OpenAI how it works , Ensage might port in a thing or 2 ? :P

Alvin92Alvin92 Posts: 48

https://blog.openai.com/dota-2/

the blog above sums up what the bots does. :doge:

Comments

  • AyushAyush Posts: 356

    @Alvin92 said:
    https://blog.openai.com/dota-2/

    the blog above sums up what the bots does. :doge:

    My Dear Friend, you asking ensage to build something like openai? It's just impossible no one here are intelligent scripters/coders ( no offence just stating obvious , most of the scripters are average) and coding something like openai requires high level intelligence ( which none have ATM in this firm ) dendi himself said it's too strong and arteezy lost 10 games with 0 wins... with ensage you can max win games in 5k mmr ( that too 25%winrate) and arteezy (almost 10k lost to it that too 10 times) now you tell me ensage building something like openAI... lol IMAO never possible , I'd best say it better to improve its support community first (lol),anyways something like openai for ensage is impossible don't get your hopes up with those average coders( no offence again,just stating obvious).

  • NeeDforKillNeeDforKill Posts: 1,976

    @Ayush написал:

    @Alvin92 said:
    https://blog.openai.com/dota-2/

    the blog above sums up what the bots does. :doge:

    My Dear Friend, you asking ensage to build something like openai? It's just impossible no one here are intelligent scripters/coders ( no offence just stating obvious , most of the scripters are average) and coding something like openai requires high level intelligence ( which none have ATM in this firm ) dendi himself said it's too strong and arteezy lost 10 games with 0 wins... with ensage you can max win games in 5k mmr ( that too 25%winrate) and arteezy (almost 10k lost to it that too 10 times) now you tell me ensage building something like openAI... lol IMAO never possible , I'd best say it better to improve its support community first (lol),anyways something like openai for ensage is impossible don't get your hopes up with those average coders( no offence again,just stating obvious).

    What you talking on? This is even not "AI" , what it only can play on sf on mid, that's all. Even Lua moones meepo doing alot more.
    It win pro player because they write some stuff like "Ensage" and writed "script" , ofc it win pro players because it can detect autoattacks, particles , timing and have alot faster reaction than human.
    Idk how to possible be so stupid, to think it's impossible to make or it's real ai. Real ai you can google for StarCraft 2 , where devs show how it works and how it analysis game. This one grab all values from game like Ensage.

    Ofc noone will make it on Ensage based, because no one need script what can stay on mid vs one hero.

  • AyushAyush Posts: 356

    @NeeDforKill said:

    @Ayush написал:

    @Alvin92 said:
    https://blog.openai.com/dota-2/

    the blog above sums up what the bots does. :doge:

    My Dear Friend, you asking ensage to build something like openai? It's just impossible no one here are intelligent scripters/coders ( no offence just stating obvious , most of the scripters are average) and coding something like openai requires high level intelligence ( which none have ATM in this firm ) dendi himself said it's too strong and arteezy lost 10 games with 0 wins... with ensage you can max win games in 5k mmr ( that too 25%winrate) and arteezy (almost 10k lost to it that too 10 times) now you tell me ensage building something like openAI... lol IMAO never possible , I'd best say it better to improve its support community first (lol),anyways something like openai for ensage is impossible don't get your hopes up with those average coders( no offence again,just stating obvious).

    What you talking on? This is even not "AI" , what it only can play on sf on mid, that's all. Even Lua moones meepo doing alot more.
    It win pro player because they write some stuff like "Ensage" and writed "script" , ofc it win pro players because it can detect autoattacks, particles , timing and have alot faster reaction than human.
    Idk how to possible be so stupid, to think it's impossible to make or it's real ai. Real ai you can google for StarCraft 2 , where devs show how it works and how it analysis game. This one grab all values from game like Ensage.

    Ofc noone will make it on Ensage based, because no one need script what can stay on mid vs one hero.

    I don't know why your reacting in a negative and disrespectful manner, and why your burnt(lol), with all respect I never told u all to make an Ai, I just told "something like Ai is impossible in this forum so don't get your hopes high" sir, and sir I don't think there is a meepo mooned script 2 days ago my subscription ended I will renew in the coming time , but I don't think there is moones meepo script I hope you get your eyes checked ( in a respectful manner ),

    And one more thing, the creators of open ai said they were building a 5 man team by the next TI so they told they wanted to challenge the previous TI winner( team liquid) so obv it won't stay mid all the time idk how you thought it will be stayongbmid all the time (this is why I said ensage does not have good coders with high level intelligence, only above average with no proper logic , only good scripted who has that level of intelligence I came across in this firm is vicktherock)

    Finally with due respect please be respectful to all the members of this firm who are buying your product @NeeDforKill

  • PennyPackerPennyPacker Posts: 102
    edited August 2017

    @NeeDforKill said:

    What you talking on? This is even not "AI" , what it only can play on sf on mid, that's all. Even Lua moones meepo doing alot more.
    It win pro player because they write some stuff like "Ensage" and writed "script" , ofc it win pro players because it can detect autoattacks, particles , timing and have alot faster reaction than human.
    Idk how to possible be so stupid, to think it's impossible to make or it's real ai. Real ai you can google for StarCraft 2 , where devs show how it works and how it analysis game. This one grab all values from game like Ensage.

