Master of Magic 2 Wishlist For Slitherine

[haphazard1]

As soon as I see/hear 'We need MORE money for this/that', I turn around and walk away laughing.

The next great breakthrough is always just one research grant away!

To be fair, this is true of pretty much every field, not just AI. As a field AI has made tremendous advances over the past several decades. Stuff that my professors talked about as the future of AI back in my CS undergrad days has actually been achieved and is in common use today. And I expect the field will continue to produce more and better stuff in the years ahead. It just always seems to take rather longer than the enthusiasts predict. But eventually the tech goes get there.

Cloud-driven 4X game AI is likely to happen. I think it is going to be a while yet before we see it, though. And until then game devs still need to come up with something that works for their games.

[Jeff Graw]

Many of the reviews (even the bad ones) all seem to talk about how hardcore the AI is so that's promising.

Nah. It is not promising because Pandora AI in its present state is completely fan-written. Proxy Studios (Pandora devs) are ruining now the WH40K universe with their Gladius game (based on Pandora engine). These guys can't make a decent AI. As well as a good and original game at all (Pandora is a cheap and shallow rip-off of Sid Meier's Alpha Centauri).

And Slitherine don't develop games no more, now they are publishers only. And most of their published games have big problems with quality, alas. Broken promises, lack of support, developers abandoning games and so on.

The guy who wrote the Pandora AI (Ail) is also involved with Gladius in a more official capacity. At least, that's why he told me he was stepping back from Dominus Galaxia.

Even if they aren't currently making games, Slitherine does have their own developers and maintains their own game engine. So they probably have the ability to do everything in-house, although they'd probably need to do a hiring push. Whether or not they do that is another matter of course.

[Jeff Graw]

-Whenever you implement a feature or game mechanic, immediately implement the full AI for it. Don't use game mechanics that aren't AI friendly. The AI should be the responsibility of one person, otherwise it'll end up doing self-contradicting stuff. This person should have the right to veto any game mechanic decision if it's not playable for the AI. (Yes, this restricts the options quite a lot, but it's necessary.)

I largely agree with the sentiment, although a veto may be a bit much -- really depends on the person and the extent to which you're confident in that person's design ability. Whether or not you want more than one person working on it is also going to be situational. But ideally, you have at least one individual who is primarily tasked with AI.

But besides that, "not playable for the AI" seems a bit reductive. You can divide obviously problematic mechanics into several groups:

-Too easy for the AI to do well because of a lack of strategic depth. E.g. A mechanic that is just a bunch of complex mathematics, or otherwise requires rote repetition.

-Is exposed as being unfun when the AI plays optimally. E.g. the game board becomes overwhelmed with AI spam. Tech trading is a good example of something that becomes unfun if the AI uses it properly. But systems like tech trading also spice up the emergent narrative of the game, so most developers just hobble the AI in those cases. (For Dominus Galaxia, I've actually designed a tech trading system that should less exploitable and will debut in the next release).

-Is one-sided. E.g. traditional diplomacy modifier implementation. The AI is compelled by the diplomacy modifier, but the player isn't.

But merely being "not playable" isn't so obvious. First, it's a misnomer. What I think you really mean is "extremely difficult to play well." And this is a gradient. To make things worse, although this isn't always the case, there seems to be a correlation between strategically deep systems being more difficult for AIs than they are for humans. So do strategically deep but computationally difficult systems get the axe? Or does that just mean that one has a challenge for AI development they should attempt to meet over time? A cost-benefit analysis is needed.

Don't hide information. This is a strategy game. We need to know about all stats and unit abilities, we need to know all game formulas. All of it needs to be readily available by right clicking on the appropriate part of the UI.

A certain amount of information is always going to hidden. Ideally you want to display the most information possible with the least amount of clutter and noise, but that's pretty much impossible. The hard part is getting as close to that ideal as you can.

That's what the card game I didn't like did. Making decisions blind is not acceptable for me as a concept in a strategy game.

