|
|
rpgyay
Known Hero
|
posted May 17, 2011 10:10 AM |
|
Edited by rpgyay at 11:26, 17 May 2011.
|
|
Thanks again Quantomas for the new release, unfortunately I am at a loss on how to implement NFC Creatures so that the new AI Beta can read them, I used RTMG to place NFC Creatures on the map and when I noticed alot of peasants I figured the AI Beta wasn't implementing them so I manually added some NFC Creatures using H5_RPE_MapEditor and placed them near my city. The NFC Creatures are working correctly when using the RPE_Game just not with the new AI Beta.
|
|
magnomagus
Admirable
Legendary Hero
modding wizard
|
|
posted May 17, 2011 02:02 PM |
|
|
@Quantomas,
I have a few more questions about the AI,
1. How is the AI difficulty improved on hard and heroic difficulty? Does the AI cheat or is it just resource settings?
2. Is your latest build still affected by the following settings in defaultstats, if so what do you recommend?:
<HireHeroesBase>1</HireHeroesBase><HireHeroesPerTown>1</HireHeroesPerTown>
<HireHeroesGlobalLimit>8</HireHeroesGlobalLimit>
3. Also if this is set to 2, then does the AI get double amount of creatures?
<TownCreatureGrowthCoef>1</TownCreatureGrowthCoef>
4. Can you slowdown AI townbuilding here?
<TownBuildPeriodInDays>1</TownBuildPeriodInDays>
5. Are all those threat settings important?
<ThreatsMultiplierEnemys>1</ThreatsMultiplierEnemys>
<ThreatsMultiplierNeutral>2</ThreatsMultiplierNeutral>
<ThreatsMultiplierAllys>3</ThreatsMultiplierAllys>
<ThreatsMultiplierOurs>4</ThreatsMultiplierOurs>
<ThreatsDistanceInDays>4</ThreatsDistanceInDays>
<ThreatsAccComingLimit>15000</ThreatsAccComingLimit>
<ThreatsAccComingCalmingDivisor>1.125</ThreatsAccComingCalmingDivisor>
<ThreatsAccComingCoef>0.6</ThreatsAccComingCoef>
<ThreatsTownInDangerThreshold>500000</ThreatsTownInDangerThreshold>
<ThreatsTownProtectDistanceInDays>1</ThreatsTownProtectDistanceInDays>
6. Is the AI really affected by things like this?:
<ValueOfMinorArtifact>3000</ValueOfMinorArtifact>
<ValueOfMajorArtifact>9000</ValueOfMajorArtifact>
<ValueOfRelicArtifact>12000</ValueOfRelicArtifact>
____________
MMH5.5 Downloads | MMH5.5 Translations | MMH5.5 FAQ
|
|
Quantomas
Responsible
Famous Hero
AI Wizard
|
|
posted May 17, 2011 04:36 PM |
|
|
Hall of Decision Map Design
I would like to invite all talented map makers to a small contest to design the map for the forthcoming Hall of Decision endgame in Heroes V.
This can incorporate all kinds of ideas to offer a truly interesting challenge. I would also like to invite all veteran players who have an indepth understanding of the game to come up with a rough design sketch. If you don't feel you are good at map design but have the right ideas for the map and the gameplay on it, we can accept that and have other people work out and beautify the map itself.
You can find more details here:
Hall of Decision Map Design
@Magnomagus
The AI does not cheat. I personally disabled this creature growth modifier. The threat variables are completely obsolete, the treasure variables are currently still in effect but about to be replaced sooner or later. The same goes for the hero hiring variables.
