Heroes of Might and Magic Community
visiting hero! Register | Today's Posts | Games | Search! | FAQ/Rules | AvatarList | MemberList | Profile


Age of Heroes Headlines:  
5 Oct 2016: Heroes VII development comes to an end.. - read more
6 Aug 2016: Troubled Heroes VII Expansion Release - read more
26 Apr 2016: Heroes VII XPack - Trial by Fire - Coming out in June! - read more
17 Apr 2016: Global Alternative Creatures MOD for H7 after 1.8 Patch! - read more
7 Mar 2016: Romero launches a Piano Sonata Album Kickstarter! - read more
19 Feb 2016: Heroes 5.5 RC6, Heroes VII patch 1.7 are out! - read more
13 Jan 2016: Horn of the Abyss 1.4 Available for Download! - read more
17 Dec 2015: Heroes 5.5 update, 1.6 out for H7 - read more
23 Nov 2015: H7 1.4 & 1.5 patches Released - read more
31 Oct 2015: First H7 patches are out, End of DoC development - read more
5 Oct 2016: Heroes VII development comes to an end.. - read more
[X] Remove Ads
LOGIN:     Username:     Password:         [ Register ]
HOMM1: info forum | HOMM2: info forum | HOMM3: info mods forum | HOMM4: info CTG forum | HOMM5: info mods forum | MMH6: wiki forum | MMH7: wiki forum
Heroes Community > Heroes 5 - Modders Workshop > Thread: New (unofficial) patch for ToE / AI mod
Thread: New (unofficial) patch for ToE / AI mod This thread is 46 pages long: 1 2 3 4 5 ... 10 20 30 40 ... 42 43 44 45 46 · «PREV / NEXT»
Greenpakto
Greenpakto


Adventuring Hero
posted June 15, 2020 07:03 PM

magnomagus said:
I'm sorry to hear that greenpakto, but that slot did contain some of the highly overpowered skills that needed to be removed, to prevent forcing classes into the same buildup every time.

I thought I removed the slot from the UI, but possibly textures gets overwritten by another mod/map

theoretically 24 new balanced skills could be added in a reinstated 5th slot, but that faces lots of technical difficulties and complex skillwheel adjustment


I dont want these skills to be overpowered. But 24 new balanced skills would be AMAZING. The last abilites are highly missed.

 Send Instant Message | Send E-Mail | View Profile | Quote Reply | Link
thGryphn
thGryphn


Promising
Famous Hero
posted June 15, 2020 09:03 PM

Greenpakto said:
magnomagus said:
I'm sorry to hear that greenpakto, but that slot did contain some of the highly overpowered skills that needed to be removed, to prevent forcing classes into the same buildup every time.

I thought I removed the slot from the UI, but possibly textures gets overwritten by another mod/map

theoretically 24 new balanced skills could be added in a reinstated 5th slot, but that faces lots of technical difficulties and complex skillwheel adjustment


I dont want these skills to be overpowered. But 24 new balanced skills would be AMAZING. The last abilites are highly missed.


You say you don't want skills to be overpowered but also add that the last abilities (which were originally overpowered) are highly missed (I assume this last part is not a generalization). Sounds like contradiction...

Plus, I don't understand why you miss a single skill slot, while you are currently equipped with so many perks and skills that are way more diverse than the original. It's like a person yearning for a 51st car... Yeah, it would be very neat to have one more skill slot (as long as it's balanced) but it's not like the game is lacking anything because they don't exist now.


 Send Instant Message | Send E-Mail | View Profile | Quote Reply | Link
Greenpakto
Greenpakto


Adventuring Hero
posted June 15, 2020 09:29 PM
Edited by Greenpakto at 21:43, 15 Jun 2020.

thGryphn said:
Greenpakto said:
magnomagus said:
I'm sorry to hear that greenpakto, but that slot did contain some of the highly overpowered skills that needed to be removed, to prevent forcing classes into the same buildup every time.

I thought I removed the slot from the UI, but possibly textures gets overwritten by another mod/map

theoretically 24 new balanced skills could be added in a reinstated 5th slot, but that faces lots of technical difficulties and complex skillwheel adjustment


I dont want these skills to be overpowered. But 24 new balanced skills would be AMAZING. The last abilites are highly missed.


You say you don't want skills to be overpowered but also add that the last abilities (which were originally overpowered) are highly missed (I assume this last part is not a generalization). Sounds like contradiction...

Plus, I don't understand why you miss a single skill slot, while you are currently equipped with so many perks and skills that are way more diverse than the original. It's like a person yearning for a 51st car... Yeah, it would be very neat to have one more skill slot (as long as it's balanced) but it's not like the game is lacking anything because they don't exist now.




No its not a contradiction. You just misunderstand me.

Like I said, I want the FEATURE of ultimate abilities to return in a balanced form. The old abilities can be deleted and replaced with completely different ones. I just love having a unique goal to hunt for the end game, and one that is different for almost EVERY character.

PS: I understand you like it the way it is, but in my opinion having ultimate abilities makes the game way more fun to replay and removing that was the biggest flaw of 5.5 (the ONLY flaw in fact). Anyway, I have derailed the thread enough.

 Send Instant Message | Send E-Mail | View Profile | Quote Reply | Link
Velote
Velote


Hired Hero
posted June 15, 2020 09:48 PM
Edited by Velote at 22:11, 15 Jun 2020.

Quantomas indeed has an extremely vague vision of 5.5's inner workings... like if he had played just a few games and didn't look deep under the hood.

funked15 said:
but the only reason people are still here to notice is because of MM5.5.

Ex-actly.

magnomagus said:
But if it comes to your approval, I'm completely open to deactivating stuff currently existing in H55 to make you approve of it and adapt it to any new AI that makes it work better.

Great news, let's hope Quantomas finally realizes that a lot of 5.5 changes that he assumes to be set in stone and/or simply not transparent to his AI can actually be very easily rigged to meet his new developments.

