Page 6 of 8 FirstFirst ... 4 5 6 7 8 LastLast
Results 51 to 60 of 73

Thread: [Feedback] Bot Item Build Changes (still WIP)

  1. #51
    Basic Member
    Join Date
    Dec 2012
    Posts
    5
    Has anyone tested what "AggressionFactor" does?

    It only appears in Crystal Maiden and Warlock:
    Code:
    "HeroType"			"DOTA_BOT_PURE_SUPPORT"
    "AggressionFactor"	"0.8"
    "LaningInfo"

  2. #52
    Basic Member
    Join Date
    Dec 2011
    Posts
    1,903
    Quote Originally Posted by jptarrega View Post
    Has anyone tested what "AggressionFactor" does?

    It only appears in Crystal Maiden and Warlock:
    Code:
    "HeroType"			"DOTA_BOT_PURE_SUPPORT"
    "AggressionFactor"	"0.8"
    "LaningInfo"
    There was a thread ages ago about CM being rather suicidal. She'd often run into other heroes on her own with her max aura 0 farm build thinking she could achieve anything, and basically every single hero in the entire game will just outright kill her if she tries that, so the AggressionFactor is there to tone down her aggression. I suppose it just multiplies all her aggressive behaviour values by that number to make her more likely to decide other things.
    I think it's one of the reasons she's a bit more passive in the laning phase nowadays though, while a CM is actually pretty scary during the laning phase.

    Warlock wasn't in at the time this was changed, but I suppose Warlock has it because of how little he can achieve on his own against another hero unless he pops his ultimate.

  3. #53
    Basic Member
    Join Date
    Dec 2012
    Posts
    5
    Quote Originally Posted by Arxos View Post
    There was a thread ages ago about CM being rather suicidal. She'd often run into other heroes on her own with her max aura 0 farm build thinking she could achieve anything, and basically every single hero in the entire game will just outright kill her if she tries that, so the AggressionFactor is there to tone down her aggression. I suppose it just multiplies all her aggressive behaviour values by that number to make her more likely to decide other things.
    I think it's one of the reasons she's a bit more passive in the laning phase nowadays though, while a CM is actually pretty scary during the laning phase.

    Warlock wasn't in at the time this was changed, but I suppose Warlock has it because of how little he can achieve on his own against another hero unless he pops his ultimate.
    Thanks for the info on that :)

    Shouldn't someone make a general thread about modifying the bots builds and such?
    Last edited by jptarrega; 03-23-2013 at 08:40 AM.

  4. #54
    Basic Member
    Join Date
    Dec 2011
    Posts
    1,903
    http://dev.dota2.com/showthread.php?t=76241&page=13
    That's more or less the main thread about it.
    I suppose it has a bit of a deceptive name though as the thread was kind of derailed into bot modding, so a new thread might be a good idea.

  5. #55
    Basic Member
    Join Date
    Mar 2013
    Location
    Russia
    Posts
    27
    Arxos, maybe it would be better if you posted a new thread with a summary info about modifying bot's builds (your tutorial and Darklite's post) and then asked the moderators to make it sticky?
    And also we might make a thread which would contain only tested and approved by community bot item builds and their general info (roles, laning info) to help devs with AI development.

  6. #56
    Basic Member
    Join Date
    Jun 2012
    Location
    London, UK
    Posts
    678
    Quote Originally Posted by Karyoplasma View Post
    Another question: I tried to make Jugger farm but he sells his Midas over the healthpot, quelling blade or stout shield... How can I prevent this from happening? I don't wanna make Midas not sellable. Omni does it fine. He sells his Drums over the Soul Ring, but at least he doesn't keep his QB 50min in.
    This is I think a current bug or missing feature to do with how bots sell items. I don't think they evaluate what they are selling, they simply sell what is the first 'sellable' item in their inventory (at least this is what I have always observed).

    So to avoid that problem as much as possible you have to list their starting items in a specific order, imagining how their slots will fill/empty throughout the match. If you list consumables first, they will get used up early and your first built items will fill earlier slots. If they are marked as sellable they will be sold ahead of even consumables like tangos that haven't been cleared.

    For example on my Omni build I list QB first, since I predict that the consumables listed after will be cleared before, meaning later the QB will be the first item sold guaranteed. The stout shield will be plopped into slot 2/3 most likely (salve or clarity having been used) - and that will be sold next, to make way for the Manta.

    edit: have since seen a match where Omni used no consumables in lane before farming his Midas, went to the shop, sold the QB, and instantly sold the midas to make room for bracer. Yeah.... You can't mark first-bought core items as sellable and rely on them being kept.

    However, there is a limit to how much control you can have doing this, of course. It is only semi-predictable. And not only can weird things happen when they don't use up consumables as you might expect, but later on you might be forced to sell one item before another, even if you'd rather they were in the inventory the other way around - for example selling Midas before Drums.