Except in this genre that's not an option. How am I supposed to attack something if they might receive a move order this turn and the two armies miss each other when the move is executed? 

The effect of simultaneous execution is really dependant on the specific game, and more specifically the movement system. So, sequential vs simultaneous in MoM (or any other tile based game) makes a huge difference. Not so much in a point-based movement system like MoO.

But everything else being equal simultaneous movement has huge advantages. More streamlined multiplayer where players don't need to wait for someone else to finish before beginning their turn, or the ability to have a multi-threaded AI that executes in the background and is usually finished by the time the player hits the turn button -- simultaneous execution can help a lot with pacing.

That said, it might be possible to make a good MoM game with simultaneous execution. It would be a bit different, which isn't necessarily a bad thing like you say, but you could try to design around some of the apparent issues. For example, you could have a distinction between move and attack, where moving into a tile will create a combat encounter when two opposing units share that tile, while attack will move towards a specific unit. E.g. if I want to attack an army and that army moves during turn processing, my army will move towards the enemy army itself rather than the tile it occupied previously. Of course, you can't really know if something like that changes the game to the extent it doesn't feel much at all like MoM until after the fact. 

But I'm doubtful anything so novel will be attempted. The industry, especially the large games, seem more or less content with one tragically flawed extreme or the other.

Even in CoM the AI isn't allowed to behave like a human in some aspects because it would be too unfun. Spell Blast, Disjunction, and all of the diplomacy system being a few examples. Diplomacy uses the concept "the human can do anything they want but everyone else is bound by the rules/treaties". That doesn't work with multiple human players. So an entirely new concept of diplomacy is needed, that's a lot more restrictive on human players and multiplayer friendly. And this is just one example. We are drifting further and further away from a MoM sequel and closer to an entirely different 4X game that has nothing to do with it. You can't try to close a 25 years gap and still call it a sequel, it's a too big jump at once. That becomes a completely different game. Maybe better, maybe not, but definitely not a sequel. 

Restricting options is one path. Most of the time I think it's better to come up with better options, or add interesting constraints to those options. AIs tend to suck at diplomacy, and for Dominus Galaxia the initial thought was to restrict the diplomacy system to a fairly simple subset to level the playing field, but in hindsight this removed a lot of the soul from the MoO 1 formula. I've created a new diplomacy implementation for the next release that adds in a lot of stuff, but is also more intelligent about how each option works. For example, asking an AI to declare war on a third party has traditionally been very exploitable, but in this implementation it's only possible when there's an alliance between the parties, and the request is more of an obligation. Refusing your obligation can then have a host of negative penalties, and the AI is also free to refuse if it feels that those penalties are preferable to declaring war.

The big tension here is between designing an AI that plays well, and one that role-plays well. The holy grail is when you can design things so that an AI that playing well reinforces its role instead of subverting it.

Speaking of business, I don't think making a quality game and one that's viable as a business is possible at the same time. The two are mutually exclusive. Quality games require time, effort, dedication while being a better game doesn't necessarily sell better so it's not worth it financially. So business is not an argument. If that's what they make it for, it'll be a bad but profitable game.

Nah, I don't think they're mutually exclusive even if it's difficult gain both. It's difficult to make good games as a large team for a host of reasons, including philosophical differences between members, too-many-cooks, that coordinating a large number of people requires a degree of enforced rigidity ("Just follow the GDD, don't think about it or try to iterate"), and you're more likely to be led by people who are undeserving or otherwise ill-fitted of their leadership positions -- the Alex Kurtzmans of the world only thrive in sufficiently large organisations. 
 
On the other hand, solo or small teams with limited funds obviously have a disadvantage everywhere else, and those are the areas are where larger, better funded teams excel. But there's nothing about good on-boarding, productions values, marketing, etc. that contraindicate a good game. And a game with those attributes that is also good should sell a lot better if everything else is equal. It's just extremely difficult to do.

