H7 vs taking things into our own hands

[Quantomas]

Heroes of the Ancient Order is the title for a potential spiritual successor, an analogue to a Heroes 8.

Certainly, if I could get Ubisoft to do a H5 platinum edition I'd insist that it wouldn't be a simple re-release but upgraded in major ways. It wouldn't sell otherwise.

In some ways H5.5 is incompatible with the AI that I developed. The AI does millions of computations of paths with a deep lookahead to weigh the pros and cons of actions and to determine the best path to choose. If you simply write some scripts to override the AI's evaluations, scripts that technically do no lookahead at all, what kind of gameplay do you get?

The features H5.5 has are nice, but IMO features need to be implemented properly and fully integrated with the AI, if you want an outstanding strategy game. And that is technically the reason why I do the work, that there aren't truly that many strategy games worth playing. Too many fail at the AI, even well received titles like Endless Space or Endless Legend quickly loose their appeal once you realize that the AI is either a push-over or a cheat. The game I came back to most often is Master of Orion 2. It's old now, and I have played it too often, but you get the picture. Heroes VI and Heroes VII aren't even strategy games in my book.

[Galaad]

Ok that's good to hear. Personally what I do love about HOMM is not only the strategical element -which ofc we agree has to be present- but also the atmosphere. To me the original h5/tote is unplayable because I find it visually horrendous, so in such case I would insist it is very important that at the very least Xuxo's textures can make it. If doable, I think it would be worth it to find a way to implement H5.5 enhancements as well, then you'd really have the ultimate H5 game IMO
A good argument would be their mistake with H3 HD which was only RoE HD. A complete edition would have sold much better, and so would have been the reception (I think most if not all the steam negatives come from the fact it's just RoE). They did argue they don't have the source code, but if Polish students with Hota could reverse engineer so should be able to do the staff of the market giant they are in the game industry, I think?

Although I'd be much more looking forward to Ancient Order, if it has the same new AI but another engine, more up to date graphics, different towns and new lineups etc, but that would be crazy work maybe requiring a kickstarter of some sort, I don't know if you work alone but if so that would take forever to make, you need a team for modeling another for animation another for scripting... Have to say there's a lot of mystery going around that potential spiritual successor so hard to figure how realistic it is.

[mr.hackcrag]

To me the original h5/tote is unplayable because...

I revisted H5 recently and one thing that stood out to me is how significantly imbalanced spells/abilities affect combat. I'm not a fan of the take control of enemy spells even if they can be countered with cleansing. For example, seducer skill wins matches so easily that I've never once considered why I would ever choose the chain shot upgrade. A lot of the alternate upgrades seem like one is almost always superior to the other, even with 5.5 version.

In addition to better AI (especially for combat), I would love to see properly balanced units and spells.

[Galaad]

In addition to better AI (especially for combat), I would love to see properly balanced units and spells.

I think that's an utopia, both h3 and h5 are very imbalanced in many aspects yet both their gameplay is highly praised. I think imba is fun. These games were never meant to be e-sports. That said some tweaks when well thought-out definitely improve the experience and IMO 5.5 did a pretty solid job on that (as well as Hota up to 1.5 for me, after I don't like much what they did with spell research or removing resistance).

[Panda Tar]

The imba thing I thought nice to play was H2 style of gameplay, when there was 'no problem' not having upgrade for all troops or not having an uber powerful unit in Castle's faction. But I concur with the fact of balancing upgrades of a same unit, or there is no point having differences at all, as filler content.

[Pol]

H4 was also giving you a nice choices
H3 is too simplified.

[Galaad]

Roughly, just give me h2 towns and artstyle with h5 gameplay but h4 magic system.
And LOTS of content.

[Pol]

Any additional content is up to map makers & modders. For creators is the engine and show off demos. Rock'n'Roll baby, it's avalanche effect!

[Panda Tar]

Heroes of the Ancient Order is the title for a potential spiritual successor, an analogue to a Heroes 8.

Still, this is the only and THE ONLY sort of interest I have or would have in anything related to Heroes games henceforth. I'm really not interested in the next Ubisoft's screw up. Having those many years of work on the scale of something as fortuitously unstable as Ubisoft is not something I would encourage anyone. So, I do hope Quantomas finds his own way into bringing his work into light by other means.

[mr.hackcrag]

Whatever happened with this?

[Quantomas]

Very good. Maybe it's time for some additional explanation to put that in context.