    If you have sellable branches/stats items you will want to list them first in an inventory.

    Omni and DP items end game.

    Another example is what I did for DP: branches, branches, mantle, tango, clarity. Clarity gets used first - bottle goes in slot 5. Then tango - soul ring goes in slot 4 [edit: you might argue these should be the other way around..]. She builds phase boots (slot 6). Then she has to sell the branches to make room for building Soul Booster. But luckily Bloodstone builds into that so she doesn't sell anything else - if she did it would be the Mantle, so no big loss. When Bloodstone is cleared she builds Linkens in slot 2. Again, could sell Mantle, but doesn't that I have seen. She'll sell it next to build her end game items (and then Soul Ring).

    Basically the bots very much need a way to evaluate items for sale purposes, or we need a way to specify a sale priority
    Last edited by cinephile; 03-25-2013 at 12:59 AM.

  7. #57
    Basic Member
    Join Date
    Dec 2011
    Posts
    1,903
    As cinephile says, all you can do sadly is try and organize the initial purchase in such a manner to avoid these issues, as indeed they will simply sell the first item in their inventory that is marked as sellable, they will gladly sell their Drums over a branch.
    In the end though organizing the inventory won't be able to fix everything, it will only reduce the probability of it occuring, plus it really limits you sometimes.
    For example what's a lot trickier to do is give the bots say 3 trash branches. I personally like to do this a lot, there's really no point leaving gold unspent unless you're bottle rushing or some such, may as well spend everything on trash branches.
    However, it is quite possible that of the say 3 trash branches you buy, throughout the game you get rid of 2 of them, but after the 2nd you complete your Drums. The next thing that happens is that the bots will sell the drums instead of their 3rd branch.

    A good tool to figure out what to sell when would be when it shows up in the item list. Obviously trash items sold at the start of the game are likely to be less valuable than trash items bought later on. However, this is not guaranteed to be so. Even so, it'd be a quick temporary fix as the current method is atrocious.

  8. #58
    Basic Member
    Join Date
    Feb 2013
    Posts
    222
    Quote Originally Posted by Arxos View Post
    http://dev.dota2.com/showthread.php?t=76241&page=13
    That's more or less the main thread about it.
    I suppose it has a bit of a deceptive name though as the thread was kind of derailed into bot modding, so a new thread might be a good idea.
    Haha. It was still a rather good thread though and served a purpose for a time. However the thing that concerns me is that everyone is modding in their own direction and tastes, which is great for amusement purposes but not quite as good for giving feedback since the findings aren't as consistent without everyone working from the same data.

    If we want to make the base bots better at what they do, we honestly need to come more together and work on one bot at a time. Even before that though, the biggest issue I want to resolve personally though is making bots dealing with Viper better (either disabling him altogether, making it so they can deal with him or making him cheat less) and improving their tactical AI (not pushing when losing/shorthanded, getting other strats besides 5 man dota) but that's just me.

  9. #59
    Basic Member DarkLite's Avatar
    Join Date
    Dec 2011
    Location
    Patch Anticipation Station
    Posts
    3,919
    Quote Originally Posted by Peachrocks View Post
    If we want to make the base bots better at what they do, we honestly need to come more together and work on one bot at a time. Even before that though, the biggest issue I want to resolve personally though is making bots dealing with Viper better (either disabling him altogether, making it so they can deal with him or making him cheat less) and improving their tactical AI (not pushing when losing/shorthanded, getting other strats besides 5 man dota) but that's just me.
    That's quite an interesting idea. Focusing on one bot for, say, a week and improving them as much as possible?

    Also, bots durr horribly against Venomancer. I can't help but wonder if they just vastly underestimate the lethality of DoT effects - if they're seeing it as "oh, I'm only taking 50 damage, that's fine" without realising they're going to take that 50 damage for the next 15 seconds...
    Please help keep the forums clean by reporting rude or abusive posts!
    Got a problem? PM me or another moderator!
    Forum Rules


  10. #60
    Basic Member
    Join Date
    Mar 2013
    Location
    Russia
    Posts
    27
    That's quite an interesting idea. Focusing on one bot for, say, a week and improving them as much as possible?
    Consider you've got me in. However, the problem is that we can't focus on 1 bot for one week. I suggest that we take 2 or 3 days for each one (everyone has a lot of his own adjustments, so it could be made quicker).
    Also, bots durr horribly against Venomancer. I can't help but wonder if they just vastly underestimate the lethality of DoT effects - if they're seeing it as "oh, I'm only taking 50 damage, that's fine" without realising they're going to take that 50 damage for the next 15 seconds...
    Not only against Veno - Queen of Pain and all other heroes with DoT effects on their spells would just decimate that poor AI...
    This issue with DoT effects and AI was brought a lot of times. We'll have to wait until it would be fixed.

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •