HoMM is disappointing

[Humakt]

Yeah, but at least it was just the factions that were unbalanced, not the whole concept of the game (the way heroes were done on the field in H4 was unbalanced in favour of them).

It wasn't just the factions. Both secondary skills as well as spells were terribly unbalanced, from overpowered (Dimension Door, Town Portal, Logistics) to practically useless (Eagle Eye, Mysticism, Hypnotize).

Heroes in H4 weren't nearly as unbalanced as you claim. In many maps hero power rapidly increases compared to creatures, thanks to vast amount of experience compared to the rest of the resources (gold and dwellings primarily). In low experience, high resource maps the tables would be turned.

It was very unlikely you could manage well without creatures especially at start. Usual army setup, I used at least, was three to five heroes with rest being creature stacks.

Biggest problem was low pricing of immortality potions (should be 3000 gold a piece). Retreating with heroes was way too easy, enabling hit and run tactics too much (though if you failed to retreat in time, game over).

[Banedon]

Agree with some of humakt's assessment, disagree with others. Logistics was pretty OK because every race could get it. Dimension Door wasn't, because although every race can get it once you get it it's impossible to defend against. Eagle Eye, Mysticism etc are fine because everyone can just avoid it. There are useless skills in the game, but that's OK as long as what's left is playable. H4 was pretty balanced except for one faction (Death). Sure heroes stomp armies, but so what - you just switch to using heroes and the game remains strategically viable.

Anyway from first impressions H6 is not looking good for solo play. Maybe the AI doesn't cheat enough or something, but when I invaded his land he was so underdeveloped I had no trouble just walking into his town (he didn't even defend it). "One big battle to end the game" doesn't even happen because all I've got to do now is wait 7 turns to eliminate him, although I doubt I'd be that patient. H6 AI is even more of a pushover than the H5 AI. Oh well. Since early-game expansion is a lot easier now (with all the powerful defensive abilities each faction has) there isn't even a need to practice that. All that's left is MP and the campaigns ... oh well.

@Kristo - I *think* quick combat evaluates directly based on what creatures are in the other army and what are in yours, that is, it does not actually fight the battle. The combat AI is pretty good, although it spends a bit too much time healing when it needs to get close to deal damage. Also if its purpose is to attrition the player's army, they need to spread out their damage, hopefully overwhelming the player's defensive skills and getting a couple of kills.

@humakt - maybe, but aside from rolling over the AI you need to find a human opponent, and how long does that take? I can't commit three hours to playing a single Heroes game ...

[ThunderTitan]

Biggest problem was low pricing of immortality potions (should be 3000 gold a piece). Retreating with heroes was way too easy, enabling hit and run tactics too much (though if you failed to retreat in time, game over).

That would have made heroes harder to use when the enemy units had enough numbers to take them out in one attack...

[MattII]

Do you still get the hero's Tactics skill bonuses after he's been 'killed'?

[Humakt]

Do you still get the hero's Tactics skill bonuses after he's been 'killed'?

No, that's why he is usually the most important hero to keep alive.

[Dalai]

From what i understand you suggested that quickcombat could actually go through a battle, just not showing it on screen. Compared to fight value calculation, such approach would give less predictable results.

I followed some of NWC interviews closely, and they commented on this. They said that there are several hundreds battles calculated each AI's turn. If they were actually modeled, AI turn would take tens of minutes. So it is calculated using some simple scheme.

Can't comment on post-NWC era though.

[ThunderTitan]

I followed some of NWC interviews closely, and they commented on this. They said that there are several hundreds battles calculated each AI's turn. If they were actually modeled, AI turn would take tens of minutes. So it is calculated using some simple scheme.

Say hi to multiple core 3+GHz each processors...

And why would there be so many battles?! Are they including the AI making decisions on what to attack or something?! The they could just use the modelled battles when it actually attacks... (remember, as players we don't care that it's more work for them to make the changes).

[Dalai]

Say hi to multiple core 3+GHz each processors...

Yes, they are here. But they are not AS MUCH more powerful than previous. I can assume 10 times more. Are you ready to wait 5 minutes instead of 50? But implementation is the key factor here, of course.

And why would there be so many battles?! Are they including the AI making decisions on what to attack or something?! The they could just use the modelled battles when it actually attacks... (remember, as players we don't care that it's more work for them to make the changes).

If my memory serves me well, they calculated each possible battle where AI's troops can reach, and they did it several times to find an average (or may be the best) result.

There could be, let's say, 5 AI players with 8 armies each, and every army could reach 2 to 6 neutral or enemy armies. That results in more than 150 battles, each can be simulated 5 times - more than 750 simulations. Sure not every turn is like this, but you could split any of your armies, and combinations became countless. And even 1 such turn may be enough to make a frustrated user.