The difference on the higher difficulty levels is mainly that the AI is less inclined to attack other AI players' heroes but rather go for the human player's heroes. But this is moderated because the AI gives precedence to the impact on the overall power balance in the game, and only attacks human heroes more likely if their is no discernible difference. On the weaker difficulty levels the AI mostly battles against itself first, to give more leeway to the human player.
|
|
magnomagus
Admirable
Legendary Hero
modding wizard
|
|
posted May 17, 2011 04:57 PM |
|
|
Thanks for clarifications. Personally I don't have anything against a cheating AI because I think that's what difficulty settings are for. Providing an exceptional challenge for anyone who is willing to accept it. Many people have been complaining in the past even on heroic the AI wasn't hard enough so I hope your non-cheating AI is up to this task.
____________
MMH5.5 Downloads | MMH5.5 Translations | MMH5.5 FAQ
|
|
Markus404
Tavern Dweller
|
|
posted May 23, 2011 09:29 AM |
|
|
|
A request for Quantomas, it is possible to export the xp table in editable file format ?
|
|
stachnie
Known Hero
|
|
posted May 24, 2011 01:06 PM |
|
|
Hi, Quantomas.
I have downloaded the most recent updates and I tested the 3.1.8 on the Axis of Evil map (as Necropolis, at Hard).
I had no crash at all (Win XP). I had no problems with the 7th creature slot in the special Necro building (and with the building itself), so perhaps the problem was solved also for other town types.
As for AI: both Haven players behaved rationally. They tried to capture poorly defended towns, flagged dwellings and mines etc. However, my Dungeon ally had big problems. All Dungeon heroes got through the one-way portal to the island with the second Dungeon town and they did not manage to go back to flag remaining dwellings and mines, collect artifacts etc. It was impossible to get back through Town Portal but I think they could summon a boat or hire another hero in the first town. At the end their first town was captured by the strongest Haven hero when I was going to win (one Haven player eliminated, the other was going to lose its second town).
Now I test the Crescent Allied map (also at Hard) and my ally seems to be very competent: while I am amassing troops and try to avoid losses, it collected most available artifacts in our starting area.
S.
|
|
Hellglory
Hired Hero
|
|
posted May 26, 2011 05:41 PM |
|
|
Quote:
Hi, Quantomas.
I have downloaded the most recent updates and I tested the 3.1.8 on the Axis of Evil map (as Necropolis, at Hard).
I had no crash at all (Win XP). I had no problems with the 7th creature slot in the special Necro building (and with the building itself), so perhaps the problem was solved also for other town types.
As for AI: both Haven players behaved rationally. They tried to capture poorly defended towns, flagged dwellings and mines etc. However, my Dungeon ally had big problems. All Dungeon heroes got through the one-way portal to the island with the second Dungeon town and they did not manage to go back to flag remaining dwellings and mines, collect artifacts etc. It was impossible to get back through Town Portal but I think they could summon a boat or hire another hero in the first town. At the end their first town was captured by the strongest Haven hero when I was going to win (one Haven player eliminated, the other was going to lose its second town).
Now I test the Crescent Allied map (also at Hard) and my ally seems to be very competent: while I am amassing troops and try to avoid losses, it collected most available artifacts in our starting area.
S.
If I could take some of the blame here, the problem with AI's and one way portals appears to be long lasting. I've done many tests watching the computers bahavior with portals and have found their "portal lust" to be a bit extreme. They will dive through one way portals be it Quantomas' AI or the basic one. The two sets of portals givin to the Haven AI's had to be carefully placed so they understood them as an extension of their territories to avoid portal foolishness (to great success) but two ways are still easier to handle for the AI's. I am surprised to hear that the dungeon AI acted as foolishly as it did. I usually did my tests while set on the dungeon side assuming the what I saw in the necro would be roughly mimiced by the dungeon player should he be controlled by an AI, who did indeed take their whole island and fight effectively. However this sounds not to be the case for the dungeon. The purpose of the one way axis portals is 2 fold. To maintain a strong defense on the main island (a two way would add an attack angle for an already dangerous enemy) while making the secondary town isolated and difficult location to hold. Infact, there are no shipyards on the secondary Axis isles and the only way off is to summon an existing ship. This is an easy thing for human players but the AI doesn't seem to understand this quite as well and does not moderate the number of heroes it sends on this one way trip. I believe the AI decides its moves based on building proximity. The dungeon secondary is rather close and likely tempts the dungeon AI to portal dive moreso than the necro AI. I have some ideas and how to improve the dungeons decisions by altering the environment. Currently I'm attempting to turn the rivers to sea for asthetics per Magnomagus' request. This is something that's turning out to be a bit time consuming as I may have to terraform nearly all of the terrain to do this neatly. With this patch I will couple some alterations to dissuade the dungeon/necro AI from portal diving.