When I was starting to work on the HoMM AI, back in 2007, it was a tremendous help to review the original AI code for Heroes 3 written by Gus Smedstad. Particularly the algorithms for ordered tile processing and the priority queues proved invaluable tools. What I wrote back in 2011 regarding AI for strategy games still holds: http://www.bonddisc.com/ref/h5/ai.htm

That's a line of development I could have pursued, simply building on the best practice of AI to date, and polish Heroes V.

When I started this AI project in earnest, in 2011, I began to look at the scope of what Artificial Intelligence truly means. In terms of predictability and a tool at the core of strategy games, first, but as the work evolved it became a question of building a machine that can analyze a situation and come up with a plan to improve it. That's really the core challenge in the quest for artificial intelligence, and as this is yet unsolved, it requires work, lots of it.

The Heroes of Might and Magic games are not only unique in the gameplay they offer, but they are also rich and diverse in the strategic challenges they can offer. In a way they are an ideal touchstone to examine what type of challenges a machine needs to be capable of handling properly. Everyone who has played HoMM knows that in principle you can make a decision which way to go at every tile you traverse, in which direction you go, what the travel costs of the different paths are, which objects to visit, and much more. It's easy to see that even in a single turn, especially if you have multiple heroes, there is a large variety of paths you can choose. If you look beyond one turn, the number of possible variants grows exponentially, and if you look at many turns, let's say 40+, the number of variants is practically unlimited, you stare into infinity.

The first years I worked on multiple approaches to get a grasp on the entire space of variants and to figure out methods that attempt to hone down the variant space by focusing on the more valuable targets and possible opposition, mostly by looking at probabilities and statistical distribution of forces. But in the end it turned out that the number of variants still remains unfeasibly high and that the predictions are of little value.

You might think these are years wasted. Not at all, the grasp of the subject and what matters grows steadily. It became clear that you have to understand the mechanics, how different variants are evaluated to prove their advantage, to arrive at a way that is feasible to process. Again HoMM is a fantastic touchstone for this work, because you have multiple factions that vie for multiple targets, with a wide variety of tools at their disposal, but ultimately it's all decided in battles.

As fate would have it, three years ago I had a severe poisoning and had to do a lengthy detox. Luckily the medicament I got was a star, and it not only helped me to recover but to cure ailments I had no hope to get rid of. You know, the ailments that you accumulate during your life and that you simply attribute to the inevitable progress of aging. But it didn't stop there. Once you begin to come back to your senses literally, having them much more refined allows you develop a feeling for yourself and to identify what is good for you and what is not. That goes for cloth, living enviroments, the food you eat. All in all it allowed me to put my life on a different path.

I don't know, whether it was a result of the detox or simply that the work on the AI had matured, I could see more and more how the variants reorganised themselves, how intelligence in algorithmic form could be used to make the variants compete and how they could be processed efficiently to arrive at solid predictions. The mechanics itself.

This was a long way. What made it so arduous were not only the infinite variant spaces which structure you have to grasp abstractly, but also that there is an infinite variety of potential solutions for the tools you need to develop to describe the variants and their dependencies. Somehow you have to be that proverbial rock in a stream in a storm and to weather all to grasp the algorithms that work from the myriad of possibilities that whirl around you.

Anyway, this work on the AI core algorithms is done, mostly. There is a bit of work left, a good day to write the AI core down. So all good.

What we get from this is an AI core that can almost instantly come up with a good estimate for what the heroes of a faction can do on map. This should give us fearsome AI opponents that can nearly act without delay. You know the H3 AI turn response times, which where mostly instant. On top of that the AI can spend more time to probe variants more deeply for the veteran players who desire that special challenge.

Please don't forget that after the AI core is done, the HoMM mechanics need to be implemented in detail to make the general AI algorithms work in tandem what the heroes on the map can do. While that will take some time, it's a much more straightforward work, and technically identic to the work on the 3.1j AI I did back in the day.

With all that work and the arduous climb to solve the challenges that drive the new core AI, I regret a bit that I didn't do a Kickstarter in 2011, because the work was a lot harder doing mostly everything on a shoestring budget. But then again, I would not have guessed that this work would take that long.

For all the people who supported me financially, thank you very much. Know that your contribution made a difference and was not in vain. The best may be yet to come.

[Pol]

Thanks Quantomas, seeing your site that's a lot to read. Though you probably should update it with a new information?

When I was in WoG Team, I was dreaming about simple AI libraries. Well, this is still a future. And generally I'm continually surprised like game developers decide to hone visuals instead of story and AI.

[Panda Tar]

Great to read all that, Quantomas. As they say, Lakad Matatag!

[Quantomas]

@Sebastian
Actions speak louder than words. Thanks a lot. I was indeed worried this month, so it helps! If you work on something big you need faith, and so far it has always worked out. Not the wrong faith, hey it will surely be fine, won't it? But faith that if you do your best, you will find a way.

