A quick primer on how to balance a game and how this applies

[elliott20]

David Sirlin's Lecture

For those of you not familiar with this name, Sirlin is a game designer who has worked on balancing a number of games such as Street Fighter 2 HDRemix, Puzzle Fighter, and has written extensively about a large number of other games. His writing here is not the end all, be all, but I think it provides a lot of good ideas and good food for thought.

the next part will assume you've read the presentation.

so, applying the model described, here's what we need to think about:
1. Viable Options - are there a lot of options that are not trumped by a single most dominant option?
2. Fairness - are all the starting options, despite their differences all start off balanced?

the answer to question 1 is a resounding no. Why? because every single race you play you will always have the one true strategy you need to shoot for. That is, go for the super unit. For high men, that's the paladins, for halflings, that's the slingers, and so on so forth. So what stops people from getting to the super units? well, primarily that most super units have a built in cost structure that prevents them from being accessed in the beginning. this is not a good enough solution as it just means that the only games which are strategically interesting are the early games.

the question to number 2 is also a resounding no. As you can see from your race selection, not all races are created equal, and not all schools of magic are considered of the same strength. While you have the possibility to expand into other races later on, this only further shows how imbalanced the races are, as if the new race has such a great advantage, it would make more sense to just start with that particular race in the first place. (poor klackons)

discuss

[Lucern]

Good post Elliot. The short of it is that I agree with your overall assessment. The long of it is that I wrote a long post a couple of days ago, hoping to tie it together thematically or something. Eh. I'll post it anyway – it's not like we're awash in posts here. I think the best way to balance MoM is first to do nothing to it, see how it actually plays out when it comes to fruition, and then mod it out with different emphasis on different aspects of the game. With my mod, I'm going for variety and expanding strategic and tactical possibilities, so, (see below) simple strategies relying on a couple of spells and one uber-unit are an affront to the variety that the game would have.

...
Thanks for posting that Elliot...it's humbling to see that much written about a subject that on first glance seems much more straightforward. I found myself on the defensive in my desire to add more, more, moar to the game, but I found myself agreeing with him a lot. He loses me in the intuition stage, but his point is taken that balance isn't about solvable, mathmatical problems. Rather, it's an ongoing challenge. I like that he refers to Magic: The Gathering so much, particularly since our own beloved game borrows so heavily from it. It's highly asymmetric but variable in the extreme, much like our own game. Though we'll never have the grace of balance of games like Chess or the much more complicated (and yet, in many ways much less complicated than MoM) Starcraft, every build does have something to offer to the possible ways to play the game. (Even if the main offering is a much greater challenge (ahem, klackons ;) ) I've long admired Master of Magic's style of balance much more than the bland sorts of blanket equality offered by many games of this type. The balance is in the fine details - every unit is lacking in some important ways, every spell school fails you in ways that you wish it had mastery over, and best of all to me, for all of your planning, you don't have as much control over Arcanus as you wish to have. You may plan to go the life/paladins route, but you may find yourself in a world populated with gnolls and under pressure from the other wizards early. You'd be setting yourself back if you didn't adapt to having their relatively quick halberder raiding parties at your command if you wiped them all out and started fresh. I'm most interested in the chaos of a MoM game, providing you with new tactical and strategic situations that you couldn't have anticipated, keeping you on your toes and expanding your thinking about such a familiar game. I'm probably prone towards unusual builds because of this predilection rather than the classic overpowered plays - and there's so much variety in this game for that to destroy. The greatest offense that these overpowered play styles is that they narrow the scope of experience in the game - you've got to prepare to counter them, and that may well be MoM's greatest flaw in a multiplayer game.

It's this in-game contingency where you don't know exactly what to expect (especially with other humans in the mix) that makes MoM IME so compelling where even very complex games that SIrlin talks about can never be. We can talk about races being unbalanced, but even at the middle-game stage, most of my games give me at least 3 races under my command. Furthermore, MoM is in many ways a game of chance, and I bet many of us (me especially) would have a hard time achieving high score 'impossible' victories when you can't play the save/load game. You know this game: Summon champion: "Aww man, the Illusionist? I wanted Mortu." LOAD. Or attacking a sky drake node with your hero stack can be redone over and over until your heroes come out on top without losing anyone. I do it all the time lol...just because I enjoy the game with a bit more control. In a multiplayer situation, however, you've got the anxiety of uncertainty playing against you. You cannot load when the meteor hits you, when you lose that one capitol to raiders you couldn't see (which is now ruins), or when crack's call takes your demi-god Alonna the Elven Archer away. This may cause players to play more conservatively, with defensive, careful moves towards building up empires. Unfortunately, I think it will cause many to play the odds in the grandest sense, as in, that ONE strategy for your empire's makeup that will most likely produce success: Longbowmen, Hammerhands, Paladin, Warlocks - whatever your one trick is, SPAM. Personally, I look forward to seeing how this game pans out online, but it's one I'll play just as much offline when I have more control and time to enjoy the pleasures of empire building, random treasures, and developing hero parties. My own sense in balance in the game is that just about everything is internally balanced somewhat. We may not agree with the balance, and some races seem to get the shortest end of several sticks, but I don't see any 'god tier' or 'garbage tier' races. Playing the gnolls or the klackons is harder than most others, but it's still possible to win. Playing the high men or with life magic is easier than most others, but it's still possible to defeat these choices. All units can be brought down.