However, I shouldn't underestimate Quantomas and perhaps he might have some solutions to improve the AI's decision making when it comes to one way portals.
|
|
Quantomas
Responsible
Famous Hero
AI Wizard
|
|
posted May 27, 2011 02:06 PM |
|
|
Quote:
Quote:
Hi, Quantomas.
I have downloaded the most recent updates and I tested the 3.1.8 on the Axis of Evil map (as Necropolis, at Hard).
I had no crash at all (Win XP). I had no problems with the 7th creature slot in the special Necro building (and with the building itself), so perhaps the problem was solved also for other town types.
As for AI: both Haven players behaved rationally. They tried to capture poorly defended towns, flagged dwellings and mines etc. However, my Dungeon ally had big problems. All Dungeon heroes got through the one-way portal to the island with the second Dungeon town and they did not manage to go back to flag remaining dwellings and mines, collect artifacts etc. It was impossible to get back through Town Portal but I think they could summon a boat or hire another hero in the first town. At the end their first town was captured by the strongest Haven hero when I was going to win (one Haven player eliminated, the other was going to lose its second town).
Now I test the Crescent Allied map (also at Hard) and my ally seems to be very competent: while I am amassing troops and try to avoid losses, it collected most available artifacts in our starting area.
S.
If I could take some of the blame here, the problem with AI's and one way portals appears to be long lasting. I've done many tests watching the computers bahavior with portals and have found their "portal lust" to be a bit extreme. They will dive through one way portals be it Quantomas' AI or the basic one. The two sets of portals givin to the Haven AI's had to be carefully placed so they understood them as an extension of their territories to avoid portal foolishness (to great success) but two ways are still easier to handle for the AI's. I am surprised to hear that the dungeon AI acted as foolishly as it did. I usually did my tests while set on the dungeon side assuming the what I saw in the necro would be roughly mimiced by the dungeon player should he be controlled by an AI, who did indeed take their whole island and fight effectively. However this sounds not to be the case for the dungeon. The purpose of the one way axis portals is 2 fold. To maintain a strong defense on the main island (a two way would add an attack angle for an already dangerous enemy) while making the secondary town isolated and difficult location to hold. Infact, there are no shipyards on the secondary Axis isles and the only way off is to summon an existing ship. This is an easy thing for human players but the AI doesn't seem to understand this quite as well and does not moderate the number of heroes it sends on this one way trip. I believe the AI decides its moves based on building proximity. The dungeon secondary is rather close and likely tempts the dungeon AI to portal dive moreso than the necro AI. I have some ideas and how to improve the dungeons decisions by altering the environment. Currently I'm attempting to turn the rivers to sea for asthetics per Magnomagus' request. This is something that's turning out to be a bit time consuming as I may have to terraform nearly all of the terrain to do this neatly. With this patch I will couple some alterations to dissuade the dungeon/necro AI from portal diving.
However, I shouldn't underestimate Quantomas and perhaps he might have some solutions to improve the AI's decision making when it comes to one way portals.
The problem is a bit more involved. Currently, the AI has a fixed hero hiring limit, something like the number of towns owned + 2.
Also, if the AI perceives something of value behind the one-way portal, it will send heroes there. With a limited lookahead range, it will not note that the return path is very long and probably problematic. Even with the strategic endstage, planned until 3.2, and an extended lookahead estimation of 2/3/4 weeks, the problem is only alleviated, not resolved. For example, the AI may go for the second town behind the castle pretty early, while a real threat to its first town will materialize not earlier than in week 5.