    Ofc noone will make it on Ensage based, because no one need script what can stay on mid vs one hero.

    Where did you see that it's grabbing values from game like ensage and that it's like a script?
    It doesn't appear to be on their website.

    They said they never coded a strategy into the bot.

    https://www.twitch.tv/videos/166172514

    If you skip to 7:59:50, they talk about how their bot became good. Never did they mention writing scripts.

    Interviewer: "What's the big break through?"
    OpenAI Dev: "It's hard to say there is a particular big break through, it's a series of small improvements. They reached the level of casual players, then, they learn to play aggressively, and then, they learn to bait."

    Definitely doesn't sound like a script to me.

    As for it only being able to play SF on mid. 1v1 mid with one hero is challenging enough. They have achieved this with huge success. Now they said they will work on 5v5 for the next TI. Also they want the bot to be able to play on the same side as human players.

  • StiflerStifler Posts: 38

    Definitely doesn't sound like a script to me. Any script on Ensage that learns to improve itself?

    =))) Let them make "Hello world" guide already and i can write dll that learns to improve itself

    with all respect I never told u all to make an Ai, I just told "something like Ai is impossible in this forum so don't get your hopes high" sir

    It should be an opensource, so he can hope that after release it will be ported into ensage

  • NeeDforKillNeeDforKill Posts: 1,976

    @PennyPacker написал:

    @NeeDforKill said:

    What you talking on? This is even not "AI" , what it only can play on sf on mid, that's all. Even Lua moones meepo doing alot more.
    It win pro player because they write some stuff like "Ensage" and writed "script" , ofc it win pro players because it can detect autoattacks, particles , timing and have alot faster reaction than human.
    Idk how to possible be so stupid, to think it's impossible to make or it's real ai. Real ai you can google for StarCraft 2 , where devs show how it works and how it analysis game. This one grab all values from game like Ensage.

    Ofc noone will make it on Ensage based, because no one need script what can stay on mid vs one hero.

    Where did you see that it's grabbing values from game like ensage and that it's like a script?
    It doesn't appear to be on their website.

    They said they never coded a strategy into the bot.

    https://www.twitch.tv/videos/166172514

    If you skip to 7:59:50, they talk about how their bot became good. Never did they mention writing scripts.

    Interviewer: "What's the big break through?"
    OpenAI Dev: "It's hard to say there is a particular big break through, it's a series of small improvements. They reached the level of casual players, then, they learn to play aggressively, and then, they learn to bait."

    Definitely doesn't sound like a script to me.

    As for it only being able to play SF on mid. 1v1 mid with one hero is challenging enough. They have achieved this with huge success. Now they said they will work on 5v5 for the next TI. Also they want the bot to be able to play on the same side as human players.

    Did you trust anyone who say it? They need to make show and hype, "AI" win Dendi . You really think someone in TI or who look TI was interesting how it works? . I told you there is no proofs how this bot analysis game? Why Google +Blizzard Project for Starcraft 2 have alot proofs how it works, how it analysis game.

    Ofc it will good, cos they have tons of coders and mb it will in future works like a "AI" with learning , bot not now.

    P.S There is bot with "AI" it have some learning and analysis algorithms, maded by 1 guy, based on Bot of Legends

  • NeeDforKillNeeDforKill Posts: 1,976
    edited August 2017

    @Ayush написал:

    @NeeDforKill said:

    @Ayush написал:

    @Alvin92 said:
    https://blog.openai.com/dota-2/

    the blog above sums up what the bots does. :doge:

    My Dear Friend, you asking ensage to build something like openai? It's just impossible no one here are intelligent scripters/coders ( no offence just stating obvious , most of the scripters are average) and coding something like openai requires high level intelligence ( which none have ATM in this firm ) dendi himself said it's too strong and arteezy lost 10 games with 0 wins... with ensage you can max win games in 5k mmr ( that too 25%winrate) and arteezy (almost 10k lost to it that too 10 times) now you tell me ensage building something like openAI... lol IMAO never possible , I'd best say it better to improve its support community first (lol),anyways something like openai for ensage is impossible don't get your hopes up with those average coders( no offence again,just stating obvious).