[Seravy]

That said, it might be possible to make a good MoM game with simultaneous execution. It would be a bit different, which isn't necessarily a bad thing like you say, but you could try to design around some of the apparent issues. For example, you could have a distinction between move and attack, where moving into a tile will create a combat encounter when two opposing units share that tile, while attack will move towards a specific unit. E.g. if I want to attack an army and that army moves during turn processing, my army will move towards the enemy army itself rather than the tile it occupied previously. Of course, you can't really know if something like that changes the game to the extent it doesn't feel much at all like MoM until after the fact.

I honestly don't think such a system is possible - you either block the movement of the target of the attack to ensure they can't slip away, in which case an enemy might block yours so you fail attacking anyway - or you don't. In either case you have no way of knowing if the attack you ordered to happen really goes through or the enemy can get away. MoM also has variable costs for tile movement and a hard limit on how many units can be one a tile at once - which is a fairly important limitation as number of units is highly relevant in combat.
Surprisingly it worked fairly well for MoO but it had a completely different movement system. Or maybe it was because the AI didn't try to actively dodge your fleets and bomb your unprotected planets, idk. Haven't played that game in a very long time.

For example, asking an AI to declare war on a third party has traditionally been very exploitable, but in this implementation it's only possible when there's an alliance between the parties, and the request is more of an obligation.

Nice, that's actually fairly similar to what I have in CoM. Although the AI can't refuse unless they have another obligation that contradicts it - an alliance or peace treaty with the target.

Nah, I don't think they're mutually exclusive even if it's difficult gain both.

The issue with that isn't so much what teams are capable of vs individuals, but time. A commercial game has to make money - within a reasonable amount of time. The money made doesn't linearly grow with game quality. Past a certain point, any additional effort invested will be a loss of profit. And unfortunately, at least what I call a good game, is very far beyond that point. Also, what makes a game sell and what makes it good generally don't overlap. Those finer details that make a game great or average are all stuff players will only realize after they spent their money on it. But it doesn't even get that far, I mean, they put minimal effort even into the basics. Maybe I'm wrong on this but there has to be SOME reason why almost every game I tried playing the past 2-3 years had fatal flaws. Playing games nowadays feels like a chore, not something I do for fun... it's like, whoever made them never tried to even play their own game, they are so unfun to play, missing something completely obvious. And these were major titles from major companies... just an example, the material needed to craft pretty much every unlockable equipment in a game wasn't available until after halfway through the second playthrough. Why would I want "basic sword+1" after beating the final boss, what were they thinking? How can anyone miss THAT?


I completely agree with everything else in your post.

[Blake00]

Well.. so much for this being a wish list thread to give to Slitherine.

[Domon]

"Just hire Seravy,  cover him in money if he resists"

[Arnuz]

It was, unfortunately "my wish: well done multiplayer pls" seems to be contentious on this forum despite the year, century and millennium ^^
Decent AI is "just" the result of that approach.

Haphazard: good find on the cost! There weren't many articles on that and i missed that. Interesting!
Not convincing though. First of all it's estimated by externals. But even assuming they got it right:

1. We're talking about the AI beating reliably the world champion. Obviously this isn't needed or even desirable. With the much much cheaper imitation learning, where AI imitates player behaviour, deepmind [URL="https://www.alexirpan.com/2019/02/22/alphastar-part2.html]was still able to reach the StarCraft ladder mp gold layer (the top of the ladder)[/URL] which would be more than enough.

2. Even when wanting to beat the world champion... That estimate is the cost of a technology that hasn't yet even left r&d. It's obviously going to change quickly.
And, just to give you an idea of HOW quickly it changes, well.... You've found an article concerning alphago from 2017 estimating $35m for simulating 72 hours of go.

Well, I've found a similar one for alphastar: https://medium.com/@mattiasappel/alphast...3e11078bc8

1.5 years later it costed $25m to simulate 60000 years of StarCraft. That's a reduction of 6 orders of magnitude per year.

