H7 vs taking things into our own hands

[Steven Aus]

Quantomas, the Celestial Heavens web-site might be placed in hiatus (suspended animation) soon. Do you have any more news about the new game? I guess you will still post on HC, but it would be good if you could do a last update here.

[Quantomas]

I honestly believe that Heroes of the Ancient Order can revive the community big time. We can revive the popularity of Heroes V and build a much more advanced game, with a lightning fast and competent AI to start with. CH may blossom again and we might make it a popular destination once more.

My advice: don't suspend the Celestial Heavens just yet.

[Dalai]

That's funny how history goes in rounds. I remember the moment when I was convinced, that Ubi team is hopeless. They don't understand the most basic things about the game, and no amount of funding and "development experience" can change it. It was before H5 launch. All that happened after that proved me right. When I said it aloud people replied "give them a chance". Ok, we now have 3 failed games with I don't know how many patches and addons, and nothing changed.

Quantomas tried to resuscitate H5, and after a lot of work and effort arrived at same conclusion - the patient is dead, stop torturing him.

I don't really follow anything after H4 seriously, but I think there were no other optimists to resuscitate H6 or H7, which sais a lot.

Ironic

As an explanation I can only guess that the management does not recognize the central value a proper AI has for a strategy game and its playing experience and the effort it requires to develop one.
...
tell me that they were also lacking a good lead programmer.

You can count a dozen ultra-important things they don't recognize or lack. But it can be summed up to one and only problem: they don't care. As long as fans, who have been shown a middle finger not once, not twice, but three times (!!!) continue preordering Heroes - why should they care?

I also do not understand why Ubisoft and Limbic did not inquire about the AI work I do. Actually, I spoke to the CEO of Limbic in depth about H6, way back before H6's first expansion. So, they definitely have my contact details.

You seem eager to work for the company who destroyed our favorite franchise. So they have couple more chances to rip us off with crappy games with excellent AI? You seem to understand that it's not AI that is problem. So fixing AI won't fix the main problem, and will not result in a good game. They milked the great reputation "Heroes" had, and now you want to give them a chance to milk yours?

Imagine they do hire you. Imagine you do job right, and the game is success. It will only prove my initial point - they are a bunch of ignorant professional impotents. Do you think that is what they want? So what are the chances of that happening?

Do you want to "demonstrate what a proper AI can do for the game. It will be an eyeopener" or do you want to just get hired?

Mostly all parts of the AI have now been written.

Do you have it described anywhere? As an algorithm? H4 strategic AI was terrible, and I would very much like to know your your approach to it, as you seem to specialize on it.

[Quantomas]

Not so.

IMHO a remaster of H4 would need to address more than one issue: the AI of course, the battlefield, and somehow a smart solution will be required to rework the heroes' presence on the battlefield.

Do you think I only could do that if I were hired?

Mostly it requires access to the source code and a small budget, little more.

Ironic is that back in the day I purchased Heroes IV as soon as it was available. Coming from H3, I started the game on a higher difficulty -- and it was a complete disaster! Once my hero arrived at the enemy territory, there were only underdeveloped towns and hardly any opposition. Exploring further I discovered these small tombstones that indicated fallen heroes. They simply had run into wandering monsters! Apparently NWC didn't even do minimal testing on a higher difficulty, and with the game ruined for me I were so frustrated that I resolved not to buy the expansions. Possibly a lot of fans did this. Today, with the situation as it is and knowing a good deal more about HoMM, I wonder whether this was a good choice.

Given this fan behaviour and considering that H6 sold reasonably well, indicates that fans were quite satisfied with H5 TotE in the end. Considering the lackluster sales of H7, fans weren't that pleased with H6.

H5 TotE was just as addicting as H3 in respect of gameplay, but a faster and more competent AI as well as a much speedier autosave is a must to keep the game enjoyable in the long run. THAT IS COMING. And we will see.

H4 could live as well, if someone at Ubi would see merit in what I outlined at the beginning of the post.

[Dalai]

Do you think I only could do that if I were hired?

Not necessarily. But if I ventured a guess, I'd say you have a job which you like less than working with games' AI. So it's your hobby. And you can spare not so much time on it. If you had it as a job, you would be more motivated AND spend more time on it. So - although you could probably do it anyway, you would do it much better and faster if it were your job. How far am I from truth?

Was I wrong in my impression that you want to work for them?

Ironic is that back in the day I purchased Heroes IV as soon as it was available. Coming from H3, I started the game on a higher difficulty -- and it was a complete disaster! Once my hero arrived at the enemy territory, there were only underdeveloped towns and hardly any opposition. Exploring further I discovered these small tombstones that indicated fallen heroes. They simply had run into wandering monsters! Apparently NWC didn't even do minimal testing on a higher difficulty, and with the game ruined for me I were so frustrated that I resolved not to buy the expansions. Possibly a lot of fans did this. Today, with the situation as it is and knowing a good deal more about HoMM, I wonder whether this was a good choice.