    What you talking on? This is even not "AI" , what it only can play on sf on mid, that's all. Even Lua moones meepo doing alot more.
    It win pro player because they write some stuff like "Ensage" and writed "script" , ofc it win pro players because it can detect autoattacks, particles , timing and have alot faster reaction than human.
    Idk how to possible be so stupid, to think it's impossible to make or it's real ai. Real ai you can google for StarCraft 2 , where devs show how it works and how it analysis game. This one grab all values from game like Ensage.

    Ofc noone will make it on Ensage based, because no one need script what can stay on mid vs one hero.

    I don't know why your reacting in a negative and disrespectful manner, and why your burnt(lol), with all respect I never told u all to make an Ai, I just told "something like Ai is impossible in this forum so don't get your hopes high" sir, and sir I don't think there is a meepo mooned script 2 days ago my subscription ended I will renew in the coming time , but I don't think there is moones meepo script I hope you get your eyes checked ( in a respectful manner ),

    And one more thing, the creators of open ai said they were building a 5 man team by the next TI so they told they wanted to challenge the previous TI winner( team liquid) so obv it won't stay mid all the time idk how you thought it will be stayongbmid all the time (this is why I said ensage does not have good coders with high level intelligence, only above average with no proper logic , only good scripted who has that level of intelligence I came across in this firm is vicktherock)

    Finally with due respect please be respectful to all the members of this firm who are buying your product @NeeDforKill

    It was not to offense some one. In some times ago, Ensage was Lua hack, with some paid scripts. It have moones combo for all heroes in dota, and meepo script , what have alot functions in auto mode. It farm/push lines, kill enemies, farm/stack jungle, dodge enemy skills, avoid enemy players, makes low hp meepos to go to base, and alot more suff.

    About how it not possible to make , this script for bot of legends, maded by one human.

  • PennyPackerPennyPacker Posts: 102
    edited August 2017

    @NeeDforKill said:

    Did you trust anyone who say it? They need to make show and hype, "AI" win Dendi . You really think someone in TI or who look TI was interesting how it works? . I told you there is no proofs how this bot analysis game? Why Google +Blizzard Project for Starcraft 2 have alot proofs how it works, how it analysis game.

    Ofc it will good, cos they have tons of coders and mb it will in future works like a "AI" with learning , bot not now.

    P.S There is bot with "AI" it have some learning and analysis algorithms, maded by 1 guy, based on Bot of Legends

    I don't trust anyone who says it. No.
    I think the reason why Deepmind and Blizzard have more details, is because they made an agreement I think it was last year to use Starcraft 2 as a test bed for AI researchers.

    Valve and OpenAI haven't made such a deal to my knowledge. So I wouldn't expect the same arrangement or details to be released. Who knows in the future.

    I do expect OpenAI to release a paper on their SF bot at some point in the future. They are doing this for research purposes after all. Computer games are a test bed for AI.

    There's no proof it's a script either. But a company that has been put together and been invested in by Elon Musk, isn't making a script to beat some pros at a computer game.

    That iKita bot, is made by an individual who has monetary gains to make from it. I wouldn't trust someone who's sole purpose is to make money. It may have some machine learning in it but I wouldn't think it's very advanced.

    OpenAI is a non-for-profit company that has a goal to make generalized AI. I don't believe it is going to waste its time on a script to beat pro players.

    Elon Musk's tweets:

    "OpenAI first ever to defeat world's best players in competitive eSports. Vastly more complex than traditional board games like chess & Go."

    "Would like to express our appreciation to Microsoft for use of their Azure cloud computing platform. This required massive processing power."

  • AyushAyush Posts: 356

    @NeeDforKill said:

    @Ayush написал:

    @NeeDforKill said:

    @Ayush написал:

    @Alvin92 said:
    https://blog.openai.com/dota-2/

    the blog above sums up what the bots does. :doge:

    My Dear Friend, you asking ensage to build something like openai? It's just impossible no one here are intelligent scripters/coders ( no offence just stating obvious , most of the scripters are average) and coding something like openai requires high level intelligence ( which none have ATM in this firm ) dendi himself said it's too strong and arteezy lost 10 games with 0 wins... with ensage you can max win games in 5k mmr ( that too 25%winrate) and arteezy (almost 10k lost to it that too 10 times) now you tell me ensage building something like openAI... lol IMAO never possible , I'd best say it better to improve its support community first (lol),anyways something like openai for ensage is impossible don't get your hopes up with those average coders( no offence again,just stating obvious).

    What you talking on? This is even not "AI" , what it only can play on sf on mid, that's all. Even Lua moones meepo doing alot more.
    It win pro player because they write some stuff like "Ensage" and writed "script" , ofc it win pro players because it can detect autoattacks, particles , timing and have alot faster reaction than human.
    Idk how to possible be so stupid, to think it's impossible to make or it's real ai. Real ai you can google for StarCraft 2 , where devs show how it works and how it analysis game. This one grab all values from game like Ensage.