Neutral stack sizes and what not in 5.5 are largely attempts to compensate for certain flawed AI behaviors, as (and if) the AI gets better Magnomagus will surely adjust all of that stuff to give no unfair advantage neither to the human player nor to AI itself.

Probably the biggest current problem is that AI suffers much bigger losses in combat with neutral stacks, which triggers a lot of 5.5 much needed adaptations, no matter how clumsy or variety-killing they could look to a human strategist.

Greenpakto said:
PS: I understand you like it the way it is, but in my opinion having ultimate abilities makes the game way more fun to replay and removing that was the biggest flaw of 5.5 (the ONLY flaw in fact). Anyway, I have derailed the thread enough.


Actually less replayable because the vanilla approach to the ultimate abilities was more complex (more layers / 'moving parts') but resulted in less depth (fewer viable choices because you had to repeat the same hero build to acquire a doomsday weapon, ditching the possibility amounts to defeat at the hands of anyone who finally gets theirs).

Ultimate abilities need to be remade from scratch, if ever allowed, that's 8 factions multiplied by 3 hero classes, 24 (!)
____________

 Send Instant Message | Send E-Mail | View Profile | Quote Reply | Link
Quantomas
Quantomas


Responsible
Famous Hero
AI Wizard
posted June 16, 2020 04:11 AM
Edited by Quantomas at 09:39, 16 Jun 2020.

It is not the question whether MMH55 offers nice-to-have features. As I said it is a mod in its own right. I can respect that. But it doesn't solve the issues I see as critical to create a grand strategy game.

If you continue to belittle my work or suggest my mind is clouded or write disclaimers in the style of a lawyer, this does little to convince me, because my main arguments were not addressed.

My interest is mostly the strategy game and MMH55 lacks for me the persistence and integrity required of a strategy game.

I suggest that you read my careful explanation above to understand my point of view. Why we have two mods now.

I don't see why we can't cooperate in the future.

But if you continue bashing this door, it will eventually close and it will be hard to open it again.

If you are interested to see the new advanced AI of Heroes V come to live or take part in the project, you are welcome to join us on Discord. Here is an updated link: discord.gg/s7yaCz
____________
Do you love strategy games? Join us on Discord: discord.gg/JKU6tey

 Send Instant Message | Send E-Mail | View Profile | Quote Reply | Link
thGryphn
thGryphn


Promising
Famous Hero
posted June 16, 2020 12:13 PM

Quantomas said:
It is not the question whether MMH55 offers nice-to-have features. As I said it is a mod in its own right. I can respect that. But it doesn't solve the issues I see as critical to create a grand strategy game.

If you continue to belittle my work or suggest my mind is clouded or write disclaimers in the style of a lawyer, this does little to convince me, because my main arguments were not addressed.

My interest is mostly the strategy game and MMH55 lacks for me the persistence and integrity required of a strategy game.

I suggest that you read my careful explanation above to understand my point of view. Why we have two mods now.

I don't see why we can't cooperate in the future.

But if you continue bashing this door, it will eventually close and it will be hard to open it again.

If you are interested to see the new advanced AI of Heroes V come to live or take part in the project, you are welcome to join us on Discord. Here is an updated link: discord.gg/s7yaCz


Quantomas, just to be clear, nobody representing MMH5.5 responded to your previous post. Just users. Well, there is nobody representing MMH5.5 other than magnomagus himself

The way I see things, once your new AI is out, magnomagus can make adjustments in MMH5.5 (according to your recommendations to make the new AI actually work) for true compatibility. At least, that's what I gather:

magnomagus said:
But if it comes to your approval, I'm completely open to deactivating stuff currently existing in H55 to make you approve of it and adapt it to any new AI that makes it work better. Please stop assuming that how it looks now is some rigid form that can't be changed and then needs to be criticized as not adaptable to an AI that exists only in theory.



You have your own vision for an AI for a strategy game, and magnomagus has his own reasons to make the changes in MMH5.5 he did (to summarize, I believe: to improve the challenge posed by the AI, plus to balance the strength across factions and skills).  

But with his willingness to course-correct for your new AI, I really don't see a reason for negative feelings.

One last thing: it is correct that 3.1j AI doesn't do well in combat, which really dampens its development and eventually spirals the AI down. You shouldn't take this as people putting your AI down, or take the AI-boosting features as an affront. It is what it is, and I hope you can take this as constructive criticism. Everyone here wants better AI, and combat performance is a huge part of what AI needs to do well in a heroes game.

Cheers!

 Send Instant Message | Send E-Mail | View Profile | Quote Reply | Link
magnomagus
magnomagus


Admirable
Legendary Hero
modding wizard
posted June 16, 2020 01:23 PM
Edited by magnomagus at 13:30, 16 Jun 2020.

@Quantomas

Ok let me first say that this post is meant as friendly, since I know forums have limited emotional expression, I know I can enter a discussion 'strongly', but you also made some very strong statements so this gives me the impression you prefer a tough debate. In order to make any cooperation possible I would still like our issues to be resolved, because I believe in large part they are based on unnecessary misunderstandings

Quote:
If you continue to belittle my work


That works both ways, me and others have spend hundreds of hours to mathematically analyse every tiny detail of the game and there are hundreds of pages here and on dicord discussing and addressing every balance issue based on human vs human play, but somehow you discard the whole thing as 'not a strategy game', 'lacking integrity' and tell me that I'm 'fooling' players.

Ok so then I'm still willing to take a pause and see what issues you still 'see as critical to create a grand strategy game', because believe me if h55 has any such issues I would like to have them adressed. But then you bring up these arguments

Quote:
You made most hero skills and feats stronger and gave those synergies that could be exploited by intent


Exactly the opposite philosphy of what h55 had from day one, the purpose was always to nerf the skills that are too strong and buff what is too weak and remove all exploits. On top of that primary skills were nerfed so they should actually be weaker. Consequently heroes can't be stronger than in h5 or h3 still=>

Quote:
As a response you created tools that made the neutral stacks that guard mines, chokeholds and treasure vaults many times stronger.