Both people who love H4 and people who hate it agree that it required more time to develop and test. It was rushed, 3DO had their reasons, and the rest is history.

Given this fan behaviour and considering that H6 sold reasonably well, indicates that fans were quite satisfied with H5 TotE in the end. Considering the lackluster sales of H7, fans weren't that pleased with H6.

May be you are right, may be fans make decisions based not only on previous performance, or may be fans change their decision making process based on previous performance (that's me, btw) - it's not something I want to debate right now.

Mostly it requires access to the source code and a small budget, little more.
...
H4 could live as well, if someone at Ubi would see merit in what I outlined at the beginning of the post.

If I made a hocus-pocus and pulled H4 AI source code out of my sleeve (but no budget, small or otherwise) and offered you to work on H4 AI to prove your point - would you agree or would you come back to your exciting secret project? And if the guy from Limbic called you next day and offered to help them - what would you do then?

[Quantomas]

To clarify a couple of things: I studied computer science quite some time ago and used that knowledge to work on highly advanced tech. I worked for large and small companies and had the good fortune to pick up a good deal of organizational and management skills as well. Today I could work fulltime on the AI and Heroes of the Ancient Order, or any other interesting project. However, there are other issues beyond my control that keep me currently from working as much as I wish. I hope to be clear of these soon and get much more done.

I am normally not interested in entry level positions if this answers your question. I hold myself to highly professional standards and this includes ethics as well. While it might be tempting to fix a gem like Heroes IV, it would go against my professional ethics to release it without permission of the IP holder, i.e. Ubisoft. Fixing it as a proof of concept in order to convince Ubisoft of the project's value might be a valid consideration though.

[Dalai]

I am normally not interested in entry level positions if this answers your question.

No, I don't feel that any of it does, but that's ok.

You skipped my question about availability of some sort of AI algorithm, description or smth like that for a Heroes game, so I have to infer there is none.

If there was another, less talented, less experienced and less ethical programmer, who just wanted to fix H4 AI and ready to type all his code by himself with no help from assistants and junior programmers - what would you tell him? To read, to keep in mind, to center his thoughts around - anything really. Just to avoid typical mistakes and narrow traditional approach.

[Quantomas]

If there was another, less talented, less experienced and less ethical programmer, who just wanted to fix H4 AI and ready to type all his code by himself with no help from assistants and junior programmers - what would you tell him?

Start with a more simple project. Try to get a grip on the AI in the source of Heroes III. It is excellently written and lots of techniques can be gleaned from it. Still, this will most likely be the most complex project you ever approached. If you fail to understand this code completely, which a bright and skilled programmer can possibly accomplish in a time span of two years, you are most likely not suited for writing AI code.

You skipped my question about availability of some sort of AI algorithm, description or smth like that for a Heroes game, so I have to infer there is none.

Indeed. AI is very much an unsolved challenge. There are no tools that would be suitable to model an AI for a Heroes game nor for any other complex strategy game afaik. The complexity is simply mind boggling. On the one hand you have this next to infinite (in the Greek sense) space of possibilities, in which every hero is an actor and can visit a large number of places and objects on the map if you have an advanced lookahead, with a complex force model and on top of it the heroes can all interact with each other. On the other hand the result of high level AI meta-functions will be impacted by every small and intricate rule, of which there are an abundance in Heroes. For example if a hero picks up a boot that enhances travel speed, it will impact the interactions substantially, and not in a way that can be modelled easily.

You have to excel in math and in theoretical physics and quantum mechanics. Writing an AI very much resembles the structures of these subjects, so I highly recommend their study.

Additionally I found while designing the AI that there is so much information that it is not viable to write it down because finding the bits you need will cost way too much time. So you have to memorize everything and transform the design in your mind. Not for the faint-hearted. It is also a substantial barrier for building a team that attempts to develop an AI together.

In short you will need a programmer who is extremely skilled and dedicated and who would agree to give three or more years of his live to get the project to completion.

You can easily conclude why there are no interesting and complex strategy games that feature a fast and non-cheating AI. The work Gus Smestad did for Heroes III was probably the best that had been done until that point. The other game that comes to mind is Master of Orion II. Essentially back then the know-how of designing AI for strategy games was very much a living art, with programmers transferring and teaching their know-how to others. But apparently no one of the big publishers and big studios after 2000 did care about cultivating AI programming skills. In a way the know-how was lost, except for the source code of the Heroes games.