The AI does summon boat. We have a reference map for testing this extensively, Island Hoppers. There are a lot of small islands with a stat boost. I observed the AI summoning boats if its own boat was summoned away earlier. So the problem would be rather that there are no boats to summon or it doesn't have the spell.
The solution to this is (a) to have the AI do full return path checks, (b) dismiss secondary heroes that are stuck and hire new ones, or (c) special logic that a remote hero can request a secondary hero to build a ship in some other place. Once I get to implement the full algebraic factorization for the strategic end stage (after 3.2) the AI will have all these capabilities automatically.
I will think about something to have a workaround in one of the next builds. If you like, we can include Axis of Evil in the set of our EE reference maps. It appears to be a good test scenario how the AI deals with portals and asymmetric return paths. The concept of the reference maps is that they aid us to improve the AI. For example on Dread Realms, there is a problem with heroes awaiting reinforcements on subterranean gate exit cells. Most players will not notice it, because the Heroes will eventually move on, but that's another issue I have to look into.
For the time being, I suggest a simple workaround on the map. Place a shipyard on the distant island. Or alternatively another one-way portal to a remote staging area, if that is your main intention, probably connected to a long tunnel to be traveled first. Or give the AI an incentive to build ships, so that boats are available for summoning. Hope that helps. 
|
|
Hellglory
Hired Hero
|
|
posted May 27, 2011 09:31 PM |
|
|
Some observations of mine of the AI on this map and general logistical behavior:
Originally there were 8 shipyards on Axis of Evil (one per island) and the Haven AI's created ships liberally and would begin to pepper the coasts with them mid game. Things would quickly get to the point ships were of no value. I removed 4 shipyards and found ships to be more reasonable in availability. None of the AI's have any great difficulty getting ahold of ships. Now in the early game there is a "summon war" with the first wave of ships but lack of availability quickly gives rise to another wave of manufactured ships and soon that conflict (a fun one) is over. Anyone who wants a ship simply walks to the shore and summons one. Now deeper into this we get to the portals. I noticed the AI's stupidity with portals quite early in my molding of Axis of Evil and in these early stages I removed all concept portals to keep the AI from acting foolishly. After I began to learn a bit more of the AI's behaviors from dozens of long tests, I noticed that only vast distance could finally reduce the AI's portal fever. Even still one way portals were wildly hard to manage for the AI and would not think clearly about concept of being "stranded" or ultra distant from reinforcements and more pressing treasure hunting at home. Town seems to be the driving motivation on this particular map for the AI to enter portals and travel great distances from the capital. The outlaying dungeon town is very near the coast to were you can almost glance it from the across the channel. However, the one way portal is clear across the continent, and same for necromancer as well. Now I've watched the necromancer many times and he appears to act intelligently about moving to his outlaying town about the time he's consumed most of this home territory. His secondary town is quite far away... So here's what I think is happening with the dungeon AI (total theory now). The dungeon AI looks across the water and thinks: "Hmm, that town is fairly close. I'll plot a course there as if the water doesn't matter, as if it was just a couple days journey and hugging my territory perimeter." If the water wasn't there it would indeed be a smart move. Hopefully some of this information is helpful.
I do have some ideas on how to adapt the map to work around this behavior. Namely adding more dangerous guards to the one way portals and secondary axis towns so when the AI looks ahead, it will see death. At that point the AI should reconcentrate on its home territory long enough to consume it and heavy use of the one way will happen at a time where ships are to be summoned easily. This will make the game a touch harder for the axis but it will only delay acquisition of the town 2-3 weeks. The last resort would be to change the one ways to the more intelligently used two ways. Seeing as the one ways add the flavorful dimensions of home territory protection and isolation of the expansion town, I likely won't do this. Also, Axis of Evil is played at its best with a friend and of course a human being will understand this. That doesn't mean I won't stop until the AI provides a decent ally as I often play this single player too.