[Blake00]

Oh enough about bloody AIs lol!!!! I posted wishlist requests in a few places and got great feedback and positive conversation. People came forward, listed the things they wanted, and left it at that. While here (the place I had biggest and highest hopes for) I've got 5 pages of bickering about super AIs and pessimistic Slitherine bashing which HELPS NO ONE.

I can pretty much sum up the 5 pages in 2 wishlist items for Slitherine:
1. Add multiplayer mode and take advantage of modern improvements to such as simultaneous turns BUT still give option in settings for traditional turn based. Leave single player as turn based or once again provide settings option for both ways to keep everyone happy. DONE.
2. Make a good AI and don't f*ck it up, see examples above. There DONE. If you really wanna debate about different AIs do it in another thread, not a MoM feature wishlist thread.

Maybe another 4 or 5 posts here (that weren't part of the bickering) that I can use too. Thank you to those people. I'll make sure I copy across your suggestions.

[asw3]

I can't remember a single sequel to a classic game which didn't disappoint. Even the "spiritual successor" sequels by a different name are failures, Path of Exile coming to mind as the single one that matched and in some elements outdid its ancestor (and GGG outdid Blizzard at their own game). While said disappointments and failures I can think dozens of, maybe even hundreds at this point.

I doubt Slitherine's MoM game will be decent, let alone as good as MoM, and people that think it can be better probably post from the madhouse...

The only wish I have for Slitherine - release the game's source code. MoM doesn't have the popularity to quite go the OpenXcom/Daggerfall Unity/OpenMW etc. route by itself (even though some attempts and progress has been made, they are not quite on the same level as the ones mentioned, no offense to the people behind them). MoM doesn't even get the effort that Rocco and co. are putting into MoO2, let alone to get something on the level of what Heroes III is getting with it's Horn of the Abyss and HD mods (though CoM is doing as great a work as it can under the circumstances I think). So the only hope I see for a classic MoM experience, but improved, is for the source code to be released, which might create enough interest for a source port with extensive editing and modding capability to happen. Anything else will result in disappointment imo.

Though it could be interesting to think about changes to MoM that I would want to see in a MoM like game.

Well my two biggest problems with current MoM is the level if micromanagement tedium in regards to town and unit numbers (and proximity in regards to towns) on the map.

So what I would like is a style of game more oriented towards "tall" playstyles, where players don't own more than a couple of cities, maybe a dozen at max if they conquer the whole map and every AI city in addition to their own, but offer more variety and choice about how you develop those cities. So for example lets say you have 2 Barbarian cities, each one should offer so much mutually exclusive choices in regards to development (sometimes only possible because of city location etc.), that they function differently enough to one another, for example they might have no or little overlap in units produced. That would be better than the non choices which we have to do dozens and dozens of times now in town management.

In regards to units on the map I personally think the Heroes of Might and Magic style of heroes leading armies, and you having only a limited number of heroes that you can employ at a time, to be simply superior. But it has to be less than HoMaM's 8. Maybe it can be related to map size, or with the fame categorization of heroes, or related to how powerful summoning spells you know, but each player/AI shouldn't have more than 4-5 heroes at max on the map, each having a stack under him. This will lead to two improvements - no more playing whack-a-mole with dozens of weak AI units running around, now each encounter will be more challenging, as the AI will pool its units more, and also open up more interesting choices about unit compositions, since now you being able to field only a limited number of units at a time, you would always want to have the strongest ones possible for your strategy, which of course means having the resources to get and maintain them, which feeds into economy etc. This aspect obviously exists now, but is muddied by the sheer quantity of garbage cheap units that a player/AI can push out. Obviously increasing the power disparity between units will make the game even more interesting in this regard.