I made these items scale up to the difficulty required by human players, because that is just how good human players are, I can't change that, also the growth is user-customizable

Quote:
You added a gouvernor feature and skills that give your heroes additional free troops each week.


Governance mainly serves to boost secondary hero experience to improve their tactical possibilities, it generated a lot of positive feedback, the troops bonus is almost neglectably small and not free, the only items that give free troops are 'legion' artifacts that also exist in heroes 3, in heroes 3 they can be active from 4 slots, while in h55 only from 1 slot, since they overrule each other. Which brings me to this=>

Quote:
To streamline the game you introduced a townportal spell that is in line with the most exploitive version of Heroes III's townportal with expert earth magic, which is by most veteran strategy players perceived as overpowered and decisive.


This line is incomprehensible to me, the town gate spell in h55 is an optional feature that can be switched on and off and is far more limited in use compared to the h3 equivalent, but then later in your post you claim you want to make a strategy game more in the vein of 'HOTA', that has a more dated AI compared to 31j and advanced town portal.

So lets take a look at parts of your arguments I do understand:

1. hero movement parameters

these are 'duct tape' solutions that were made at the time to make things work 'for the time being' since there were no better technical solutions available. Personally I'm ok with any solution 'univeral penalties, native terrain, no penalties' whatever works. Since these are just numeric parameters, it will be very easy to adjust. It just seems so pointless to bicker over this.

2. cheats

removing those is literally deleting a few code lines, if there is any cooperation between us, I don't mind giving you total freedom to how you want to make the AI work for the players, you can take their criticism as guidelines, not mine.

3. AI not seeing, town conversion and governance

one of the reasons they were made optional, at first I thought they were going to be multiplayer only, but observing the Ai using it with 'the fog of war' removed. I didn't notice any issues, maybe this is because 31j or 31 was written to be very adaptable since campaign scripts can also cause sudden big changes 'on the ground'.

...

the fact that some your criticism actually seems to support H55 philosophy gives me hope that some day you will notice how h55 and your AI project can benefit from each other, I know a lot of people like to see this cooperation happen and I owe it to them to make my best effort. If at some point it feels like 'bashing' on a door then, yeah it will be better to move on.
____________
MMH5.5 Downloads | MMH5.5 Translations | MMH5.5 FAQ

 Send Instant Message | Send E-Mail | View Profile | Quote Reply | Link
Quantomas
Quantomas


Responsible
Famous Hero
AI Wizard
posted June 16, 2020 08:40 PM

I explained it already a few times, here and to you on Discord, that many claims you make simply don't check out.

I don't see why you are so obsessed with convincing me of something that is obviously not true. Let it rest, otherwise I might take indeed offense.

Most of your latest claims don't hold up under scrutiny. Let it only be noted that your claim that MMH55 gives not more free creatures than the vanilla silently omits the DBC mechanic that gives players entire stacks of Manticores and Death Knights.

It simply doesn't make sense to go over these claims one by one. Instead it is better to explain the issue TheGryphon brought up as this reflects on the points you made as well.

thGryphn said:
Quantomas, just to be clear, nobody representing MMH5.5 responded to your previous post. Just users. Well, there is nobody representing MMH5.5 other than magnomagus himself

I don't agree here, to simply exempt "just users". Even regular users get their opinions from somewhere, and more often than not it is the messages that are fed to them by leaders. They somehow learned about me without ever talking to me before, or knowing what I do. It's these baseless claims (making disclaimers in the style of a lawyer was actually referring to Magnomagus) that are repeated again and again and never questioned.

Regarding the subject of cheats in relation to the performance of the combat AI, funny that you mention that because I explained the exact same thing to Magnomagus already, but he apparently doesn't accept it. So you have to set his opinion aside for a moment and think for yourself.

Let's have a look at the facts.

First, the combat AI in the vanilla and in 3.1j is identic.

Back in the day when we released 3.1j, which was downloaded 100000 times from our server alone, we received tons of feedback, like the AI not building their towns not ideally, the AI being too aggressive or not aggressive enough, not gathering reinforcements smartly and so on. But one thing we didn't receive were reports that the AI wasn't capable of dealing with neutral stacks or breaking out of zones.

It is worth noting that 3.1j doesn't use cheats, so in truth, on top of the basegame it doesn't need these to deal with neutral stacks. It just performs fine.

There are reasons for this.

The losses the AI takes in combat against neutral stacks are not determined by the combat AI. Read this again.

Instead the losses against neutral stacks are computed by a formula that is in place at least since Heroes III. It was developed by NWC and has proven itself as a very good heuristic over and over. It takes into account the strength of both armies, the creature power values come into play here, plus adjustments for the hero leading an army, and produces a result that is the percentage of losses. This is the formula that has been used by the game played by millions of players over decades.

It works in Heroes III, is still used by HotA, and proved its value again in Heroes V with 3.1j. The reason it didn't work for Heroes IV is that the AI didn't take into account the mobile stacks introduced in this installment.

But back to 3.1j, there is absolutely no credible reason to believe that 3.1j is underperforming against neutral stacks.

On the other hand MMH55 made the heroes in combat much more powerful. It is silly to disclaim this. Just watch Aruarian's livestreams of his multiplayer matches. There is abundant evidence how armies with a handful of troops can defeat much, much more powerful opponents.

Did you never ask yourself why players requested many times more powerful neutral stacks? Why you actually had to develop the map-mixer tool to modify the maps from the very large pool of well-crafted vanilla H5 maps? It is simply self-serving to blame this on players wanting extra challenges or 3.1j not performing well in combat.

If you don't believe me, simply try the current beta 3.1.18 released today and play a vanilla map with it. If you don't trust me, as 3.1.18 contains modifications, take pure vanilla + 3.1j for a ride. It has been tested countless times, on campaigns like LotA or any of the thousand of handcrafted maps on maps4heroes.

