Tired of small sample sizes claiming high win rates

[vinfinite]

Tired of small sample sizes claiming high win rates - 24+hr run stats

So I just wanted to vouch for this bot. It doesn't have the greatest logic, and more often than not makes terrible terrible plays, but honestly you can't expect a bot to get much better than 50% WR (and those complaining at 50% don't understand what the point of this bot is). I was tired of people posting like 10-30 games with 50-70% win rate or what not, obviously this is possible when you have a very small sample size. I've had runs that starts 7-1 and then ends with 7-20. What i'm trying to say is, a small sample size does not give you consistent results so I left my bot on for 24+ hours to give you the following results:

(notice that it says 9:XX:XX, but every time you start/stop the bot it resets the time, and I check it every so often to see how many wins my rogue currently has)

Another note: I think rogues may be one of the poorer classes for the bot to handle, because it does not utilize the dagger very well. I was able to achieve consistent 50% with my paladin, since the hero power is just a simple 1/1 creature.

I am going to leave my bot running for another day, and if you guys are interested I can show you the stats of that later.


ALSO, I love version .50, it hasn't crashed for over a day, excellent work!

EDIT: I'm not posting this to brag, because I know 40% really isn't that good. But I just wanted to show the potential of this bot, and also I've been using it for 100% ranked play.

 

[Endus]

Honestly as long as the bot can get near 50% playing ranked it's good enough. I can usually get in the high 40s with most decks when built properly, and the bot can only improve from here.

 

[greenxeyezz]

50% is the perfect "regular" bot play with standard logic. HOWEVER, it depends on peoples decks.

So since this bot is being developed by an individual with an individuals play style in mind, HIS deck should achieve ~50% with HIS playstyle with HIS logic, HOWEVER, the basic logic should be finalized before implementing profile logic, which in turn should result in a higher than 50% win rate because then the bot will play closer to YOUR play style with YOUR deck choices.

 

[vinfinite]

50% is the perfect "regular" bot play with standard logic. HOWEVER, it depends on peoples decks.

So since this bot is being developed by an individual with an individuals play style in mind, HIS deck should achieve ~50% with HIS playstyle with HIS logic, HOWEVER, the basic logic should be finalized before implementing profile logic, which in turn should result in a higher than 50% win rate because then the bot will play closer to YOUR play style with YOUR deck choices.

I agree and understand what you're saying. But the BOT will never be good enough to achieve a high rank in ranked play. The bot cannot understand the nuances of the game, even if we are able to put in custom API. At low ranks, the bot can and will get much smarter and thus more wins, but at higher ranks when playing against competent players with competent decks, the bot will never out think their player counterpart.

And yes to address what you said, I know it's built by an individual and marketed towards a diverse group of players and that is why I'm saying, even at 40% this is wonderful considering the bot wouldn't be able to play your deck the way you made it to be played.

 

[greenxeyezz]

I agree and understand what you're saying. But the BOT will never be good enough to achieve a high rank in ranked play. The bot cannot understand the nuances of the game, even if we are able to put in custom API. At low ranks, the bot can and will get much smarter and thus more wins, but at higher ranks when playing against competent players with competent decks, the bot will never out think their player counterpart.

And yes to address what you said, I know it's built by an individual and marketed towards a diverse group of players and that is why I'm saying, even at 40% this is wonderful considering the bot wouldn't be able to play your deck the way you made it to be played.

Definitely not trying to argue here, but in RANKED play, yes I forsee it being in about 40-50% range is "perfect" and ideal to grind out those 500 wins, golden cards/portraits etc. In casual, I can see the winrate being ~50-60% with a good deck and good custom API profile.

Where the bot should be going to improve is in Arena. Im not looking for 12 win games, but with a good custom API, and especially if it can use a program as simple as ArenaValue, to "build" the decks i think we can achieve around 3-6 wins on average. Which to me is ideal. It may not always break even with gold (as in 150 gold for arena ticket, you win and get 1 pack and 50 gold in a box) so essentially it is like buying a pack) but at 3-5 wins we get an "extra" box and can definitely be UP on either cards/dust/gold.

Therefor we can use bot to grind our daily's, grind our 100 gold cap, and then play our arena to break even or higher.

I think accepting 40-50% win rate as an optimal bot is not where it SHOULD be. a 40-50% win rate can be won with the free "color" bot that is out there that literally plays the first card to the left of the hand and has ZERO logic built into it but smash the hero. (I used it extensively with a hunter deck and came just shy of 44% with it, and my deck was NOT optimized)

So i think with a bot that has LOGIC and the ability to be profile API, we SHOULD be able to, in some game types, achieve a better solution that a mouse clicker bot, else there is no point for the LOGIC in the bot.

So yes, I agree with you but needed to elaborate on what I meant.

 

[vinfinite]

Yeah, exactly what I was thinking. I was speaking in terms of ranked play only. I don't really understand the point of casual play (since you can't earn golden, and the slightly higher win rate doesn't mean much). But yeah, in casual play where the rank matters less so, I'm sure the win rate can be improved drastically.

And yes, I did not take Arena into consideration. I've played many arenas (the only mode I play), and it will be hard for the bot to play it correctly because even if there is an API released, choosing a "general" action per card won't work well in Arena. It would work really well in a constructed deck since you know the synergy you are going for and therefore will program your card action towards that end. However, in arenas, choosing cards due to ranking is not always the best and using a specific card for a normal constructed action will not always yield you the best case. THAT being said, I think the bot would at least be able to achieve 3 wins (first 3 wins are always so easy), which would make Arena-ing with your bot somewhat cost efficient (if it manages to achieve slightly higher than 3 every so often).

But yeah, I'm not trying to argue about any of the points stated earlier, I was simply showing what the bot could achieve 'at the present moment'. I know that the behaviors have immense room to grow and being able to program your own action would help tremendously. But again, the bot seems far from that and as is right now, it is relatively good.

 

[greenxeyezz]

At least you are with the minority in understanding that there is a difference between BOT logic and PROFILE logic. It seems when people are making posts about issues/card specific issues/or bad bot plays most are talking about PROFILE logic and not BOT logic, and at this point in time, will not help anyone or anything. So I think the next best thing to do is to separate all the logic's and make it easier on the few developers that are working on the bot right now to help fix these issues and get it closer to API release lol.