Finally I would be pleased to contribute Axis of Evil to the project. I look forward to a more intelligent, challenging AI and if this map helps I can't see a reason to refuse.
|
|
magnomagus
Admirable
Legendary Hero
modding wizard
|
|
posted May 28, 2011 04:05 PM |
|
|
@Quantomas, would you support adding new battlefield shapes and sizes? I' m also interested in adding special tiles for cover or special positions for range/melee and magic units (like disciples 3).
____________
MMH5.5 Downloads | MMH5.5 Translations | MMH5.5 FAQ
|
|
Quantomas
Responsible
Famous Hero
AI Wizard
|
|
posted May 28, 2011 04:51 PM |
|
|
Quote:
@Quantomas, would you support adding new battlefield shapes and sizes? I' m also interested in adding special tiles for cover or special positions for range/melee and magic units (like disciples 3).
The magic elements will provide you with this capability. We can also make the battlefields variable. I have a couple of ideas in this direction, for example adding special combat arenas for the fight in a garrison, but in general for the combat arenas (and the magic within) corresponding much more strongly to the surrounding environement (where the battle takes place).
@Hellglory
Cheers, I will include Axis of Evil in our set of reference maps once I have figured out a workaround what the hero should do regarding one-way portals. Probably implementing the strategic end stage should be fair enough.
|
|
Hellglory
Hired Hero
|
|
posted June 04, 2011 05:50 AM |
|
|
|
Would it be a waste of time to test the latest beta build in multiplayer?
|
|
Quantomas
Responsible
Famous Hero
AI Wizard
|
|
posted June 04, 2011 12:29 PM |
|
|
Quote:
Would it be a waste of time to test the latest beta build in multiplayer?
Don't waste your time on this. Multiplayer except hotseat is broken in the current build.
|
|
Quantomas
Responsible
Famous Hero
AI Wizard
|
|
posted June 05, 2011 04:16 PM |
bonus applied by Galaad on 28 Aug 2017. |
|
Took a good deal longer in the end, but we have 3.1c now.
This new build is focused on gameplay enhancements and new options to customize the game to your liking. We have now an adjustable lookahead range, neutral champions who defend creatures and treasure vaults, artifact guards that you have to fight, taxes for a more measured game pace, and of course the eight skills enhancement. You will see, it can be quite useful to fend off the neutral champions.
I will provide more details soon. In the meantime, check out the readme.
I will also write a note who deserves credit for these upgrades. Watch this space. 
@Magnomagus
There was zero time to produce additional builds and do additional test rounds for having the hero class and skill fix included. I wouldn't chance this with 3.1c anyhow. From the error reports, and what I learned during the work on the neutral champions skills, I think the cause of the crash might be the .exe probing the additional 999 skill slots. That the crashes happened right after combat, and random at times, supports this, it might be just the game rolling random skills for level-up. We have to find a better solution for this.
Current stable build:
3.1c public release (7.8 MB)
Latest beta build:
3.1.8 beta release (11.0 MB)
Project website:
Heroes 5.5 - Eternal Essence
edit: it should read additional 999 skill slots, the suspected cause
|
|
magnomagus
Admirable
Legendary Hero
modding wizard
|
|
posted June 05, 2011 05:17 PM |
|
|
Quote:
There was zero time to produce additional builds and do additional test rounds for having the hero class and skill fix included. I wouldn't chance this with 3.1c anyhow. From the error reports, and what I learned during the work on the neutral champions skills, I think the cause of the crash might be the .exe probing the additional skill slots. That the crashes happened right after combat, and random at times, supports this, it might be just the game rolling random skills for level-up. We have to find a better solution for this.
I don't understand, I have never had any crashes with the class extention (they were caused by ncf). If the crashes you encountered are caused by the skill extension then why did you implement the additional skill slots and not the classes?