If you are interested to see the new advanced AI of Heroes V come to life or take part in the project, you are welcome to join us on Discord. Here is an updated link: discord.gg/Fg9DZM
____________
Do you love strategy games? Join us on Discord: discord.gg/JKU6tey

 Send Instant Message | Send E-Mail | View Profile | Quote Reply | Link
Greenpakto
Greenpakto


Adventuring Hero
posted June 16, 2020 11:08 PM
Edited by Greenpakto at 23:09, 16 Jun 2020.

Quantomas said:
I explained it already a few times, here and to you on Discord, that many claims you make simply don't check out.


It works in Heroes III, is still used by HotA, and proved its value again in Heroes V with 3.1j. The reason it didn't work for Heroes IV is that the AI didn't take into account the mobile stacks introduced in this installment.

discord.gg/Fg9DZM


I have to say that I dont think Heroes 3 AI (hota as well) works very well. It works in small maps, but when we play larger maps the AI is extremely easy to beat when end game comes (they have lost way too many troops). Thats what made me loose interest in Heroes 3  (hota included), it simply did not challenge me enough. 5.5 does something very well, the AI can still put up a decent fight at endgame because they haven't lost many troops (but there are flaws here as well, they sometimes runs away or runs past me instead of actually fighting me when they have a HUGE chance of taking out at least 90% of my forces or even winning, which is a flaw). But the fact that its extremely hard to beat the endgame AIs main army is a good thing.

 Send Instant Message | Send E-Mail | View Profile | Quote Reply | Link
dredknight
dredknight


Honorable
Supreme Hero
disrupting the moding industry
posted June 16, 2020 11:12 PM
Edited by dredknight at 23:20, 16 Jun 2020.

Quantomas said:
The losses the AI takes in combat against neutral stacks are not determined by the combat AI. Read this again.

Instead the losses against neutral stacks are computed by a formula that is in place at least since Heroes III. It was developed by NWC and has proven itself as a very good heuristic over and over. It takes into account the strength of both armies, the creature power values come into play here, plus adjustments for the hero leading an army, and produces a result that is the percentage of losses. This is the formula that has been used by the game played by millions of players over decades.


I cannot agree with that - unlike heroes 3 or other similar games that you mention, heroes V does not fall into the category of games that there is a single formula of success for army calculating mechanics. This is simply because of:
- development of hero depth is way higher than in the examples, which means there is a possibility for higher synergy between ability picks, which means hero has the chance to become high performer on the combat field way more than in the other games.
- Some Spell effects that are impossible to be calculated in terms of efficiency
- Heroes V is the first heroes in the game series since 1,2 and 3 that spellpower actually gives real battleground privilege to Mage heroes.
- Also how does AI compares efficiency between hero combat strike and Haste? or difference between Armageddon and Blind? I am not even speaking about long term game plans here.
- Using a "casualty formula" based AI will never compensate for a real player option to abuse the game through very narrow specifically tailored strategies for holding zero casualties battles. A small example from ToE - if Tent healing cast value is left to default value experienced players will have 0% casualites in battles where they face PvE melee only units because they will delay the end of the battle until all Tent heals are exhausted. (FYI this was nerfed in 55).
- Also when AI levels up how does it calculate the choice of what skill to pickup when it is given attack and Dark magic ? What if the hero is balanced mage class where that game it has bad roll of primary skill level ups on spellpower? Also how does the AI does calculation on this information compared to the possible main hero build of his enemy (which the AI can  see in the Tavern). Can the AI make a choice that will simultaneously try a counter build?

The complexity of this game for PvP is so ridiculous that if Starcraft 2 AI is one of the top game AI of what we have then we are really far from getting a comparable AI for heroes 5 if it is just based on 1 formula.

By default ToE is very abusable game on its own so I don't understand why you insist that this is what we do to MMH55 when 9 out of 10 skill changes in all updates are nerfs.

Quantomas said:
On the other hand MMH55 made the heroes in combat much more powerful. It is silly to disclaim this. Just watch Aruarian's livestreams of his multiplayer matches. There is abundant evidence how armies with a handful of troops can defeat much, much more powerful opponents.


Aruarian usually play Deadly Environment map which is not an example of normal MMH55 game.
____________
Join our official discord channel | NCF Utility Beta

 Send Instant Message | Send E-Mail | View Profile | Quote Reply | Link
magnomagus
magnomagus


Admirable
Legendary Hero
modding wizard
posted June 16, 2020 11:45 PM
Edited by magnomagus at 23:53, 16 Jun 2020.

Quote:
I don't see why you are so obsessed with convincing me of something that is obviously not true. Let it rest, otherwise I might take indeed offense.


I'm sorry if I offended you, but I was trying to explain we are actually on the same page about something, namely that heroes should not be too strong, in the hope this could create a better groundwork for cooperation, the balance issues exposed in aruarians stream have been in the process of being dealt with recently.

Quote:
Regarding the subject of cheats in relation to the performance of the combat AI, funny that you mention that because I explained the exact same thing to Magnomagus already


I also apologize for spreading misinformation about the combat AI, I do not recall you ever explaining this to me in clear terms. In my experience playing vanilla+31j the AI did not provide the same level of challenge as the original TOE AI regardless of the fact that it moved around smarter and was calculated faster. Since I knew the 3.1 had cheats I blamed this on the loss of those cheats and become convinced of this since I got back what I wanted after reintroducing them. I am of course curious how 3.1.18 performs in comparison.

Quote:
They somehow learned about me without ever talking to me before, or knowing what I do.


There is a large part of h55 related activity I have no control over (and I shouldn't, because then I would indeed be obsessed), Sometimes someone tells me dredknight made the whole mod by himself and I don't mind, and don't even correct it. When people talk info gets distorted. Please do not consider me a 'leader' who is on purpose influencing people to spread a certain opinion on a forums, all people around here are speaking on their own behalf.