@Pol
Do you rather want me to maintain the site instead, so that everyone will have to wait longer for the AI?

Regarding CH, I like the site. I would give CH a shot to use it as a forum for the AI project, if two requirements could be met. First, your notification of PMs received needs to be fixed. I didn't see a PM from a friend for half a year, because your notification did not work. And on response the friend didn't see my PM for 3 or 4 months. That's abysmal. There is a bug in phpBB that causes this, so you would need to configure the board to notify your users on *every* PM received. The other thing is all posts need proper time stamps. Right now you don't see the year, which has been confusing in a number of instances. Maybe when the AI is out, we can generate enough traffic that you can dispose of the ghost users, but that is not a requirement.

Simple AI libraries that offer you any function you want to be performed in a heartbeat? That's wishful thinking!

@Panda Tar
Human beings have to set out and achieve something! Otherwise we are just human cattle.

[Groovy]

The other thing is all posts need proper time stamps. Right now you don't see the year, which has been confusing in a number of instances.

The date format can be configured in User Control Panel. I'm using "d F Y, G:i", which shows the timestamp of your last post as "17 August 2018, 16:31".

[Quantomas]

Cheers, Groovy!

Maybe it should be made the default.

[Pol]

Ok, let's make some little offtopic comments:

Maybe it should be made the default.

CH default is: M j Y, G:i

Which shows "Month DD YYYY, HH:MM". Documentation format is here.

There's good to know that this is not phpBB default. There was a discussion about that in 2012? Sigh, I don't remember but from that moment on, our format was set up as a new default.

First, your notification of PMs received needs to be fixed. I didn't see a PM from a friend for half a year, because your notification did not work. And on response the friend didn't see my PM for 3 or 4 months. That's abysmal.

From the point of phpBB the e-mail delivery works perfecly, all mails are sent.

The problem lies in mail protocol in generall and numerous anti-spam protections which are being used. The board as ours produce a tons of confirmation e-mails, some people set on notification, then forget it and then mark it as a spam. Which always get us temporarily on some blacklist.
Also, as time goes and if you have easy password, your account may be compromised. The last remnants of it can be seen here. And for that reason is good to check our Bug thread regularly.

And now back to the topic:

Do you rather want me to maintain the site instead, so that everyone will have to wait longer for the AI?

Purely rhetorical question..

However it would be good to document it somewhere, and there's gold rule, that sooner is better.

..libraries
Yea. When people create mod, they should also modifiy the AI to handle it correctly. And having AI libraries would speed that up, tremendously.
(Apart from starting to build their own AI.)

[Quantomas]

First, your notification of PMs received needs to be fixed. I didn't see a PM from a friend for half a year, because your notification did not work. And on response the friend didn't see my PM for 3 or 4 months. That's abysmal.

From the point of phpBB the e-mail delivery works perfecly, all mails are sent.

The problem lies in mail protocol in generall and numerous anti-spam protections which are being used. The board as ours produce a tons of confirmation e-mails, some people set on notification, then forget it and then mark it as a spam. Which always get us temporarily on some blacklist.

Negative. I don't use anti-spam and the mail delivery protocol, which uses technically a handshake, had never failed on my mail delivery as far as I can tell, for twenty years across numerous providers and 10,000+ mails. I am aware of a couple more instances in which your mail notification failed. And I have seen the same bug on other boards that were configured similarly. I studied CS and know the ins and outs of the tech, so you won't get away with excuses here.

[Pol]

Ok, another offtopicness:
You may not be aware of anti-spam solutions which your provider use and these may apply even if you decide to switch them off in your e-mail. You may simply not know about them.

If you are completely on your own solution or you are really sure that there's no anti-spam filter along the way, then you need to send me some details about the mail, not older than a three days. If that will occur I will check our logs and see it.
(Nonethless some things cannot be seen that way and I will need to know exact steps to reproduce such error.)

For that please use PM or Bug thread - but there just without details.

At this moment I stay with my word that there are no problems, as I don't have any details from which I can check further. Neither anyone else complain.

And as you said, this software is used on many forums, so such generic problem how you try to outline, would been likely complained by many people before.

That's about current state of PM. Notifications are a bit different matter but I would need some steps to reproduce them. Historically speaking, yes in the past there were (one or two) problems with them, which were usually fixed with another release.

I hope that this is enough for now, let's not clutter this topic too much. It's interesting and deserve to live on its own.

[Blackfayr]

I really like those Ideas is there any eta on when will they be implemented into the game?

I think Improving the AI is what missing from making this game from a good game to a great game.