALWAYS be a factor in achieving a higher rank.

 

[Endus]

Technically the logic behind the 50% win rate is solid, but you have to keep in mind that you could build a deck so bad that you'd only be getting a 50% win rate at like rank 40 (which doesn't exist, obviously). So since the true bottom rank is 20, if your deck is bad you may only have a 10%-30% chance of winning.

Which makes the current challenge with the bot to make a deck that wins 50% of the time @ rank 20 or higher. Once the bot is finished (and the API is public to allow for custom profiles) I'd assume it would be much more feasible to obtain a 50% rate @ higher ranks rather than at the bottom of the barrel as we are now currently.

 

[zuabros]

Technically the logic behind the 50% win rate is solid, but you have to keep in mind that you could build a deck so bad that you'd only be getting a 50% win rate at like rank 40 (which doesn't exist, obviously). So since the true bottom rank is 20, if your deck is bad you may only have a 10%-30% chance of winning.

Which makes the current challenge with the bot to make a deck that wins 50% of the time @ rank 20 or higher. Once the bot is finished (and the API is public to allow for custom profiles) I'd assume it would be much more feasible to obtain a 50% rate @ higher ranks rather than at the bottom of the barrel as we are now currently.

Currently, with warlock decks, we can get very very near to the 50% winrate, using lots of low-life minions that buff themselves, just like a cheap version of a murlock deck.

As it is now, bot logic is very primitive, and it fails at the mechanics to draw cards using special cards (like the hunter buzzer or the priest card). That's why warlock decks work better with this bot.

Once the bot logic can take us to rank 19, then the 50% winrate is guaranteed.

Nowadays no bot deck (with no epics) was able to get us over rank 20, so we are still below the 50% winrate. We are probably at rank "21" (if it was possible).

 

[Mario27]

you could make the bot the same as a human player would play all you need to know is to have the understanding of the cards how they work like big game hunter would only destroy a minion with 7 attack or higher board clear with paladin is around 4 5 Minionwith equality pyromance equality consecration equality Avenging wrath if you make the bahavior like that we would have more wins.

 

[vinfinite]

you could make the bot the same as a human player would play all you need to know is to have the understanding of the cards how they work like big game hunter would only destroy a minion with 7 attack or higher board clear with paladin is around 4 5 Minionwith equality pyromance equality consecration equality Avenging wrath if you make the bahavior like that we would have more wins.

Obviously they are trying to make the bot more human-like, I think people seem to believe that making an intelligent bot is easy. It is one thing to program a bot to do exactly what a card is made to do, and another thing to make the bot choose to combo cards based on board situations. Good AI is really hard to make, and the HS bot is very much still in beta. The fact that it can achieve 40% regularly with a decent deck is a pretty great feat already, not saying that I don't expect it to get better of course. Eventually I'd like it to be good enough to get at least a few wins in arenas, beyond that though, I don't care since I just need it to grind the gold.

 

[Mario27]

Obviously they are trying to make the bot more human-like, I think people seem to believe that making an intelligent bot is easy. It is one thing to program a bot to do exactly what a card is made to do, and another thing to make the bot choose to combo cards based on board situations. Good AI is really hard to make, and the HS bot is very much still in beta. The fact that it can achieve 40% regularly with a decent deck is a pretty great feat already, not saying that I don't expect it to get better of course. Eventually I'd like it to be good enough to get at least a few wins in arenas, beyond that though, I don't care since I just need it to grind the gold.

ye we need to have a board clear combo with all classes im a paladin player i was that in wow aswel a holy one and if we get a better board control with those combos the bot would own all players with board control i mean you only use aoe spells around 3 4 minion for paladin its equality pyromancer equality consecration equality avenging wrath etc. we need to get a bahavior settings like in the other bot so we can make our own settings

 

[Fibula]

That 50% win rate thing is true but it's more like a 55%~ win rate thou and it only applies to casual play.
In casual play you are matched by your MMR and Blizzard stated in one interview that you lose MMR much faster then you gain it. For example if you lose 3 games your MMR will drop, if you then win 3 games it will not be back at the same level as before it will still be lower. I boted around 2700-3000 games in casual mode, first with Bot of Stone then with HSB. At first I was getting something around 30% win rate but then after some time my MMR dropped so low that bot got matched against really easy opponents. Then I started to get around 53-55% win rate. I managed to get few classes to 60th level and all of them had 55% win rate. When I'm watching bot I can see that players that he is fighting with are still on the same level of game, they have almost only basic cards, rarely a legendary and they make a lot of stupid mistakes.

So as you can see I have quite a big number of games played with it and I can say that with a win rate around 55% you still lose enough MMR to hold you at the same very low level.

But ranked is a different story. In ranked you are matched only by your rank. On rank 20 you will be matched with people at rank 20 most of the times. If you have 0 stars you will be matched with someone on lower rank (21) and with 3 stars you will be matched with someone on higher rank (19). Things are a little different on ranks just before legend (4-1) but that's because there are not that many players at that ranks.
Because of that many people are asking for auto concede option to lower their rank. For example if bot would get to rank 18 he should concede himself back to 20. That concedes should not be counted towards win rate because they would mess the stats.

So you can make a deck and AI that would win like 80% of games - you only need it to concede himself (and not count that concedes to win rate) to keep him at lower rank.

At this point I am boting ranked games and it's already quite good. In mornings I would get above 50% win rate most of the times. Because of that I need to check him and derank him manually because he tends to rank himself up quite fast with win streaks. He got to rank 17 many times already (he would probably go even further but I always deranked him at that point). In weekends and evenings it's lower, something around 40-45%.

With API in our hands and option to concede himself to get back to lower ranks (or even with option to auto concede against certain classes) it will be quite easy to get 70% and I do believe there will be people able to make AI logic and deck that will get 80% and more.