____________
MMH5.5 Downloads | MMH5.5 Translations | MMH5.5 FAQ
|
|
rpgyay
Known Hero
|
|
posted June 05, 2011 10:27 PM |
|
Edited by rpgyay at 22:27, 05 Jun 2011.
|
|
Wow this is interesting, I look forward to testing this out more.
|
|
Quantomas
Responsible
Famous Hero
AI Wizard
|
|
posted June 05, 2011 11:02 PM |
|
|
First, I would like to thank a couple of people who helped me:
Many thanks to Cepheus, who did the work on the interface that brought you the eight skills enhancement. Cepheus did more than just providing an interface, he systematically analyzed how the interface works (the game database really) and meticulously adapted the UI elements. It really needs to be stressed that this is the true value of the work. I believe with the understanding Cepheus has now about the UI and what I gleaned from it in return, it also helped me to design the interface for the new options, we can tackle almost anything on that front, even if our understanding is not complete yet. I will write more to that topic in the next days.
Cheers and thanks to Markkur, who did a good deal of awesome map work to get the Hall of Decision started. It helped me a good deal, to have other people working on other fronts, instead of having the feeling of carrying the whole burden. 
Special thanks to TheBaron, who not only created a most inspiring backdrop for the EE campaign (more on this soon) but also helps us as a community to get our heads together. This is important in many ways, if we really want this project to fly.
I will write more to your posts in the other threads in the coming days. I am a bit tired today and having a clean head will help.
|
|
rpgyay
Known Hero
|
|
posted June 06, 2011 09:13 AM |
|
|
|
Is it possible to come up with a version more compatible with the RPE? The NCF support isn't a problem for me since I don't use them (atleast yet), I think support for the classes is all that is really needed? With the 3.18 beta release I never used NCF and never had any crashes.
|
|
Quantomas
Responsible
Famous Hero
AI Wizard
|
|
posted June 07, 2011 02:38 AM |
|
|
Quote:
Is it possible to come up with a version more compatible with the RPE? The NCF support isn't a problem for me since I don't use them (atleast yet), I think support for the classes is all that is really needed? With the 3.18 beta release I never used NCF and never had any crashes.
Sure. This has been stated already multiple times. But don't expect this be done today, or even next month.
There is a ton of tasks:
- multiplayer is not working
- townscreen and other game database research
- first steps for an EE campaign
- Hall of Decision
- fixing bugs in the combat arena, failing perks, damage calculation
- work on the AI which is missing its strategic end stage
- lots of other tasks listed in the developer diary
You have no idea how much work it really is. Technically I spent three months to improve merely the .exe to surpass 3.1 vanilla, from a codebase that was substandard, hardly any work on the AI at all. What do you think Black Hole will give you for an estimate, if you asked them to implement all the advanced options included in 3.1c? My guess is they would say something like, ETA 6 months, if they feel optimistic. All things considered, I would say I provided a development effort worth roughly half a million euro on maintaining H5. It gained me nothing, my focus, what ultimately is my benefit is the AI, but nothing really on this account.
The next upgrade will focus on the AI, finally.
Treat 3.1.8 as a gift. Something dedicated to the RPE project.
You can use it to play with RPE if you like, and on other times play with the latest enhancements of EE without RPE. Is that so bad? And one day, there might eventually be a version that incorporates both.
Don't make it sound, like this is the next thing that has to be done.
I regret that my wizard lab assistant made fun of Hero Psathka. Somehow, punishment always has a tendency to hit the wrong ones, it's particularly sad if it hits the polite and mild mannered folks.
But if you keep asking, I will not be able to constrain my wizard lab assistant any longer.
Last warning, no quarters given.
|
|
rpgyay
Known Hero
|
|
posted June 07, 2011 03:07 AM |
|
|
|
Sorry Quantomas, I know you are busy with alot of other things. I wasn't sure if modifying the new build was an easy task but apparently it's not, keep up the good work and I will look forward to a future build that supports RPE.
|
| |
|
|
This thread is pages long: 