I'd also like to suggest that Sirlin is using a bit of an outmoded game theory here: he seems to be assuming that everyone's playing with the same goal: victory, and not much else. Now, I won't say that any of us aren't playing to win, but I'll suggest that there are other factors in our game that are enjoyable whether one wins or loses. There are a lot of ways to play Master of Magic, and yes there are, sadly, some quite overpowered and necessary (in some contexts) strategies. However, who cares if you win if you just took 10 life spell books and spammed Torin and paladins? There are ways to counter any one strategy in this game. They may not be ideal, but it can be done. I don't really want to be playing with anyone who's just going to spam one unit for a quick victory rather than players who enjoy competitive empire-building and spell-slinging as the game plays through. Besides, I bet I could get a good alliance going to wipe out that kind of player :). But that's one way to play, a popular one, I suspect, and I can never fully disapprove of playing a game in a way that it allows you to.

[elliott20]

A good post, Lucern.

I agree, in order for us to really be able to make claims, we need to first see the game played in a multiplayer environment, with highly competitive players trying their best to grind out the best and even most broken strategy possible. The whole thing with about the chaos of MoM is definitely true too, and it is even more applicable than say M:tG, since in M:tG, you at least get to pick the cards you build your deck with. On the other hand, in a game of MoM, for the exception of the standard starter package, just about everything else is random. This means that any traditional model of comparing combinations and tiers pretty flies out the window.

We need a different model that utilizes the same principles but is fitted more for a game that is inherently chaotic and diverse.

[elliott20]

after thinking about this some more, I think one of the many things we need to first do is to considered the difference between colors and races and what niche each component is supposed to fill. That is basically MtG tries to have the checks and balances built in.

Life: good buffs, good healing, good longevity, and of course, Torin. Life summons actually have one of the best BANG for the buck because it makes your normal units that much more powerful. Life generally works best with races that has a diverse unit selection. They are also great for infrastructure growth and have very powerful heroes to go along with it. With races like highmen, high elves, barbarians, and such races, this would easily be considered a top tier combination.

death: Debuffs, insta-deaths, and crippling effects. Death has a very very high reliance on the wizard's ability to overcome the oppositions resistance since it's about making the other side SUCK as much as humanly possible. Unfortunately, this means that you only got a handful of combination to maximize this trait. The archmage retort, and boosting your casting skills. Another problem is that a lot of the spells that rely on overcoming an opponent have a very small cap on how much mana you can pump into it, making this overcoming process a lot harder. so this limits the usefulness of a lot of the spells against stronger opponents. So once again, in order for death to compete against higher level foes, death needs to rely on buffing, summoning, and infrastructure building speed, something that death is TERRIBLE at doing.

nature: summons, field control, and creature healing. They also have good growth buffs. But for the most part, it's the first three I mentioned. They're not good at actually buffing their creatures, but they are pretty good at keeping your fighting force alive and replenishing them if they can survive the fight. My gripe with nature is that their summons are not all good enough, and it introduces a lot of summons that just don't do what they were intended to do.

I'll do sorc and chaos later, since I need to study up on those.

[elliott20]

Sorcery: Transportation, world field control, spell stopping and many many other utility spells. While not as solid as say, life magic, it is the one school that has all the game breaking spells in them. This is the one school that bears the most studying and scrutiny as a lot of the spells here can yield some very powerful combinations if used correctly. (i.e. invisibility + any flying unit, etc) This school also has two of the more powerful summons: storm drake and storm giant, with phantom warriors being a surprisingly good summon for it's price. (in fact the in combat summoning feature almost makes this almost one of the best low level summons out there)

Chaos: Creature buff, direct damage, and more combat spells than you can shake a stick at. Because of the fact that it's so effective at killing units, it's probably one of the match ups that white needs to look out for the most, since white is more unit dependent than chaos.

Those, from a brief glance, are the basic principles and niches of each school.

Balancing can mean several things. 1. we increase the capability of each school so that they are about on equal footing with each other in most aspects save one (boring!!!) or 2. we focus on increasing their niche field, while we build in more arcane spells that can provide a reliable means, if less efficient and less stable means of dealing with each tactic. Having some universal spells that can negate most of the problem issues will serve as our defensive buffer from accidentally adding into too many game breaking new spells and abilities.