Quote:
Did you never ask yourself why players requested many times more powerful neutral stacks?


They didn't, the on-map stacks are unchanged, the creatures growing in battle sites are a feature I added from H4 and serves a purpose for players who enjoy playing XL maps, mapmixer wasn't made by me, it was made for adding NCF creatures and remixing the artifacts for replayability.
____________
MMH5.5 Downloads | MMH5.5 Translations | MMH5.5 FAQ

 Send Instant Message | Send E-Mail | View Profile | Quote Reply | Link
Quantomas
Quantomas


Responsible
Famous Hero
AI Wizard
posted June 19, 2020 01:35 PM bonus applied by Galaad on 21 Jun 2020.
Edited by Quantomas at 19:42, 19 Jun 2020.

I uploaded a new test build that fixes an issue with hero class names and adds a license that gives redistributors and modders clear guidelines how they can use the mod.

This test build is limited, but you can play multiplayer and RMG maps. Note: the shroud is lifted for AI observation.

If you are interested to see the new advanced AI of Heroes V come to life or take part in the project, you are welcome to join us on Discord. Here is an updated link: https://discord.gg/45SFAg

Edit: the link is valid until June 20th.
____________
Do you love strategy games? Join us on Discord: discord.gg/JKU6tey

 Send Instant Message | Send E-Mail | View Profile | Quote Reply | Link
Gidoza
Gidoza


Famous Hero
posted June 20, 2020 11:30 PM
Edited by Gidoza at 23:33, 20 Jun 2020.

Quantomas said:
I uploaded a new test build that fixes an issue with hero class names and adds a license that gives redistributors and modders clear guidelines how they can use the mod.

This test build is limited, but you can play multiplayer and RMG maps. Note: the shroud is lifted for AI observation.

If you are interested to see the new advanced AI of Heroes V come to life or take part in the project, you are welcome to join us on Discord. Here is an updated link: https://discord.gg/45SFAg

Edit: the link is valid until June 20th.


D'oh - I was away from the forums just long enough to miss the due date - oh, well.



As for the rest of the discussion - I've read through everything and I'll give my short responses to the whole conversation here.

First - I exclusively use H5.5, as I do think it's an upgrade over standard H5.  That's not an assessment of anything - it's just where I'm coming from.

Second - I can see how some of the H5.5 changes could cause problems for the new AI.  I guess it's up to Quantomas to adapt or H5.5 to adapt to make things work together there.  I think it would be a shame if it didn't work with H5.5 because I wouldn't even bother with H5 at all if it wasn't for H5.5, but that's up to Quantomas.

Third - I absolutely LOVE the pro-H5.5 hubris coming from the H5.5 dev team, especially when it comes to H5.5's so-called perfection and "mathematically provable" stuff.  No, it's not perfect; no, I've yet to see a single mathematical "proof" of balance posted anywhere; and no, others are not mandated to accomodate H5.5's heavenly form.  The reverence of Neutral stacks in particular tends to make H5.5 boring for me, regardless of how balanced it may be - I'm playing H5 to fight Heroes, not bigger and bigger Neutral stacks.  On the other hand, I really like H5.5 because it tackles some of the internal numbers that aren't easy to tackle (e.g. experience point values) that really ought to be tackled and can't be by a regular Joe like me - that's what I'm thankful for in particular.  And no, I'm not using "heavenly form" ironically here - I've read lots of posts here from lots of people, and it seems to me that questioning H5.5's balance decisions is like trying to question a religion.  I'm not all-knowing either, but I'm perfectly capable of making my own analyses - thank you.

Last - I mean I'm looking forward to the new AI and would love to see how it works (and if any of my own modding will break it) - but hey if it doesn't fit it doesn't fit.  There's probably something better I could be doing with my time anyways.  In the mean time, I'm taking as many hand-written notes as I can to try and help the project as much as possible, because I think it's a good project, I enjoy helping out on this sort of thing, and I know that I have a good intuition of the sorts of things that need to be hammered out - I might not be able to program anything, but I can point to things that matter most.

Chow,
-G

 Send Instant Message | Send E-Mail | View Profile | Quote Reply | Link
Quantomas
Quantomas


Responsible
Famous Hero
AI Wizard
posted June 20, 2020 11:40 PM

Here is an update: note though that it is only a test build and not a proper release. I suspect there is currently a bug that keeps the AI from building up its towns. Shouldn't take long to iron out though.

If you are interested to see the new advanced AI of Heroes V come to live or take part in the project, you are welcome to join us on Discord. Here is an updated link: discord.gg/DHhpbY
____________
Do you love strategy games? Join us on Discord: discord.gg/JKU6tey

 Send Instant Message | Send E-Mail | View Profile | Quote Reply | Link
Gidoza
Gidoza


Famous Hero
posted June 23, 2020 09:39 PM

Quantomas said:
Here is an update: note though that it is only a test build and not a proper release. I suspect there is currently a bug that keeps the AI from building up its towns. Shouldn't take long to iron out though.

If you are interested to see the new advanced AI of Heroes V come to live or take part in the project, you are welcome to join us on Discord. Here is an updated link: discord.gg/DHhpbY


Says that the invite is invalid?

 Send Instant Message | Send E-Mail | View Profile | Quote Reply | Link
Quantomas
Quantomas


Responsible
Famous Hero
AI Wizard
posted June 23, 2020 09:44 PM
Edited by Quantomas at 21:49, 23 Jun 2020.

Here is an updated invite:
____________
Do you love strategy games? Join us on Discord: discord.gg/JKU6tey

 Send Instant Message | Send E-Mail | View Profile | Quote Reply | Link
Gidoza
Gidoza


Famous Hero
posted June 24, 2020 12:01 AM
Edited by Gidoza at 00:23, 24 Jun 2020.

OK thanks.

