|
|
Warmonger
Promising
Legendary Hero
fallen artist
|
posted June 17, 2011 07:43 PM |
|
|
Quote: On a separate note: Does the AI plan strategically? That is, if it thinks that inferno and dungeon are its greatest threats, would it consider picking up Protection for its hero?
Where was the last time you saw AI doing such a thing?
____________
The future of Heroes 3 is here!
|
|
Quantomas
Responsible
Famous Hero
AI Wizard
|
posted June 17, 2011 08:08 PM |
|
|
Quote:
Quote: On a separate note: Does the AI plan strategically? That is, if it thinks that inferno and dungeon are its greatest threats, would it consider picking up Protection for its hero?
Where was the last time you saw AI doing such a thing?
The strategic end stage I am currently working on allows for the AI to customize its strategies. It will definitely be able to select the best strategic targets, factions it pursues and so on with this.
The algebraic factorization, later on, will allow this in full, that the AI goes out to get the right skills, spells, creature lineup and so on.
Right now, it's an advantage the human player still has, to skill better and be more adaptable. But it's direly needed, because the AI has many other strengths, it's insanely good at counting (like all comps) to a perfection humans just can't match. For example it can count the exact movement cost from portal A to town B.
What also needs to be done is to align the adventure map and combat AI, currently it is nearly two distinct entities, with the creature power value and similar values the only link between the two. Ultimately, it should be that the AI does full combats and makes proper assessments before the battle, and choices along the road, to optimize its chances.
|
|
magnomagus
Admirable
Legendary Hero
modding wizard
|
posted June 17, 2011 09:57 PM |
|
|
Quote: Ultimately, it should be that the AI does full combats and makes proper assessments before the battle, and choices along the road, to optimize its chances.
Aren't you afraid in the end all this is going to make the AI slow again, I think playability should be the main goal. Could you explain for example why the H3 AI is so fast while still satisfying for a lot of players?
|
|
Quantomas
Responsible
Famous Hero
AI Wizard
|
posted June 18, 2011 12:26 AM |
|
|
Quote:
Quote: Ultimately, it should be that the AI does full combats and makes proper assessments before the battle, and choices along the road, to optimize its chances.
Aren't you afraid in the end all this is going to make the AI slow again, I think playability should be the main goal. Could you explain for example why the H3 AI is so fast while still satisfying for a lot of players?
The H3 AI is technically 3 days lookahead range (the EE AI has minimum 4 on easy) plus some form of strategic end stage. That it works rather brilliantly is a sheer insane amount of (or genius) balancing of the AI core values. I would have liked to have met the programmer who created this masterpiece. It requires a clean mathematical concept, similar to the algebraic factorization that is planned here, but while they balanced core values only we will do much more.
No, it's not going to slow the AI down, it rather adds intelligent behaviour, use of town portals and more in a highly optimized way.
|
|
Jukhta
Hired Hero
|
posted June 18, 2011 05:17 AM |
|
|
Quote:
Quote:
Quote: On a separate note: Does the AI plan strategically? That is, if it thinks that inferno and dungeon are its greatest threats, would it consider picking up Protection for its hero?
Where was the last time you saw AI doing such a thing?
The strategic end stage I am currently working on allows for the AI to customize its strategies. It will definitely be able to select the best strategic targets, factions it pursues and so on with this.
The algebraic factorization, later on, will allow this in full, that the AI goes out to get the right skills, spells, creature lineup and so on.
Right now, it's an advantage the human player still has, to skill better and be more adaptable. But it's direly needed, because the AI has many other strengths, it's insanely good at counting (like all comps) to a perfection humans just can't match. For example it can count the exact movement cost from portal A to town B.
What also needs to be done is to align the adventure map and combat AI, currently it is nearly two distinct entities, with the creature power value and similar values the only link between the two. Ultimately, it should be that the AI does full combats and makes proper assessments before the battle, and choices along the road, to optimize its chances.
I'll be looking forward to this.
|
|
bingdevil
Tavern Dweller
|
posted June 20, 2011 04:34 AM |
|
|
at first thank you for make such AI mods
but it has somes important BUGS
ESP the AI can not get boat or summon boat
I try many times at expert difficult .
