the auto complete feature that was recently updated is now creating actual playable decks. does anyone have an idea how the new auto complete works?
I have a feeling that it actually gives you the exact deck that is currently the highest ranked on the ladder in your region. I get Bunnyhoppers list for hunter (who is in the top 10 legend). I get a Shirvallah OTK list that I've seen at around top 30 legend just yesterday in a stream. For shaman I get Kibler's dragon-evolve shaman which could very well be the highest ranked shaman deck right now.
If you put 0 cards in your deck then, yes, it'll just give you the best deck you can create with the cards you have. They flat out said that if you have all of the cards it'll just spit out whatever the #1 deck is at the time.
Thus if you normally base your deck onwhatever VS or whoever says is the best, just leave it empty and hit complete and you'll get it.
If you start stuffing cards in the deck beforehand or your collection has a LOT of empty spots in it, then it'll (in theory at least) give you a more unique deck as it'll look for whatever deck is best based on those limitations.
It's essentially a "Given what cards I demand to have in it, and what cards I own, what's the #1 deck out there?"
Note that's NOT "build a deck around these cards." If the best shaman deck is Dragon-Evolve and you put Juggler, the system will see if it can make Dragon-Evolve anyway if it can still work with Juggler. If you want a more unique deck, youhave to put in enough cards to make it too hard to fit the cards "the b est deck" needs without ruining the deck. Then the system will see what it CAN do.
It's probably the most unique for newer players. When you just don't have hte key cards, there's no way to 'work around' that limitation at all, so you have to find whatever the best deck is that contains your collection in it.
The decks you can create now are quite good. I;m not sure where they got those lists though.
That are no lists, that is all from their database. The algorithm look what cards have the highest winrate with the cards combined you already put in your deck. Means if you start it without any card in your deck it simply builds the highest winrate deck as long as it is possible with the cards you own. If you miss some cards it looks as if there are viable replacements or they try to build a new completely new deck with the second highest winrate.
Easy to understand but in the background damn complex.
Though he doesn't specifically says the words exactly, he does say that the cards added are the most successful cards based on their analytics and what cards you have/what cards you added. You can then see how Regis' statement fits: If the builder is giving you the best cards it sees and you have a full collection and didn't add any cards, then it'll simply input the best deck according to its numbers.
Testing from players now that it's out pretty much confirm this, btw. Also note that it'll keep giving you that same deck if you only put, say, 1 bad card.
edit: Given this i technically shouldn't have said "They flat out said". I misremembered where I got the quote from. "pretty much said" is probably a better qualifier there.
There are definitely some weird outputs though. The zoo deck for example looks more like a tournament deck than a ladder deck (expodinators, looks like it's built to kill greedy decks bit would fold to anything that fights for board).
For hunter it offered me a spell hunter as the 'best' deck. And when I added 2 bearsharks to try get a midrange hunter it gave me spell hunter with 2 bearsharks.
It also seems to do weird things with single copies of cards. I added prince kelseth and captain hooktusk to a deck to try get tempo rogue (unrefined arch-type). The deck had the corpsetaker package but only 1 copy of corpsetaker.
And all this is before the weirdness that happens with incomplete collections (kelseth + other 2 drops).
i test 5 autocompletes to see what it would give before i got bored and just closed game
Shaman: Dragonwock with unstable frog package
Warrior: Lameass odd dragon warrior
Rogue: Some kind of error that was supposed to be a even hooktusk deck. It had so many one offs and things i definitely own two of. Didnt have shadowstep either ewwww
Hunter:Pretty standard scavenging hunter..yawn. easy wins though.
Pal:Of course a lameass odd pal deck id never play...EVER
In order for the auto-complete to optimise one of your decks without disrupting the deck flavour you’d have to pick at least 15-20 cards of your own..
I wish it could choose the best cards to work with the archetype input you have giving it instead of just working out best cards based on power level and win rates.
Like by choosing Bwonsamdi it shoul create a deck with spirit of the dead and the like with highest win rate instead of giving you a Gallery Priest with Bwonsamdi..
I've tried this with druid and put 15 treants and treant synergy cards in the deck and it still made a mechathun deck. The deck was awful but does that mean it doesn't have data on the best treant deck or does it mean its so bad I'm still better off with a mechathun deck?