Nival revisited the Heroes III code but there is evidence that they didn't fully understand how it works to adapt it to the changed rules of Heroes V. It's no easy task. The meta-structure of the AI is not explained in the source, you have to develop it in your mind while reading all the routines. On top of that the AI doesn't work on its own but requires a very large amount of numerical tuning to get proper heuristics working. I guess NWC had a mathematical model to accomplish this.

As you can see, it is really no simple subject at all. What I am doing is building tools and implementing a prototype, kind of a blueprint that can be adapted for other strategy games as well. That's at the core of the business model of TESLA MINDS, but I do not like the idea of a middleware provider too much. So instead we are using the AI to build an advanced strategy game that should appeal to the vast majority of the turn-based community. This game will be HEROES OF THE ANCIENT ORDER.

[Dalai]

You can easily conclude why there are no interesting and complex strategy games that feature a fast and non-cheating AI.

Galactic Civilizations series. With you level of involvement with the topic of AI it is very odd you could miss the game and A LOT of discussions about it's fast and non-cheating AI.

You have to excel in math and in theoretical physics and quantum mechanics. Writing an AI very much resembles the structures of these subjects, so I highly recommend their study.

Additionally I found while designing the AI that there is so much information that it is not viable to write it down because finding the bits you need will cost way too much time. So you have to memorize everything and transform the design in your mind. Not for the faint-hearted. It is also a substantial barrier for building a team that attempts to develop an AI together.

Ok, let's sum it up. Someone basically genius has to do it. You are obviously the only one able to do it, but you are "normally not interested in entry level positions" because you "picked up a good deal of organizational and management skills as well" and ready to lead the team. Which, sadly, can not be built due to "substantial barrier" of all the rest being basically ordinary faint-hearted people.

That leaves Ubisoft and Limbic in a very tight spot with basically zero chance of success. I can understand why they wouldn't want to hire you, so they can postpone their releases for 3 years only to discover that contradicting conditions can not be fulfilled. Smart business decision, seems like.

What I am doing is building tools and implementing a prototype, kind of a blueprint that can be adapted for other strategy games as well. That's at the core of the business model of TESLA MINDS, but I do not like the idea of a middleware provider too much. So instead we are using the AI to build an advanced strategy game that should appeal to the vast majority of the turn-based community. This game will be HEROES OF THE ANCIENT ORDER.

Yeah... future tense. Good luck.

[Quantomas]

Actually, I stopped playing GalCiv because of the AI. Have you ever examined the planets of your opponents after you take them over? This AI couldn't even get the basics right and the conclusion is simply the only way that the opponents appear to be powerful are massive cheats.

Ubisoft and Limbic are not having the success because they simply ignore what made Heroes of Might and Magic successful. It is not only the gameplay, but the whole software engineering process. NWC build their codebase over nearly a decade of development. They made it a valuable asset which they reused from game to game and extended it as appropriate. So the AI and all the supporting code became a real asset. Nival and Fabrice Combounet made the right decision and built H5 on the same base. On the other hand Black Hole and Limbic apparently started from scratch. With other words they had to recreate all the code NWC had as an asset and representing scores of man years. Still, they had no larger budget than Nival and you could practically see it coming that they will fall short.

[Dalai]

Actually, I stopped playing GalCiv because of the AI. Have you ever examined the planets of your opponents after you take them over? This AI couldn't even get the basics right and the conclusion is simply the only way that the opponents appear to be powerful are massive cheats.

GalCiv AI is non-cheating, and it is not an opinion.

Ubisoft and Limbic are not having the success because they simply ignore what made Heroes of Might and Magic successful. It is not only the gameplay, but the whole software engineering process. NWC build their codebase over nearly a decade of development. They made it a valuable asset which they reused from game to game and extended it as appropriate. So the AI and all the supporting code became a real asset. Nival and Fabrice Combounet made the right decision and built H5 on the same base. On the other hand Black Hole and Limbic apparently started from scratch. With other words they had to recreate all the code NWC had as an asset and representing scores of man years. Still, they had no larger budget than Nival and you could practically see it coming that they will fall short.

Now that is opinion.

Here is mine - there is no budget, that is enough to create a good game if the team doesn't care. Horse trails? 4 resources? Convert town? What a joke! What an expensive stupid joke! Let's add a million and it will start make sense!

[Quantomas]

https://www.reddit.com/r/GalCiv/comment ... _ai_cheat/

Do not expect any more answers from me.

[Dalai]

Cherry-picking convenient bits is not the same as answering. Of course I meant GalCiv 2, not GalCiv 3, which should be obvious for anyone interested in AI. It's a well known cognitive bias though, so it's ok. Good luck

upd: from one of the reviews:
AI: It doesn't cheat except on the highest levels. But levels range from stupid and crippled to somewhere near perfect. The toughest two settings give the AI an economic advantage, of 125% and 200%. On "Intelligent," the highest non-cheating level, it watches you to figure out what you're doing and anticipates your next move, but only like a human could and not behind the scenes. That is, a human who was willing to tweak everything every turn to squeeze every last bit of productivity from their civilization.