____________
|
|
bingdevil
Tavern Dweller
|
posted June 20, 2011 06:02 AM |
|
|
at first thank you for make such AI mods
I found somes BUGS
1. the AI never use the "surrender" , th h3 h4"s AI often use it at the battles
2. the AI is very very low intesting pick up the treasure which is guarded by Neutrals. even the Neutral threat level is low.
I hope my suggest can make your MODS much better and much clever .
____________
|
|
Steiner
Tavern Dweller
|
posted June 20, 2011 11:09 AM |
|
|
Since I installed this AI enhancement I have rediscovered HoMM 5 for myself virtually from scratch! I always loved playing big maps and epic battles of Heroes 3, but with sloppy AI from the designers and unbearable turn times I had to turn away from this reincarnation of classics.
But with your help I'm now playing it again and happy, great work guys!
____________
|
|
mfdoom
Hired Hero
|
posted June 20, 2011 02:44 PM |
|
|
Quote: Since I installed this AI enhancement I have rediscovered HoMM 5 for myself virtually from scratch! I always loved playing big maps and epic battles of Heroes 3, but with sloppy AI from the designers and unbearable turn times I had to turn away from this reincarnation of classics.
Me 2.QT should be put on payroll by Ubi.
|
|
Fester
Adventuring Hero
|
posted June 20, 2011 10:20 PM |
|
Edited by Fester at 22:21, 20 Jun 2011.
|
This is regards to the computer fleeing.
Of course the computer should flee if it feels overpowered but the late games I´ve had it flees instantly which is rediculus because I think it should stick around and evaluate if it can mess up the opponents army a little before taking of.
Thats what I do when I meet a beefy army. i evaluate how many turns I can take a beating then do as much damage to the opponent as I possibly can before fleeing.
I think it strategicly important for the A.I. NOT to instantly flee because otherwise one can make it loose many armies for nothing simply by "scaring" them off
|
|
Quantomas
Responsible
Famous Hero
AI Wizard
|
posted June 24, 2011 12:15 AM |
bonus applied by Galaad on 28 Aug 2017. |
|
We have now 3.1d which streamlines the AI optimization and contains a number of important fixes. It contains some additional scripting support that was missing in the official 3.1 vanilla, and is useful for maps that have complex scripting to tell a story. For example, Ultima Classic tells now a story from the perspective of your chosen faction, instead of mangling the story messages.
3.1d doesn't include the strategic end stage yet. That's my current focus and it shouldn't take long before we have an AI that uses town portal spells and plans intelligently ahead. Once this is done, you need to put on warm clothes and it will be tough to beat the AI opponents.
Current stable build:
3.1d public release (7.8 MB)
Latest beta build:
3.1.9 beta release (11.0 MB)
Project website:
Heroes 5.5 - Eternal Essence
---
@Fester
It shouldn't. If it does it's a bug. Can you test with the latest build?
|
|
Hellglory
Hired Hero
|
posted June 24, 2011 01:46 AM |
|
|
Did you make improvements to the AI with one way portals in the last patch increment or this one? From the few games I've played with the last version, the AI appeared to act more intelligently with one way portals. I raced to capture my territory and the AI actually beat me for the first time.. ever. That's without leaping through the one way too early. Perhaps it was the changes I made to the map, I'm not sure which. Well done if so. Also, they seem to still act foolishly with whirlpools, at times going back and forth for no reason. Does the AI lose units when they do? If so they shouldn't. Seems any human player can divide a stack of 2 units in the first army slot and lose 1 level 1 each passage. I'll have to start a new game with this version.