These changes of course lead to new and interesting design possibilities as well. You can bend these rules in certain situations - for example one race might have mechanics that let it have many more towns, or a specific hero might have a bigger capacity for leading armies (so bigger stack under this specific hero), and so on which would lead to a level of variety that currently doesn't exist in the game now.

These are my main ones, but there are of course other minor things that can be done. Like Sulla has said in his MoO1 videos, MoO2 in many ways is a sequel to MoM rather than MoO, so a lot of ideas and mechanics can be taken from there and adapted, or minor things from other games, like each town (or towns of specific races/realms) having a powerful guardian unit like in Disciples II, more varied map design, including impassable terrain, maybe terrain elevation, introduction of zones, improved map object quantity and variety, and generally things to do on the map which aren't waging war with the AI, etc.

The main thing to note, that I don't think many people realize, is that good design comes from limitations. If you have a maximalist approach to number of town, units and other aspects of the game, you are actually shrinking the design space you have to create meaningful and interesting choices that can keep you playing for a long time. I think Seravy said it best with the 20% thing, but getting the other 80% will be done by general subtraction of things already in MoM, and reintroducing them sparingly only in certain situations to certain races/realms etc.

I know most people won't agree or like my views, but it's my two cents anyway, not that any of these discussions matter for anything in reality.

[Seravy]

So what I would like is a style of game more oriented towards "tall" playstyles, where players don't own more than a couple of cities, maybe a dozen at max if they conquer the whole map and every AI city in addition to their own, but offer more variety and choice about how you develop those cities.

I would love that, but I don't see it actually working.
Most of the time the human player is the aggressor and the AI needs quite a bit of time in game turns to send troops at them and retaliate (even more so if the AI actually respects the player's borders during peacetime). Not enough cities means the AI goes down before their transports even reach the player's shores.
Of course, there is always a workaround - if conquering cities is a slow progress (like, you need to siege them for 10 turns each) that works but then we are back to square one - conquering cities takes painfully long.
Nonetheless, if there is a way this can be made to work, it's definitely worth doing so.
...actually I might have a good idea to solve this. If players lose the game only after losing all their units and cities, instead of only their cities, this can work. Then conquering those few cities the other player has doesn't also make them lose their main armies, so those armies will attack the aggressor and likely give the AI new cities to work with. Of course this mechanic doesn't really work if units have maintenance (you can't pay for it without cities) so this isn't as easy to do as it sounds, but it might be the correct direction nonetheless. Upkeep is a pretty bad game mechanic either way as human players absolutely hate it while AI's can't really work with it and need to cheat their way around - they lack the foresight to be able to know when they can't afford building more units. So in this regard, HOMM style gameplay might work well (there was no upkeep there and you had a few turns to recapture your towns before losing), however, I disagree on HOMM style gameplay in the number of units. I have been playing that on large maps and it's a horrible whack-a-mole even worse than MoM. If your few heroes are out conquering enemies, then their heroes can get a free pass at conquering yours. There are more cities than heroes, so you can never defend your stuff properly (yes, you can leave garrisons but a higher level hero makes the units in their stack worth many times more so...any battles you do without a hero present is a net loss in dead units anyway, and against the enemy main armies it just doesn't work at all) as long as the map is open enough that heroes can just walk past each other without actually meeting. (Also, as heroes generally move at the same speed, chasing down that one enemy hero that's 2 tiles ahead can take many many turns, keeping one of your significant stacks busy if they actually bother to move away from you.)
Sure there are fewer stacks but you also lack the amount of stacks to stop anything and...it's per player. So it's your 8 stacks against the 4*8=32 enemy stacks. So it's pretty difficult to keep up on large maps.
This isn't noticeable because campaign maps simply don't have that setup on them (usually only 1-2 enemy factions) AND you can't build new towns in HOMM at all, but try an XL size map with 8 enemy factions and 30 towns, it's a nightmare.
(Note, I do like HoMM, it's a great game...but it's also a very different game. Also, I mostly played HOMM 2, and 4, maybe newer versions are different.)