    Ofc noone will make it on Ensage based, because no one need script what can stay on mid vs one hero.

    I don't know why your reacting in a negative and disrespectful manner, and why your burnt(lol), with all respect I never told u all to make an Ai, I just told "something like Ai is impossible in this forum so don't get your hopes high" sir, and sir I don't think there is a meepo mooned script 2 days ago my subscription ended I will renew in the coming time , but I don't think there is moones meepo script I hope you get your eyes checked ( in a respectful manner ),

    And one more thing, the creators of open ai said they were building a 5 man team by the next TI so they told they wanted to challenge the previous TI winner( team liquid) so obv it won't stay mid all the time idk how you thought it will be stayongbmid all the time (this is why I said ensage does not have good coders with high level intelligence, only above average with no proper logic , only good scripted who has that level of intelligence I came across in this firm is vicktherock)

    Finally with due respect please be respectful to all the members of this firm who are buying your product @NeeDforKill

    It was not to offense some one. In some times ago, Ensage was Lua hack, with some paid scripts. It have moones combo for all heroes in dota, and meepo script , what have alot functions in auto mode. It farm/push lines, kill enemies, farm/stack jungle, dodge enemy skills, avoid enemy players, makes low hp meepos to go to base, and alot more suff.

    About how it not possible to make , this script for bot of legends, maded by one human.

    oh ok i get it.... thank you for your clarification sir....

  • AdminAdmin Posts: 96

    @Ayush said:

    @NeeDforKill said:

    @Ayush написал:

    @Alvin92 said:
    https://blog.openai.com/dota-2/

    the blog above sums up what the bots does. :doge:

    My Dear Friend, you asking ensage to build something like openai? It's just impossible no one here are intelligent scripters/coders ( no offence just stating obvious , most of the scripters are average) and coding something like openai requires high level intelligence ( which none have ATM in this firm ) dendi himself said it's too strong and arteezy lost 10 games with 0 wins... with ensage you can max win games in 5k mmr ( that too 25%winrate) and arteezy (almost 10k lost to it that too 10 times) now you tell me ensage building something like openAI... lol IMAO never possible , I'd best say it better to improve its support community first (lol),anyways something like openai for ensage is impossible don't get your hopes up with those average coders( no offence again,just stating obvious).

    What you talking on? This is even not "AI" , what it only can play on sf on mid, that's all. Even Lua moones meepo doing alot more.
    It win pro player because they write some stuff like "Ensage" and writed "script" , ofc it win pro players because it can detect autoattacks, particles , timing and have alot faster reaction than human.
    Idk how to possible be so stupid, to think it's impossible to make or it's real ai. Real ai you can google for StarCraft 2 , where devs show how it works and how it analysis game. This one grab all values from game like Ensage.

    Ofc noone will make it on Ensage based, because no one need script what can stay on mid vs one hero.

    I don't know why your reacting in a negative and disrespectful manner, and why your burnt(lol), with all respect I never told u all to make an Ai, I just told "something like Ai is impossible in this forum so don't get your hopes high" sir, and sir I don't think there is a meepo mooned script 2 days ago my subscription ended I will renew in the coming time , but I don't think there is moones meepo script I hope you get your eyes checked ( in a respectful manner ),

    And one more thing, the creators of open ai said they were building a 5 man team by the next TI so they told they wanted to challenge the previous TI winner( team liquid) so obv it won't stay mid all the time idk how you thought it will be stayongbmid all the time (this is why I said ensage does not have good coders with high level intelligence, only above average with no proper logic , only good scripted who has that level of intelligence I came across in this firm is vicktherock)

    Finally with due respect please be respectful to all the members of this firm who are buying your product @NeeDforKill

    Just because they've announced their plans for next year, doesn't mean it'll work as expected. Programming a self-learning AI can have unpredicted results and debugging is almost impossible due to working with a blackbox like system.

    More important: they didn't tell any details. For me it looks like, they used a custom lobby game to feed the bot with more accurate information, which are normally not available.
    For example these events would make programming in Ensage 100% easier, but the server is not sending them in normal games, so we have to create our own ways with particles, animation etc. which is more unreliable and in most cases slower than having a direct callback. Even simple things like the current attack target of an entity is usually not known by the client, because it simply doesn't need to (in comparison to LoL where everything is sent to the client).

    But sure, you can sit here, dream of that perfect self-playing AI and insult our average programmers, who create their assemblies for the community. They should better take their time to develop one year, fulltime together on a project, with a big funding pool and support of valve, and don't update their stupid assemblies :kappa:

  • We are not even 50% of the way to where OpenAI's bot is at this stage (and the DotA portion seems to have been coded in under a year) - I don't know if arguing with customers is going to solve this.