Well I'll give it a shot even with my variant of H5.5 and watch what happens.  The AI is pretty dumb at building up its towns in the first place, so unless it's just *not building things*, I suspect it could be fine.

EDIT:  So I played one turn of H5 and watched what happened with each of the 7 AIs.  I would consider all the AI's turns to be a flawlessly terrible play for the ultimately simple of reasons:  they all hired a 2nd Hero; one of the AIs didn't use the 2nd Hero at all: meanwhile, in the map I'm playing all the starting Ore Mines and Woodmills are unguarded and in range of the starting Town, close enough that a single Hero can take both of them.  I think every single AI failed to take both of these resource generators even with 2 Heroes available.  In my strategy book, I'd call this a critical error, because that 2 Wood or 2 Ore can be extraordinarily important at certain times.


Also - I was reading in the ReadMe that apparently the AI only bothers to attack Banks (like Dwarven Treasuries) on Heroic?  (and on other levels if the player is a "threat")  Why?  I'd think that a smart computer would engage these always and no matter what.  I usually avoid Heroic (and ultimately created my own resource allocations for each difficulty level) because stupid spawn rates for Neutrals make the game boring - I wasn't even aware of the AIs and banks feature!  I'll re-allocate those values for sure and play on Heroic as if it was "standard" just so the AIs hits those Banks - that ought to be a standard feature.


Oh, yeah - last note.  This is regarding the old AI because I haven't played enough with the new one to know enough:  I noticed with 3.1j that it doesn't like moving Heroes into 2 things.

1)  It seems really dumb when it comes to Garrisons.  Like, it'll fight the Garrison guards, then often just leave without taking the goodies at all.  It might eventually get around with it with another Hero some weeks later...but not always.  I found that removing a Garrison and just replacing the Garrison with a mixed-style Neutral stack made the AI behave more normally on the map.

2)  The AI also doesn't like what I'll call "cubbies" - or in other words, little dead-end pockets that one might guard with a...Garrison.  Pretty much anywhere on the map that's slightly out of the way is somewhere the AI rather meaninglessly avoids with all Heroes, even secondary ones that have nothing better to do.  Sometimes the primary won't even touch obviously-available mines and resources because of this, thus depriving itself of much-needed "food" and ending up being the reason I crush it all the time.


Anyways those are a few thoughts from recent memory.
-G

 Send Instant Message | Send E-Mail | View Profile | Quote Reply | Link
Quantomas
Quantomas


Responsible
Famous Hero
AI Wizard
posted June 24, 2020 09:38 AM
Edited by Quantomas at 10:26, 24 Jun 2020.

Thanks for your feedback. Although going forward I recommend to post feedback on the dedicated channel for AI discussion on our Discord. This will make it much more easy to facilitate communication between all contributors of feedback.

Regarding the hiring of two heroes at the beginning of a game, with an AI having 30K gold, I wouldn't see this necessarily as a mistake because it doubles the forces that can be carried by the main hero.

If they don't take unguarded mines in their range, and if their is nothing better to do, this is clearly a bug.

Regarding the town not building things, I suspect that there is a bug that was introduced lately. I am currently working to correct this. It takes a bit more time as I am upgrading the way how the towns are managed in general and it has a ton of features, like buildings requiring resources, having dependencies, requiring townpoints, optimizing the recruitment week by week plus adjusting to the current strategic context (threatened, secure, what faction it belongs to) and how many resources are available currently and on the map.

Regarding the treasure vaults, this feature was introduced as to not discourage weaker players playing on normal and lower difficulty. Because it can be pretty annoying if you come across treasure vaults and you regularly get the message that it is deserted. In general the game, this comes from the vanilla, has the AI underperform on lower difficulty settings. This pertain to tactical combat as well. I am open to revamp and change this behaviour based on player feedback.

Regarding garrisons, I have no clue what you mean. Garrisons don't give any loot.

In general, based on our latest tests, the AI in MMH55 is totally broken. I am afraid all you played against was Magnomagus' scripts.

If you want to verify what the AI does on 3.1j, it is also available as a standalone download from our Discord.

I'll say a bit more on Discord later today.
____________
Do you love strategy games? Join us on Discord: discord.gg/JKU6tey

 Send Instant Message | Send E-Mail | View Profile | Quote Reply | Link
Gidoza
Gidoza


Famous Hero
posted June 24, 2020 03:05 PM
Edited by Gidoza at 15:07, 24 Jun 2020.

Quantomas said:
Thanks for your feedback. Although going forward I recommend to post feedback on the dedicated channel for AI discussion on our Discord. This will make it much more easy to facilitate communication between all contributors of feedback.

Regarding the hiring of two heroes at the beginning of a game, with an AI having 30K gold, I wouldn't see this necessarily as a mistake because it doubles the forces that can be carried by the main hero.

If they don't take unguarded mines in their range, and if their is nothing better to do, this is clearly a bug.

Regarding the town not building things, I suspect that there is a bug that was introduced lately. I am currently working to correct this. It takes a bit more time as I am upgrading the way how the towns are managed in general and it has a ton of features, like buildings requiring resources, having dependencies, requiring townpoints, optimizing the recruitment week by week plus adjusting to the current strategic context (threatened, secure, what faction it belongs to) and how many resources are available currently and on the map.

Regarding the treasure vaults, this feature was introduced as to not discourage weaker players playing on normal and lower difficulty. Because it can be pretty annoying if you come across treasure vaults and you regularly get the message that it is deserted. In general the game, this comes from the vanilla, has the AI underperform on lower difficulty settings. This pertain to tactical combat as well. I am open to revamp and change this behaviour based on player feedback.

Regarding garrisons, I have no clue what you mean. Garrisons don't give any loot.

In general, based on our latest tests, the AI in MMH55 is totally broken. I am afraid all you played against was Magnomagus' scripts.

If you want to verify what the AI does on 3.1j, it is also available as a standalone download from our Discord.

I'll say a bit more on Discord later today.



You misunderstood what I was saying.