I have been using a deck that was auto-generated for me with Warlock. I put in two Kobold Librarians to get it started and it generated a rather cool Zoolock deck - different to any I've played before. (So it's not that heal-zoo version; it's kind of a weird hybrid). I gave it a try at rank 5 and after a dozen or so games, I'm pushing rank 3. Woot! It's pretty damn strong. :-) The only irritating loss was against an Odd Paladin. He was on 8 health. And my first Soulfire lost a 1/6 RnG and discarded my other Soulfire. Fuuuuuu----!!! Lol!
I wish it could choose the best cards to work with the archetype input you have giving it instead of just working out best cards based on power level and win rates.
Like by choosing Bwonsamdi it shoul create a deck with spirit of the dead and the like with highest win rate instead of giving you a Gallery Priest with Bwonsamdi..
But how would it do that? How should it work out a whole deck from thin air based on a single card?
What even is a Bwonsamdi deck? Should it just give you the deck recipe for surrender to madness?
The algorithm cannot invent new decks, only report the best performing combinations. If the algorithm can achieve a higher perfomance value by adding Gallery cards, then it will.
I wish it could choose the best cards to work with the archetype input you have giving it instead of just working out best cards based on power level and win rates.
Like by choosing Bwonsamdi it shoul create a deck with spirit of the dead and the like with highest win rate instead of giving you a Gallery Priest with Bwonsamdi..
But how would it do that? How should it work out a whole deck from thin air based on a single card?
What even is a Bwonsamdi deck? Should it just give you the deck recipe for surrender to madness?
The algorithm cannot invent new decks, only report the best performing combinations. If the algorithm can achieve a higher perfomance value by adding Gallery cards, then it will.
I suppose a possible idea would be for it to check for the highest rated deck that includes the selected card (if there is one) and run with that one (assuming it's already got access to finding the top decks for certain archetypes already)
I wish it could choose the best cards to work with the archetype input you have giving it instead of just working out best cards based on power level and win rates.
Like by choosing Bwonsamdi it shoul create a deck with spirit of the dead and the like with highest win rate instead of giving you a Gallery Priest with Bwonsamdi..
But how would it do that? How should it work out a whole deck from thin air based on a single card?
What even is a Bwonsamdi deck? Should it just give you the deck recipe for surrender to madness?
The algorithm cannot invent new decks, only report the best performing combinations. If the algorithm can achieve a higher perfomance value by adding Gallery cards, then it will.
I suppose a possible idea would be for it to check for the highest rated deck that includes the selected card (if there is one) and run with that one (assuming it's already got access to finding the top decks for certain archetypes already)
I'm guessing that was a consideration, but then you get this question:
Say that i put a Bwonsamdi then hit 'Complete'. The best deck that already includes Bwon has, at best, a 30% win rate. However, the system notices that a dragon deck has a 55% win rate that only loses 1% if you replaced a card with bwon.
So do you just look for whatever deck has Bwon or does it try to give you the highest win rate using the cards you put in while noting that Bwon is going to be in it?
From what I gather of it, the system mostly relies on the strength of the individual cards. Deck as a whole is considered but so does raw card strength. So a bad deck with a lot of bad cards that 'works well' with the cards you put are going to be low consideration.
Which makes sense. New players don't need 'fancy silly decks'. They need to know how to take their careds and make it a good deck. So for Version 1 it's an ok decision.
I would like it to be tweaked over time to have more context. I DO like the idea that if you put nothing in it drops the best deck you can make. But I feel like the only reason why people would put 1-2 cards then hit 'complete' are looking for Synergies. They WANT that 30% win rate Bwon deck if all they put is Bwon.
Basicall, I hope they listen to feedback and keep tweaking it over time.
All of you that have never played a Bwon deck, disgust me! That is all! What I orginally named a wall priest before this dumbasss grizzly enforcer garbage spread was a surrender taunt priest with bwon. Truly, a wall priest.
I hate that the deck builder does not give you the best deck USING the cards you put in, as people would expect. Instead it seems to give you the best deck DESPITE of the cards you already put in. More often than not, there is a synergy between the cards deck builder puts in, but not necessarily synergy with the cards YOU put in. For example, it will keep putting murlocs in your Shaman deck even if your cards have absolutely nothing to do with murlocs.
I hate that the deck builder does not give you the best deck USING the cards you put in, as people would expect. Instead it seems to give you the best deck DESPITE of the cards you already put in. More often than not, there is a synergy between the cards deck builder puts in, but not necessarily synergy with the cards YOU put in. For example, it will keep putting murlocs in your Shaman deck even if your cards have absolutely nothing to do with murlocs.
But what do you consider as the best deck using your cards? If you objectively look at the win rate of the deck, then adding murlocs is the best outcome.
Adding anything else might have supposedly better synergy but not the data to back that up (either no data or data showing a worse win rate).
And then how does the algorithm decide upon a synergy? Are cards marked as synergistic or does the algorithm calculate that?
Even in that case, if you added 10 cards you want to play, the algorithm has 20 cards to fill with better muloc cards – those have less synergy with your cards but much more synergy among themselves.
Rollback Post to RevisionRollBack
To post a comment, please login or register a new account.
Hi Hearthpwners!
the auto complete feature that was recently updated is now creating actual playable decks. does anyone have an idea how the new auto complete works?
I have a feeling that it actually gives you the exact deck that is currently the highest ranked on the ladder in your region. I get Bunnyhoppers list for hunter (who is in the top 10 legend). I get a Shirvallah OTK list that I've seen at around top 30 legend just yesterday in a stream. For shaman I get Kibler's dragon-evolve shaman which could very well be the highest ranked shaman deck right now.
what do you guys think?
cheers
If you put 0 cards in your deck then, yes, it'll just give you the best deck you can create with the cards you have. They flat out said that if you have all of the cards it'll just spit out whatever the #1 deck is at the time.
Thus if you normally base your deck onwhatever VS or whoever says is the best, just leave it empty and hit complete and you'll get it.
If you start stuffing cards in the deck beforehand or your collection has a LOT of empty spots in it, then it'll (in theory at least) give you a more unique deck as it'll look for whatever deck is best based on those limitations.
It's essentially a "Given what cards I demand to have in it, and what cards I own, what's the #1 deck out there?"
Note that's NOT "build a deck around these cards." If the best shaman deck is Dragon-Evolve and you put Juggler, the system will see if it can make Dragon-Evolve anyway if it can still work with Juggler. If you want a more unique deck, youhave to put in enough cards to make it too hard to fit the cards "the b est deck" needs without ruining the deck. Then the system will see what it CAN do.
It's probably the most unique for newer players. When you just don't have hte key cards, there's no way to 'work around' that limitation at all, so you have to find whatever the best deck is that contains your collection in it.
One does not simply walk into Mordor,
unless they want to be the best they can be.
could you link the blog post where they confirm this? would be nice!
Lol, I didnt know they updated this feature.
The decks you can create now are quite good. I;m not sure where they got those lists though.
Blizzard has access to all the game's data and statistics. They should be able to know which decks are best
That are no lists, that is all from their database. The algorithm look what cards have the highest winrate with the cards combined you already put in your deck. Means if you start it without any card in your deck it simply builds the highest winrate deck as long as it is possible with the cards you own. If you miss some cards it looks as if there are viable replacements or they try to build a new completely new deck with the second highest winrate.
Easy to understand but in the background damn complex.
The 'empty deck = #1 deck" came for Regis Killbin specifically,
https://www.reddit.com/r/hearthstone/comments/avxtus/more_info_on_the_smarter_deck_builder_coming_in/
If you want a Blizzard response, the closest would be Mike who gave more info on the builder:
https://www.reddit.com/r/hearthstone/comments/ay645h/hearthstone_developer_ama_year_of_the_dragon/ei132dd/?conte
Though he doesn't specifically says the words exactly, he does say that the cards added are the most successful cards based on their analytics and what cards you have/what cards you added. You can then see how Regis' statement fits: If the builder is giving you the best cards it sees and you have a full collection and didn't add any cards, then it'll simply input the best deck according to its numbers.
Testing from players now that it's out pretty much confirm this, btw. Also note that it'll keep giving you that same deck if you only put, say, 1 bad card.
edit: Given this i technically shouldn't have said "They flat out said". I misremembered where I got the quote from. "pretty much said" is probably a better qualifier there.
One does not simply walk into Mordor,
unless they want to be the best they can be.
There are definitely some weird outputs though. The zoo deck for example looks more like a tournament deck than a ladder deck (expodinators, looks like it's built to kill greedy decks bit would fold to anything that fights for board).
For hunter it offered me a spell hunter as the 'best' deck. And when I added 2 bearsharks to try get a midrange hunter it gave me spell hunter with 2 bearsharks.
It also seems to do weird things with single copies of cards. I added prince kelseth and captain hooktusk to a deck to try get tempo rogue (unrefined arch-type). The deck had the corpsetaker package but only 1 copy of corpsetaker.