    We need new solutions (not the inaccurate hacked together current solutions) to reliably last hit and win lanes.

    The assemblies we currently have are largely useless at higher levels of DotA because of the base limitations of the tool.

    I think we should look into crowdfunding some additional programmers because we have a very upsetting lack of progress at this time.

  • bemineebeminee Posts: 566
    edited August 2017

    @axischris said:
    We are not even 50% of the way to where OpenAI's bot is at this stage (and the DotA portion seems to have been coded in under a year) - I don't know if arguing with customers is going to solve this.

    We need new solutions (not the inaccurate hacked together current solutions) to reliably last hit and win lanes.

    The assemblies we currently have are largely useless at higher levels of DotA because of the base limitations of the tool.

    I think we should look into crowdfunding some additional programmers because we have a very upsetting lack of progress at this time.

    Just saying you are aware that OpenAI is most likely a multi-million dollars project (it's Elon Musk's company) and it only can play mid shadow fiend with many restrictions (no bottle, no RD, no soul ring, no jungle, 10 minute time limit etc.) meanwhile ensage has many hero assemblies that can perform in any stage of the game without restrictions.

    Let alone the fact that it's not possible because of restrictions @Admin mentioned, Ensage's aim isn't building an AI that completely plays the game itself.

    You say you need new solutions to reliably last hit and win lanes. Admin already explained why there isn't a 100% reliable last hitter. Because you can't access the event that contains unit name that is getting attacked by specific unit. So we do get face positions, dps and many other stuff and make our own prediction that runs separately from dota and last hit with this. Meanwhile that bot most likely has access to that events. (Still doesn't last hit with 100% accuracy since too many variables involved in last hitting)

    If you can play just an example Arc Warden better than Jumpering's script, if you can always know when enemy can see you/your friendly units, if you can know when you get charged - enemy used mirana ulti - enemy smoked and many more stuff I don't need to recall; you are always free to unsubscribe since yeah if you can do all that stuff it's some 15k mmr points shit.

  • I understand that Ensage has a lower budget, but what OpenAI has accomplished in a very short time is far more impressive than anything we have seen so far from this development team or any of the assembly coders.

    Ensage has been in development for 2-3+ years at this point - there has probably been an equal amount of development time on the OpenAI DotA project vs. Ensage.

    OpenAI had a clear, intelligent goal of solving lasthitting, harass, creep aggro abuse, blocking, etc. They consulted with pro players, spent hundreds/thousands of hours on research and ran countless simulations.

    Does Zynox or any assembly creator have any experience in high level DotA games? (excluding server crashers) Why are we not hiring consultants or researching how to best win games reliably?

    I don't really think money should be the excuse here - just seems like poor time management from everyone involved.

  • AyushAyush Posts: 356

    @Admin said:

    @Ayush said:

    @NeeDforKill said:

    @Ayush написал:

    @Alvin92 said:
    https://blog.openai.com/dota-2/

    the blog above sums up what the bots does. :doge:

    My Dear Friend, you asking ensage to build something like openai? It's just impossible no one here are intelligent scripters/coders ( no offence just stating obvious , most of the scripters are average) and coding something like openai requires high level intelligence ( which none have ATM in this firm ) dendi himself said it's too strong and arteezy lost 10 games with 0 wins... with ensage you can max win games in 5k mmr ( that too 25%winrate) and arteezy (almost 10k lost to it that too 10 times) now you tell me ensage building something like openAI... lol IMAO never possible , I'd best say it better to improve its support community first (lol),anyways something like openai for ensage is impossible don't get your hopes up with those average coders( no offence again,just stating obvious).

    What you talking on? This is even not "AI" , what it only can play on sf on mid, that's all. Even Lua moones meepo doing alot more.
    It win pro player because they write some stuff like "Ensage" and writed "script" , ofc it win pro players because it can detect autoattacks, particles , timing and have alot faster reaction than human.
    Idk how to possible be so stupid, to think it's impossible to make or it's real ai. Real ai you can google for StarCraft 2 , where devs show how it works and how it analysis game. This one grab all values from game like Ensage.

    Ofc noone will make it on Ensage based, because no one need script what can stay on mid vs one hero.

    I don't know why your reacting in a negative and disrespectful manner, and why your burnt(lol), with all respect I never told u all to make an Ai, I just told "something like Ai is impossible in this forum so don't get your hopes high" sir, and sir I don't think there is a meepo mooned script 2 days ago my subscription ended I will renew in the coming time , but I don't think there is moones meepo script I hope you get your eyes checked ( in a respectful manner ),

    And one more thing, the creators of open ai said they were building a 5 man team by the next TI so they told they wanted to challenge the previous TI winner( team liquid) so obv it won't stay mid all the time idk how you thought it will be stayongbmid all the time (this is why I said ensage does not have good coders with high level intelligence, only above average with no proper logic , only good scripted who has that level of intelligence I came across in this firm is vicktherock)

    Finally with due respect please be respectful to all the members of this firm who are buying your product @NeeDforKill

    Just because they've announced their plans for next year, doesn't mean it'll work as expected. Programming a self-learning AI can have unpredicted results and debugging is almost impossible due to working with a blackbox like system.