About axis of evil development, which I will keep terse, I've lost a 700dollar video card via a coolant leak (bfg out of business? great... lost a lifetime warranty) and will be out of action for a month or two whilst I build a new machine and likely sell this one. My backup is painfully slow with the editor on large maps like axis. The good news is most of the structure and balance is complete and the only serious work I had planned was aesthetic. I did decide against the adding of ocean for the time being, it disrupts ice placement appearance and waves go against the suggested winds of NE by flowing a steady south. I'll be around... good luck Quan.
|
|
mfdoom
Hired Hero
|
posted June 24, 2011 02:21 AM |
|
|
Quote:
Current stable build:
[URL=http://www.bonddisc.com/ref/H5AI_31.zip]3.1d public release[/URL] (7.8 MB)
Jesus dude.The damn loading bar is flying and i have a mere 5200.
|
|
rpgyay
Known Hero
|
posted June 24, 2011 03:04 AM |
|
|
Wow the computer turns are going alot faster, whatever you did keep up the great work.
|
|
Quantomas
Responsible
Famous Hero
AI Wizard
|
posted June 24, 2011 11:30 AM |
|
|
Quote: Did you make improvements to the AI with one way portals in the last patch increment or this one? From the few games I've played with the last version, the AI appeared to act more intelligently with one way portals. I raced to capture my territory and the AI actually beat me for the first time.. ever. That's without leaping through the one way too early. Perhaps it was the changes I made to the map, I'm not sure which. Well done if so. Also, they seem to still act foolishly with whirlpools, at times going back and forth for no reason. Does the AI lose units when they do? If so they shouldn't. Seems any human player can divide a stack of 2 units in the first army slot and lose 1 level 1 each passage. I'll have to start a new game with this version.
About axis of evil development, which I will keep terse, I've lost a 700dollar video card via a coolant leak (bfg out of business? great... lost a lifetime warranty) and will be out of action for a month or two whilst I build a new machine and likely sell this one. My backup is painfully slow with the editor on large maps like axis. The good news is most of the structure and balance is complete and the only serious work I had planned was aesthetic. I did decide against the adding of ocean for the time being, it disrupts ice placement appearance and waves go against the suggested winds of NE by flowing a steady south. I'll be around... good luck Quan.
It's a pity that we don't have a proper budget that we could at least have the people who contribute to the project give proper hardware.
What I am normally doing is to go for a budget multicore CPU, plenty of RAM and an excellent mainboard that allows to configure the memory bus flexibly. This way you can build your PC pretty cheap. Add to it a powerful graphics card and the Zalman coolers (CPU and GPU, metal no liquids), and you end up with a very quiet system with ample potential for overclocking.
My test machine is just a 4200. I think many devs make a mistake testing only on a powerful rig because this way they miss issues that are important for many folks.
Regarding the AI, it indeed does know about the whirlpool trick (which you can learn from tavern rumours) and does not loose forces. The AI tends to spend its surplus time (if it can reach its intended target in the next turn anyhow) in whirlpools or doing some teleporting.
The real improvement regarding the one way portals comes with the strategic end stage soon.
|
|
magnomagus
Admirable
Legendary Hero
modding wizard
|
posted June 24, 2011 12:01 PM |
|
|
On the subject of hardware, can your AI profit from a quad core CPU?
|
|
Quantomas
Responsible
Famous Hero
AI Wizard
|
posted June 24, 2011 12:07 PM |
|
|
Quote: On the subject of hardware, can your AI profit from a quad core CPU?
Yes, but it requires additional logic to implement.
|
|
magnomagus
Admirable
Legendary Hero
modding wizard
|
posted June 24, 2011 03:23 PM |
|
|
So you haven't implemented it yet?
|
|
Psatkha
Promising
Known Hero
H5 MODder
|
posted June 25, 2011 12:02 AM |
|
|
Current stable version "D" do not support NCF yet
|
|
Hellglory
Hired Hero
|
posted June 25, 2011 06:46 PM |
|
|
Quote:
My test machine is just a 4200. I think many devs make a mistake testing only on a powerful rig because this way they miss issues that are important for many folks.
This is true, but what is the alternative for designers who are looking to be more efficient with their time? Even testing on a slower machine does lengthen development. Another comment you made comes to mind, about pointing out the small fires AoE had for the volcanic vents were a heavy drain on resources. I was blind to any difference in performance, with or without them. Regardless, I find it more useful to have the utmost alacrity when doing something as time consuming as designing a map. To amend to that, I will keep a keen ear out for those who do point out issues specific to performance.
|
|
|