And all this is before the weirdness that happens with incomplete collections (kelseth + other 2 drops).
i test 5 autocompletes to see what it would give before i got bored and just closed game
Shaman: Dragonwock with unstable frog package
Warrior: Lameass odd dragon warrior
Rogue: Some kind of error that was supposed to be a even hooktusk deck. It had so many one offs and things i definitely own two of. Didnt have shadowstep either ewwww
Hunter:Pretty standard scavenging hunter..yawn. easy wins though.
Pal:Of course a lameass odd pal deck id never play...EVER
Fun > Meta
it works off of Team5's internal data and some neural network shenanigans, tney keep improving it too, also wild support is in future too
In order for the auto-complete to optimise one of your decks without disrupting the deck flavour you’d have to pick at least 15-20 cards of your own..
I wish it could choose the best cards to work with the archetype input you have giving it instead of just working out best cards based on power level and win rates.
Like by choosing Bwonsamdi it shoul create a deck with spirit of the dead and the like with highest win rate instead of giving you a Gallery Priest with Bwonsamdi..
I've tried this with druid and put 15 treants and treant synergy cards in the deck and it still made a mechathun deck. The deck was awful but does that mean it doesn't have data on the best treant deck or does it mean its so bad I'm still better off with a mechathun deck?
RIP hsreplay I guess.
I have been using a deck that was auto-generated for me with Warlock.
I put in two Kobold Librarians to get it started and it generated a rather cool Zoolock deck - different to any I've played before. (So it's not that heal-zoo version; it's kind of a weird hybrid).
I gave it a try at rank 5 and after a dozen or so games, I'm pushing rank 3. Woot! It's pretty damn strong. :-)
The only irritating loss was against an Odd Paladin. He was on 8 health. And my first Soulfire lost a 1/6 RnG and discarded my other Soulfire. Fuuuuuu----!!!
Lol!
But how would it do that? How should it work out a whole deck from thin air based on a single card?
What even is a Bwonsamdi deck? Should it just give you the deck recipe for surrender to madness?
The algorithm cannot invent new decks, only report the best performing combinations. If the algorithm can achieve a higher perfomance value by adding Gallery cards, then it will.
I suppose a possible idea would be for it to check for the highest rated deck that includes the selected card (if there is one) and run with that one (assuming it's already got access to finding the top decks for certain archetypes already)
I'm guessing that was a consideration, but then you get this question:
Say that i put a Bwonsamdi then hit 'Complete'. The best deck that already includes Bwon has, at best, a 30% win rate. However, the system notices that a dragon deck has a 55% win rate that only loses 1% if you replaced a card with bwon.
So do you just look for whatever deck has Bwon or does it try to give you the highest win rate using the cards you put in while noting that Bwon is going to be in it?
From what I gather of it, the system mostly relies on the strength of the individual cards. Deck as a whole is considered but so does raw card strength. So a bad deck with a lot of bad cards that 'works well' with the cards you put are going to be low consideration.
Which makes sense. New players don't need 'fancy silly decks'. They need to know how to take their careds and make it a good deck. So for Version 1 it's an ok decision.
I would like it to be tweaked over time to have more context. I DO like the idea that if you put nothing in it drops the best deck you can make. But I feel like the only reason why people would put 1-2 cards then hit 'complete' are looking for Synergies. They WANT that 30% win rate Bwon deck if all they put is Bwon.
Basicall, I hope they listen to feedback and keep tweaking it over time.
One does not simply walk into Mordor,
unless they want to be the best they can be.
All of you that have never played a Bwon deck, disgust me! That is all! What I orginally named a wall priest before this dumbasss grizzly enforcer garbage spread was a surrender taunt priest with bwon. Truly, a wall priest.
Fun > Meta
I hate that the deck builder does not give you the best deck USING the cards you put in, as people would expect. Instead it seems to give you the best deck DESPITE of the cards you already put in. More often than not, there is a synergy between the cards deck builder puts in, but not necessarily synergy with the cards YOU put in. For example, it will keep putting murlocs in your Shaman deck even if your cards have absolutely nothing to do with murlocs.
But what do you consider as the best deck using your cards? If you objectively look at the win rate of the deck, then adding murlocs is the best outcome.
Adding anything else might have supposedly better synergy but not the data to back that up (either no data or data showing a worse win rate).
And then how does the algorithm decide upon a synergy? Are cards marked as synergistic or does the algorithm calculate that?
Even in that case, if you added 10 cards you want to play, the algorithm has 20 cards to fill with better muloc cards – those have less synergy with your cards but much more synergy among themselves.