    More important: they didn't tell any details. For me it looks like, they used a custom lobby game to feed the bot with more accurate information, which are normally not available.
    For example these events would make programming in Ensage 100% easier, but the server is not sending them in normal games, so we have to create our own ways with particles, animation etc. which is more unreliable and in most cases slower than having a direct callback. Even simple things like the current attack target of an entity is usually not known by the client, because it simply doesn't need to (in comparison to LoL where everything is sent to the client).

    But sure, you can sit here, dream of that perfect self-playing AI and insult our average programmers, who create their assemblies for the community. They should better take their time to develop one year, fulltime together on a project, with a big funding pool and support of valve, and don't update their stupid assemblies :kappa:

    @Admin sorry if i assulted u in any manner but i want to say one thing you should simply not shift to new webpages and create a new core, like since the original ensage we have shifted to almost 3 webpages and everytime we did so the assemblies start from scratch , instead of that i recommand to stay in one webpage and dont simply create new core so we will have old enough assemblies with very limited bugs and high performance :) , i understand when ensage shifted to joduska there was a need ( cos from normal to reborn) but from joduska to here i didnt understand like if we had old enough assembllies this bot would have been so much better :) anyways thanks for everythomh amin

  • @axischris said:
    I understand that Ensage has a lower budget, but what OpenAI has accomplished in a very short time is far more impressive than anything we have seen so far from this development team or any of the assembly coders.

    Ensage has been in development for 2-3+ years at this point - there has probably been an equal amount of development time on the OpenAI DotA project vs. Ensage.

    OpenAI had a clear, intelligent goal of solving lasthitting, harass, creep aggro abuse, blocking, etc. They consulted with pro players, spent hundreds/thousands of hours on research and ran countless simulations.

    Does Zynox or any assembly creator have any experience in high level DotA games? (excluding server crashers) Why are we not hiring consultants or researching how to best win games reliably?

    I don't really think money should be the excuse here - just seems like poor time management from everyone involved.

    Wait you didn't understand the most important part.

    The reason why Ensage isn't going to be as good as OpenAI or even any bot is because Ensage doesn't have access to "Events", but bot does. You can even see it in normal bot game, if you try to lasthitting against a bot you will lose 70% of the time.

    It's not money. It's not poor time management. It's just that what we are doing here isn't supported by Valve or anyone that play legit Dota.

  • skitzkeskitzke Posts: 30

    @axischris said:
    I understand that Ensage has a lower budget, but what OpenAI has accomplished in a very short time is far more impressive than anything we have seen so far from this development team or any of the assembly coders.

    Ensage has been in development for 2-3+ years at this point - there has probably been an equal amount of development time on the OpenAI DotA project vs. Ensage.

    OpenAI had a clear, intelligent goal of solving lasthitting, harass, creep aggro abuse, blocking, etc. They consulted with pro players, spent hundreds/thousands of hours on research and ran countless simulations.

    Does Zynox or any assembly creator have any experience in high level DotA games? (excluding server crashers) Why are we not hiring consultants or researching how to best win games reliably?

    I don't really think money should be the excuse here - just seems like poor time management from everyone involved.

    I think this topic should be deleted from this forum, How does a certain amount of rules and only an SF mid apply to ensage at all. I don't understand. We as players have more variety than a specific amount of rules. First of all we don't 1v1 mid, unless you're in a very low bracket. Second, the scripts for an AI/Bot can be easily done, but their isn't any need to do it because people actually want to play the game, instead of afk botting mid. And lastly, they said it took them 2 months to make a mid 1v1 with a specific amount of rules? That's not impressive at all, the fact that it can beat 10k mmr players/pros. So what? I don't think any of the players that tried vsing against that bot were actually mid specific players.

    So please delete this topic and put some relevant sh*t on the forums please.

  • axischrisaxischris Posts: 9
    edited August 2017

    Look at what the bot can do - why are we not close to solving any of these mechanics after 3 years of development?

    Imagine if we just had all the current coders/developers scrap their current hero scripts and attack a single problem - maybe a contest with a $100 prize for the winner - I'm sure we could crowdfund something like that every week for a different problem.

    1) Perfect creep blocking - note how the creep blocking is levels above what any creepblock script for Ensage is - a perfect creepblock like this in the offlane has to increase our winrate by .5-1% - not to mention we can do it as many times as we want and with illusions, units, etc. No creepblock script for Ensage has ever been more than 75% reliable with ping and they wouldn't block as effectively as the OpenAI block even at 100% reliability.