I didn't say that purchasing a second Hero was a mistake.  I said that failing to take both the Sawmill and an Ore Pit on turn 1 when it is possible to do so was a mistake.

Secondly, I wasn't saying that Garrisons give any loot.  I was saying that the AI often choose not to go PAST the Garrison after it beats the Garrison defenders (the loot is behind the Garrison).



As for H5.5 Scripts - I do need a better sense of what this is referring to anyways.  The only changes in Hero movement I'm even aware of there is that extra costs on rough terrain have been flattened (all 25%), and that there's a couple artifacts whereby you lose movement when you unequip them (Boots).  What else is there?  I don't understand what scripts have to do with any of this.



EDIT:  Also made a bug report in the BugReport channel on Discord - I think that's where it goes.  But in case it doesn't, here it is.

Hi there - so I was playing a map today and noticed that the AI seems to behave just fine around normal guardians on the map.  BUT the Neutral Town nearby - the AI keeps attacking it with underwhelming forces and getting killed repeatedly.  This wasn't happening in 3.1J.

 Send Instant Message | Send E-Mail | View Profile | Quote Reply | Link
Gidoza
Gidoza


Famous Hero
posted June 24, 2020 03:43 PM
Edited by Gidoza at 16:15, 24 Jun 2020.

You know I went back here to read all this over again, because I felt I missed something.  So I'll just comment on things.



Quantomas said:
Certainly, if you alter the game so much that you can defeat 10K size neutral creature stacks in week three with the troops of two mediocre developed towns, anything that the AI plans becomes meaningless.


magnomagus said:
This nothing like H55 and can be proven mathematically, H55 uses much lower Attack & Defense scaling, to achieve exactly the opposite, giving neutrals a better odds than before, possibly streams of Deadlyenvironment clouded your judgement on this, but that map starts with build up towns, stronger than normal neutrals and a sphinx, this isn't to blame on H55. Also 10k is a ridiculous exaggeration.


I'll tend to side with Magnomagus on this one.  Quantomas's suggestion does seem like exaggeration to me.  I will admit that I'm not a fan of the huge Neutral Stack sizes in H5.5 and feel that the game would be better with smaller stacks (which I've done in my own mod) so engagements with other players can come quickly; losses occur here and there as expected.  The change to Attack & Defence scaling is also a brilliant change and infinitely superior to the original H5.

Conversely however, it seems to me that in the original H5, defeating huge Neutral stacks was always perfectly possible anyways.  All you need is a couple shooters and an Ammo Cart.  H5.5 hasn't changed in that regard, so I'm not sure how this is Magnomagus's fault.

Also @Magnomagus - you really need to explain what the "can be proven mathematically" statement is even supposed to mean.  When you have towns of various types, with players choosing various kinds of heroes, with various skills, building up their towns in various ways, and playing on various maps and taking various paths, with various artifacts, I have no idea how you could possibly mathematically prove anything, ever.  Chess can be mathematically tackled.  Heroes 5 cannot be: you can get some rudimentary numbers for balance purposes and use math to set up the game's framework, but once you get into it there's so many variables that the only competent way to assess everything that's going on is by feel.  Please show us some documents of the math at work and let us know what you're talking about.

Quantomas said:
Now, let's have a look at how you modded the game. You made most hero skills and feats stronger and gave those synergies that could be exploited by intent,


magnomagus said:
This makes no sense, H55 is exactly the opposite, it nerfs all overpowered skills and buffs the useless ones, it has also removed pretty much all exploits by rigorous multiplayer testing and will continue to do so until they are all rooted out. Vanilla H3, HOTA and H5 all have far more exploits and insanely overpowered skills compared to H55.


Definitely agree with Magnomagus here, but I don't even think Magnomagus expanded on this point enough.

First - Ultimate abilities are gone.  These were insanely overpowered and I wanted them removed from the first time I even played H5.

Second - The Skills in original H5 have all kinds of twists and turns and weirdnesses that I can't even memorize.  H5.5's skill setup is fairly straight-forward and easy for a player and I would say *ESPECIALLY* an AI to handle.  While it's true that I'll always want several skills to play with each other for optimization, the AI will never need to depend on some exterior skill in order to pick interior abilities of another skill entirely.  I can only see this as being easier on the AI.

Third - Since combo skills in H5 were so dependent on all the intertwining of everything and aiming for Ultimates and the like, the streamlined 3-pronged skillsets offer good options and are *WAY* more strategic than anything the original H5 had to offer.  If we're talking about strategy, then H5.5 definitely has an upper hand in this regard.  I might regard the end result of some of the skill choices in H5.5 as poor or mediocre, but that's still one or two levels better than what H5 had to offer.

Quantomas said:
until players who previously stated they had difficulty winning in H5 said it is a breeze now.


magnomagus said:
I have never seen anyone comment on H55 saying it is a breeze, on the contrary people often claim it is too hard and get crushed by the AI, I have done many things over time to reduce the excessive difficulty for new players, since early in the process of making h55 I expected it to be much smaller and only for advanced players.


Let me be the first to comment to Magnomagus that I find H5.5's AI to be a breeze.

Now that said - the fact that the H5 AI cheated MAJORLY probably made it harder, but it depends on what you mean by "hard."  I'm not even sure how often if ever I even got attacked by the the original H5 AI.  It was just a game of my trying to overwhelm the enemy's double or triple spawn or whatever it was.  There wasn't even a game there.  H5.5 actually does seem like a game to me - one where I always win, but I prefer it to the original.  People had difficulty winning in H5 because the AI was cheating, not because the AI was good.

Quantomas said:
You added a feature to convert towns and dwellings to your faction to further boost your factions power. You added a gouvernor feature and skills that give your heroes additional free troops each week.


magnomagus said:
These are all OPTIONAL features, that can be deactivated


I think Magnomagus's retort here puts an end to this complaint.  I don't like any of these features myself either.  BUT...