If I had to solve this problem - I would statistically analyze as many battles as possible to come out with some sort of single calculation formula of possible losses. It IS a very complex task, but it's doable. The benefit would be very quick AI battle calculation on any hardware. The difference with previous implementation would be much more precise results, because analysis would be based on full scale simulations or actual battles. And it could improve from patch to patch as more data would be coming.

[ThunderTitan]

Say hi to multiple core 3+GHz each processors...

Yes, they are here. But they are not AS MUCH more powerful than previous. I can assume 10 times more. Are you ready to wait 5 minutes instead of 50? But implementation is the key factor here, of course.

Processing power doubles every 6 months... isn't that the rule?

So after 12 years it would actually be at least 24 times more...

If my memory serves me well, they calculated each possible battle where AI's troops can reach, and they did it several times to find an average (or may be the best) result.

There could be, let's say, 5 AI players with 8 armies each, and every army could reach 2 to 6 neutral or enemy armies. That results in more than 150 battles, each can be simulated 5 times - more than 750 simulations. Sure not every turn is like this, but you could split any of your armies, and combinations became countless. And even 1 such turn may be enough to make a frustrated user.

Yeah, but like i said, they wouldn't need to do them all, just use the normal way to determine which stack to attack, and then use the more accurate and complex way to determine the outcome to that...


Of course i got side tracked a bit, frankly i don't care about the way the AI does it, i'd be fine it it was just limited to when the player uses quick combat...

[Qurqirish Dragon]

Say hi to multiple core 3+GHz each processors...

Yes, they are here. But they are not AS MUCH more powerful than previous. I can assume 10 times more. Are you ready to wait 5 minutes instead of 50? But implementation is the key factor here, of course.

Processing power doubles every 6 months... isn't that the rule?

So after 12 years it would actually be at least 24 times more...

I don't know if doubling every 6 months is correct, but if it is, then after 12 years, you would have 16,777,216 times more.

Now that I think about it, I went from a Pentium 200 MHz to a P4, 3GHz in a seven year spread. Anyone know how much better a P4 was, compared to a P1? the raw speed is 15x faster (slightly less than 4 doublings), but without a processor comparison, I can't come up with a meaningful rate during that time period.

[ThunderTitan]

I don't know if doubling every 6 months is correct, but if it is, then after 12 years, you would have 16,777,216 times more.

It's actually the numbers of transistors you can easily put in an integrated circuit doubles every two years...

And yeah, geometric progression... i always forget about it...

I really need to get more sleep...

[Humakt]

Eagle Eye, Mysticism etc are fine because everyone can just avoid it.

No, they aren't fine. You'd never voluntarily select Eagle Eye or Mysticism OR First Aid, they do not add any real depth to the game, only coincidentally screw the hero development. Granted they do not mess up the balance that badly (unless you get Eagle Eye specialist as your starting hero and something equally terrible as tavern hero) . There were too few secondary skills to select from in H3 to even further truncate selection with useless skills.

And then there was the matter of hero specialties, which also were horribly balanced.

[Banedon]

They only screw with hero development if you can't select anything else, which you usually can. Starting with the Eagle Eye specialist is silly because it means you random'ed, and why would you? There's no advantage to going it random. If you really wanted to play a random race you could start a single-player game with random race, and then use that race in your MP game, but picking your hero directly.

Creature specialists are strong but again, everyone can use them. Assuming you start with them (which you probably should) then the playing field is pretty equal.

[Humakt]

They only screw with hero development if you can't select anything else, which you usually can. Starting with the Eagle Eye specialist is silly because it means you random'ed, and why would you? There's no advantage to going it random. If you really wanted to play a random race you could start a single-player game with random race, and then use that race in your MP game, but picking your hero directly.

Because I like variety in my games? I do not like to select Sir Mullich every time I play Castle. Sometimes I like to have the dice do the selection for me, and if I get screwed I try the tavern for main hero.

But if you never select certain heroes, secondary skills or use certain spells because they are so bloody useless what's the point of putting them into the game in the first place? I, for one, am glad that there is no Eagle Eye in HoMM IV.

Creature specialists are strong but again, everyone can use them. Assuming you start with them (which you probably should) then the playing field is pretty equal.

Your assessment skill needs tuning if you believe the creature specialists are the imbalanced ones. In small, quick maps heroes with fourth level damage spell like Solmyr are the way to go. On larger maps the Logistics specialists will rule.

I'll leave the Sir Mullich out of equation, he should never been added to the game as such as he was.