    2) Perfect lasthitting - we will get infinite excuses for why all lasthitting scripts are completely useless after 3 years of development, so I'll skip this one to avoid getting a paragraph of arguing

    3) Perfect aggro abuse/creep manipulation - best we've ever gotten is a key to aggro/unaggro (somewhat useful in the sense that we can aggro on a hero halfway across the map instead of our lane opponent) - still need to be abusing this mechanic much more

    4) Perfect orbwalk/harassing - the orbwalk scripts seem to get worse over time. Moones made a very simple orbwalk script on the old Ensage that abused the (patched) fog glitch - if anyone ever used this they would be very unhappy with the state of orbwalkers today.

    5) Perfect spell usage - None of the current scripts compare to the OpenAI Raze usage, however, I would say spellcasting is the one thing Ensage has focused on and probably needs the least attention at this point.

  • To follow up on this - the creepblocking wasn't even machine learning - they literally coded it manually. (https://blog.openai.com/more-on-dota-2/) If Ensage had creepblocking at this level it would instantly be the most overpowered script ever released. What possible Bot API function excuse will we have for this?

    How many high MMR players currently use Ensage? Every YouTube video and Reddit post I see of DotA cheating lately is a 3K MMR Pudge or Skywrath or Techies. Where are the cheaters playing and winning games against pros anymore? I remember years ago that there were several high MMR coders, players etc. (Zanko being a favorite in terms of notoriety) There was a real passion to cheat and win at the highest level which seems to have evaporated over time.

  • bemineebeminee Posts: 566
    edited August 2017

    @BalenDanny said:

    @beminee said:

    @axischris said:
    We are not even 50% of the way to where OpenAI's bot is at this stage (and the DotA portion seems to have been coded in under a year) - I don't know if arguing with customers is going to solve this.

    We need new solutions (not the inaccurate hacked together current solutions) to reliably last hit and win lanes.

    The assemblies we currently have are largely useless at higher levels of DotA because of the base limitations of the tool.

    I think we should look into crowdfunding some additional programmers because we have a very upsetting lack of progress at this time.

    Just saying you are aware that OpenAI is most likely a multi-million dollars project (it's Elon Musk's company) and it only can play mid shadow fiend with many restrictions (no bottle, no RD, no soul ring, no jungle, 10 minute time limit etc.) meanwhile ensage has many hero assemblies that can perform in any stage of the game without restrictions.

    Let alone the fact that it's not possible because of restrictions @Admin mentioned, Ensage's aim isn't building an AI that completely plays the game itself.

    You say you need new solutions to reliably last hit and win lanes. Admin already explained why there isn't a 100% reliable last hitter. Because you can't access the event that contains unit name that is getting attacked by specific unit. So we do get face positions, dps and many other stuff and make our own prediction that runs separately from dota and last hit with this. Meanwhile that bot most likely has access to that events. (Still doesn't last hit with 100% accuracy since too many variables involved in last hitting)

    If you can play just an example Arc Warden better than Jumpering's script, if you can always know when enemy can see you/your friendly units, if you can know when you get charged - enemy used mirana ulti - enemy smoked and many more stuff I don't need to recall; you are always free to unsubscribe since yeah if you can do all that stuff it's some 15k mmr points shit.

    -If you can play just an example Arc Warden better than Jumpering's script -> I do better by 10x, salt salt salt. You say axischris can't do as good as jumpering's arc warden script? Even a 2k mmr kid does better than stacking everything like a brainless monkey.
    -if you can always know when enemy can see you/your friendly units -> You can't know when someones reaction appeals to the fact that you are visible? Ouch,
    -enemy used mirana ulti, such a fast projectile, I need hacks to dodge it.
    - enemy smoked and many more stuff I don't need to recal -> Where is everyone???? smoke????

    1) My legit mmr is probably double the amount of your hacking mmr.
    2) Mirana ulti isn't a projectile :fappa:
    3) Smart players only reveal wards when they can take advantage of it. Oh of course in your crappy bracket people just attack high ground without thinking about ward getting dewarded, my bad :kappajail:
    4) You don't particularly smoke as 5. 3 can smoke and 2 can be shown on map and you won't notice shit.

    And an answer to your shit about "WOOO COPYCATS LEARN TO REVERSE ENGINEER"; stop spitting your shit source 1 knowledge as you are the next hot shit. You are like a script kiddie who just reads stuff over GIVEMEFREEHAXANDTEACHMECODE.COM and talk about them without any sense nor proof. Go back to your "favorite hack" and teach them how to have fully functional OnAttack event as they use similar methods to us for their lasthitting only difference is that their's perform a bit more poorly :kappajail:

  • tawannupinwtawannupinw Posts: 138
    edited August 2017

    @BalenDanny said:

    @tawannupinw said:

    @axischris said:
    I understand that Ensage has a lower budget, but what OpenAI has accomplished in a very short time is far more impressive than anything we have seen so far from this development team or any of the assembly coders.