I don't see what's wrong with the skills that give Heroes additional free troops each week however.  Those skills were put there because the original H5 skills were so weak that they weren't worth taking.  If something additional wasn't put in, then it would be better to remove the skill entirely so the AI wasn't tempted to take it (and I wouldn't bother either).  Having free troops based on your Knowledge count doesn't hurt the AI in any way or complicate its movements, so I don't get what the problem is.

Quantomas said:
This is the reason why I chose to develop a non-cheating AI for Heroes V.


magnomagus said:
I have played 31j extensively in vanilla TOE, and it had one major problem which I have pointed out many times, the combat AI is too weak to keep up with 'good' human players, so with every neutral stack it fights it loses more creatures, when it falls behind on creatures it loses its motivation to attack things and gets weaker and weaker
Because of this the original TOE AI is harder to beat than 31j. What you call cheats are in my opinion nothing else than corrections for the lost troops to give 31j a fair chance against humans.
You have a theoretical vision on the AI that doesn't exist right now, only 31j exists and with some help it performs best as is currently available.


I'd push back against Magnomagus here referencing an earlier point about Neutral stacks - it IS the case that H5.5 uses way more Neutral stack large sizes than the original H5.  Pulling back on this seems like a better solution than introducing minor cheating.

Quantomas said:
If anything, you abuse the AI. It borders on deception that core MMH55 members tell everyone showing up on Discord that the AI always cheats,


magnomagus said:
I added a very clear option in custom game that deactivates all cheats, please do not blame me for people not taking the time to read.


I actually prefer having the AI have a *FEW* cheats because I think it actually needs them.  Until we have a fuller and thorough AI that can compete at a human level, it seems necessary.

Quantomas said:
and frequently adding a @Quantomas to give the impression that I approve of MMH55


magnomagus said:
You joined the MMH55 discord server (ironically even before me), ofc people will think you approve of it. But if it comes to your approval, I'm completely open to deactivating stuff currently existing in H55 to make you approve of it and adapt it to any new AI that makes it work better. Please stop assuming that how it looks now is some rigid form that can't be changed and then needs to be criticized as not adaptable to an AI that exists only in theory.


I'd definitely be interested in conceptually tackling some of H5.5's features to figure out what does and does not need to be there so the new AI could work.

I will agree that it's Quantomas's AI and his project so he can create it however we wants - so H5.5 needs to adapt to that, and not the other way around.  Reciprocity seems a preferable way of doing things, however.  At the end of the day, though, even if there is no bending from Quantomas's side at all - identifying exactly what makes the AI tick in H5.5 is useful so we could modify it to make the best use of the new AI.

Quantomas said:
Given all what I said here, you certainly will agree, that the only sensible solution to move forward is to draw a line between what is MMH55 and my vision for Heroes V with a quality AI.


magnomagus said:
I don't care if an AI cheats or not, I only care if it provides a challenge, 31j didn't, so it needed cheats, if 31k or 32 or whichever new version can provide a challenge without cheats then I'm ofc totally happy with a cheatfree AI. Also you overlook the fact that human players are also cheating constantly in singleplayer, by saving/loading/replaying battles. Hard to keep up with that for an AI


I'll push back against Quantomas here by noting that probably Quantomas's assessment of H5.5 overall in terms of balance is actually incorrect.  If I had to pick strategic design, H5.5 is superior by so many levels it isn't even fair.  If the new AI only worked in H5 and was as good as I hoped it to be, I would necessarily end up modding and changing skills in H5 without the H5.5 "scripts" you say that mess up the AI to bring H5 to a better level.

Let's be clear - I don't play H5.5 for the new features like Governance and Town Conversion etc...I play H5.5 because the exp system is changed, the 25% penalty on all terrain makes more sense for balance, and because the skill system isn't completely dumb.  If we were to make a list of the things in H5.5 that were actually essential, I'd opine that the crux of the changes would be minor but the strategic effect of them would be massive.

Quantomas said:
Honestly, you would be much better off, if I made the vanilla AI as fast as H3. This is easy to do and would only require a week, not many years of dedicated work. It would give you a solution that can handle all features and is adaptable by the DefaultStats. It would only go through the motions, but that is all MMH55 requires.


magnomagus said:
I would surely like to try that.


I mean if that's all we could get then I suppose it's better than nothing - but it's really the awfulness of the adventure AI and the combat AI that's the issue.  And that awfulness comes from not taking mines, not defending towns, casting dumb spells, or making dumb combat decisions.  Most of the issues there are rather small and unrelated to any H5.5 features.  Quite frankly, smaller changes to normal AI behaviour would probably benefit them in H5.5 more than actually adapting to H5.5 features would.  Seriously.





To summarize my thoughts...

1) I would tend to agree with Magnomagus that there is a good chunk of Quantomas's assessment of H5.5 that does not pertain to reality.  H5.5 is much more well-balanced and better thought out than H5.  Furthermore, when I switch from H5 to H5.5, I don't perceive that it is a different game.  It feels like proper H5, but with many things done right or done better.

2) I would tend to agree with Quantomas that H5.5 needlessly tinkers with too many things, and this makes the AI upgrade a pain, and can also make H5.5 a pain at times as well.

3) Ultimately, where I stand is that I'd love to see the new AI in motion, but it would be an eye-roll to see it in an awful environment such as the original H5; which I quit playing NOT because of the AI, but because the game was strategically dumb.  Unfortunately, there aren't any good modded alternatives *other* than H5.5.  An actual, streamlined H5.5 that tackles the heart of the issues of H5 would be perfect.

 Send Instant Message | Send E-Mail | View Profile | Quote Reply | Link
Jump To: « Prev Thread . . . Next Thread » This thread is 46 pages long: 1 2 3 4 5 ... 10 20 30 40 ... 42 43 44 45 46 · «PREV / NEXT»
Post New Poll    Post New Topic    Post New Reply

Page compiled in 0.2562 seconds