...
The AI will taunt you when it thinks it's got you over a barrel, and it will beg, wheedle, lie, and manipulate when it's in a weaker position.

You are welcome to pick only one line about toughest two settings to feel right, of course

[Steven Aus]

I've finding I'm really enjoying the latest version of the magnomagus mod MMH5.5, based on your HOMM5 source code. How are you going with your progress, and will the new game be a better version of the skills and perks, and heroes attacking until they run out of units, of HOMM5/MMH5.5?

[Quantomas]

Sorry for the late reply, but the CH notification robot failed me somehow.

Nice to hear that you like the H5 gameplay. I always found it cool, all it lacks is a competent AI. But that is being worked upon. Hopefully we can all play it in the near future. In general things are looking good and falling into place.

The new game will be pattern-based. That is everything in the world will be based on the same logical building blocks, which enables lots of interactivity. Expect interactions between the environment and skills and units. Similarly you will be able to build new traits and spells from the patterns you learn in the game.

[Steven Aus]

Do you know when the game production will start in earnest? I see that you're tackling the AI at the moment, but the company needs to be formed and the game produced, even though the AI is a really big part of the game.

[Quantomas]

TESLA MINDS LTD is already set up legally.

The main issue why the development is slow is that I only get a couple hours to work on the code each day, which is primarily a financial issue. There isn't that much left to do, at least compared to the entire task, but things have to change if we really want to move forward.

The plan is to set up a website to test whether there is demand and enough support to finance the project. I doubt that our diminished community will be able to do it alone, so we will need to spread the word to a much bigger audience. At least now, there is no other feasible way than to finish the AI first. It makes all the difference between "this is the guy working on the AI since 2009" and "play the brandnew 3.2.alpha and witness for the first time an AI that is capable of strategic planning".

Fortunately, I am not impatient and seeing the monumental task building such an AI in every detail gives me all the confidence I need that this work is bound to be successful. A historic feat in the making.

But beyond that, I don't see an alternative to working until it is done. It's a bit of a pity because there are some good guys who would like to work on the project, and given the state of the industry in general and of Might & Magic in particular, it seems like a perfect time to get going.

[bitmaid]

Looks like you've got it all figured out. Keep us posted & good luck

[cjlee]

Guys...

I have been following this thread for a while by now, but didn't feel I could comment for the past few months, because it really was going over my head. (EG Dalai's talk about Quantum Mechanics).

Heroes of Might and Magic is very different from first person shooters and the other games that crowd the market. I believe it is because it is very appealing to the most intelligent people.

Being appealing to the most intelligent people also means that the potential market is small. I do not believe the lower 80% of the population (in terms of IQ) is able to play HOMM. There are too many things that challenge the brain. It simply is not good recreation for the average construction worker or manual worker.

Not to disparage anybody, but less intelligent people play less intelligent games. Simple as that. And less intelligent games also need less intelligent AI. There probably isn't a whole lot different about the AI of various first person shooters.

If your AI is to prove a challenge to smart people, it needs to be smart also. That's not at all easy to make.

I can't provide solutions to you guys. I dabbled in programming in college, and nearly failed. I sucked at making algorithms. I know I'm only cut out to be an end user. And while I view your efforts with awe, ultimately I know that I can't contribute much. I can only stand on the sidelines and cheer your efforts.

But Quantomas, you, magnomagus, and all the other modders in the Heroes V world are fantastic. I hope one day you can produce a good Heroes game, and that you can get paid for your work via crowdsourced sales. And yes, when that day comes, my credit card will be ready.

[Panda Tar]

I can't provide solutions to you guys. I dabbled in programming in college, and nearly failed. I sucked at making algorithms. I know I'm only cut out to be an end user. And while I view your efforts with awe, ultimately I know that I can't contribute much. I can only stand on the sidelines and cheer your efforts.

Most of us are in that situation, which feels powerless and helpless somehow. If we all had more power on that question, surely HoMM wouldn't have gone down the fail lane. However I hope this endeavor to succeed, I'm going to watch holistically, while looking for another game which would appease my feeling on that genre of gaming. Still, I'm not inclined to play anything of strategy niche at the moment (even having AoW 3 and Endless Legends to play) – I would, of course, if it was Heroes. So I guess I'm attached to this franchise regardless all the ill making over it.

And I also would support it, if crowd-sourcing had to be. As long as this new game felt like Ubisoft never held it tight.