    Ensage has been in development for 2-3+ years at this point - there has probably been an equal amount of development time on the OpenAI DotA project vs. Ensage.

    OpenAI had a clear, intelligent goal of solving lasthitting, harass, creep aggro abuse, blocking, etc. They consulted with pro players, spent hundreds/thousands of hours on research and ran countless simulations.

    Does Zynox or any assembly creator have any experience in high level DotA games? (excluding server crashers) Why are we not hiring consultants or researching how to best win games reliably?

    I don't really think money should be the excuse here - just seems like poor time management from everyone involved.

    Wait you didn't understand the most important part.

    The reason why Ensage isn't going to be as good as OpenAI or even any bot is because Ensage doesn't have access to "Events", but bot does. You can even see it in normal bot game, if you try to lasthitting against a bot you will lose 70% of the time.

    It's not money. It's not poor time management. It's just that what we are doing here isn't supported by Valve or anyone that play legit Dota.

    What do you know about hacking a game you shitty low life? All you want to do is appeal to shitty monkeys because you are a 8yr kid and your nature is that.

    Oh you are back again kappa . And thanks for such a good criticism kappa

    How was hake doing? Did they implement OpenAI features now? No offense but if they did I would sub to them.



  • From the look of it I think it's pretty much the same here and there.

  • LoaderLoader Posts: 23
    edited August 2017

    @BalenDanny said:
    Once again, I see a salty bunch of shitty developers who can't do shit except simple hello world level crap.

    The event's can be registered to the client by whoever develops ensage if he know hows to properly reverse engineer.

    Example: OnAttack

    Find where the logic of hitting is applied, fire your custom event, send necessary data.

    Also axischris is apparently more intelligent than you stupid developers who are nothing but a bunch of untrained salty kids.

    1. We already have those events but they are only called in local server games, not sure if you know what you are talking about.
    2. Why are you even here? if you hate it so much and everyone is stupid, just leave and mind your own business.

    Im pretty sure you are here to push your own or a competitors agenda because of decreased usage.
    We don't want and deserve your hate! No one does!

    If you want to give constructive criticism you are free to do so, but please provide facts/evidence/info about the problem instead of accusing everyone as stupid or little kid.


    I will give you a week timeout to think about your behavior towards this community and to cool down.

  • AyushAyush Posts: 356

    @h3h3 said:

    @BalenDanny said:
    Once again, I see a salty bunch of shitty developers who can't do shit except simple hello world level crap.

    The event's can be registered to the client by whoever develops ensage if he know hows to properly reverse engineer.

    Example: OnAttack

    Find where the logic of hitting is applied, fire your custom event, send necessary data.

    Also axischris is apparently more intelligent than you stupid developers who are nothing but a bunch of untrained salty kids.

    1. We already have those events but they are only called in local server games, not sure if you know what you are talking about.
    2. Why are you even here? if you hate it so much and everyone is stupid, just leave and mind your own business.

    Im pretty sure you are here to push your own or a competitors agenda because of decreased usage.
    We don't want and deserve your hate! No one does!

    If you want to give constructive criticism you are free to do so, but please provide facts/evidence/info about the problem instead of accusing everyone as stupid or little kid.


    I will give you a week timeout to think about your behavior towards this community and to cool down.

    Haha this is the end message of every discussion.

    "We want constructive criticism."- DUDE any criticism can be constructive if you can take it in that way.

    But frankly the core is good zynox made. Congratz on that, I mean apart from zynox and moones(who are barely on the forum) rest of the coders don't really have good skills to utilize the core and make GOD-LIKE Scripts.

    Most of the scripts are available on other engines(hacks) with very little core or not as advanced as ensage, yet they do the same things ensage does in a far more efficient manner with less or no fps drop.

    So whats the point of core being so advanced if people here are not utilizing the core properly.

    Also feel free to check and evaluate what i said.

    I've been purchasing E# from quite a long time, as i never needed to shift from this but when i did try out the other hacks they were just amazing.

    Right now i use a Private Hack with almost every functionality E# has with a lot better stability and best display fps.

    Since Ensage was the only hack i knew(back in lua time) i bought it later on when it became paid thinking i can trust these guys after the lua scene, but no E# has actually been degrading since then, and have been wasting my money on subscriber ever since, this made me try new hacks and everything just changed all of a sudden.

    Before with Ensage i used to struggle winning matches in a steak now i do it with ez with win streak of more than 15 games.

This discussion has been closed.

Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!