Now, a quick scan of the ranks I feel the ranks for viability are as such

Sorcery (least predictable)
Life (works with a LARGE variety of retorts, races, etc)
Chaos (direct damage means you can, with the right combo, plow through armies)
Nature (summons are.... well... not THAT great... otherwise there are a couple battlefield control spells that can come in handy. Actually would work best with range troops like luck halfling slingers)
Death (least viable due to it relying heavily on overcoming opposition resistance, which most of the spells cannot do after a certain point of the game)

[Lucern]

Good analysis. I'd only add that I think Chaos has some amazing and world-changing (for the worse) global enchantments. Meteor Storm alone can make sure no stack without a healer or regeneration can venture outside of towns for long, killing engineers, ships (gradually), and annihilating swarms of AI troops. More rational players are still pinned down, and I should mention that the caster's troops aren't subject to the spell's effects. And then there's the effect of destroying buildings. If you have an empire of any size, I find that most turns Meteor Storm destroys more buildings than I finish. Armageddon also makes the world a less pleasant place for everyone but the caster, making my economy swing wildly depending on where volcanoes are shooting up. I had one game where Srrr'a got off a meteor storm and Armageddon right before Jafar whipped out Suppress Magic (arguably the best spell in the game). When my Disjunction spell failed due to Suppress Magic, I was none too happy, as the world was battered with meteors and rivers of lava. Add to that some nasty chaos city enchantments and I'll call Chaos's overland magic superior to most. Note that these end up giving chaos Wizards far fewer points in the single player due to all of the wreckage, but in multiplayer, I think they'll be more dangerous than many expect.

On the note of balancing magic, I started an excel file with an assessment of each spell in the game a few weeks ago, giving them scores 1-10 (and a separate score for potential in multiplayer games). I only got through Life and a few other spells I felt strongly about, but that's another angle as far as magic balancing goes. Each magic type is 40 discrete spells, a small enough grouping that you can wrap your head around and really see what the type is all about when you do this. This is best done after a playthrough with spellbooks in that area. I think Life ended up with a score in the low 300's...which means that I consider them to have an abundance of really useful (and overpowering) spells, only really falling short in summonings (which, sad to say, aren't hugely game-deciding) and a handful of lackluster unit buffs, in a sea of very game-deciding city enchantments, combat buffs, and, well, Torin. Where this exercise lacks is what Elliott gives here: the fact that the magic types are more than the sum of their parts.

When and if players decide to make a version that's better balanced for multiplayer, your #2 is very appealing. If you or anyone else comes up with any adjustments or new Arcane spells for the sake of balance, and those would benefit from some graphics, count me and my pixel-brush in :D

[elliott20]

forgot about that part of chaos. That actually does make Chaos a very good short multiplayer game contender.

If we were to be serious about balancing this game, one of the functions we would need is the ability to record games and run some numbers based on spells cast, units created, turns taken, the frequency and popularity of a particular combo, hero choices, etc, etc, etc. There are large number of matrices we would need to consider before we can REALLY observe the game at work.

the one analysis we can make right is to come up with strategic goals (i.e. what assets the player is aiming to gather) and then try to calculate what is the FASTEST possible method of getting there. Do this in comparison to other strategies and you can just about figure out which strat will yield the fastest return, and based on how each strat performs against each other, you can pretty much figure out the strength of the strat too. But all of this takes a lot of time and study.

super units for each side
life - torin
death - demon lord
chaos - doombat and chaos spawn (if paired with sorcery)
sorcery - race dependent
nature - great wyrm

[Tino Didriksen]

Chaos Channels alone allows you to strike hard at everyone. E.g., flying Paladins are beautiful.

[elliott20]

Tino's post highlights why I said it's probably more prudent to evaluate colors and races from the strategic options choice rather than inherent advantages and all that.

[Tino Didriksen]

A lot of those statistics could be automated. Some are even pure math.

Throw AI players against each other over and over, limiting what subset of spells and units they have available. Record results.

[elliott20]

it's really not that simple, Tino. Sure, you can run automated battles against armies over the course of a 1000 times. But those results, by virtue of being the result of ran by computers, are pretty much meaningless. It's like running the same chess moves against each other 40 times. You're gonna get the same result if you're just running simulations. The best chess computers in the world can only beat human beings by brute force calculation, and that's only with 6 different pieces on an 8x8 board. A game like MoM, when it comes to strategic effectiveness, is almost entirely dependent upon who uses, when he uses it and how he uses it.

That's the core of balancing. While you can create models for simulation and projections to help you learn and clarify numbers and such, pure balancing cannot be done mathematically. If it can, the game itself could be solved and that only means it's not worth playing.

[Tino Didriksen]

Hence the variation in the setups, who gets to shoot first, etc. You can get meaningful results if you control the parameters.

I am not suggesting a simple fire-and-forget purely random fight between AIs. While that would be the minimum amount of work, it would also be the most chaotic. Guiding the parameters takes a bit more time, but it's still orders of magnitudes faster than playing the games manually.

One should of course also record all manual games, but there is bound to be a limited amount of those.

[Lucern]

This raises a few questions. Do we have the capability to record the results of battles/whole games, via 3rd party program or otherwise? Do we have anything that can generate the kinds of stats you guys are talking about from the game? Can we even create games with all AI players? (I mean in the future, when they'll go after each other).

The experiences of veteran players playtesting might be the best we can feasibly do as far as analyzing balance. That wouldn't be terrible, I think. The 10 highest posters on this board have probably played 500 or more games between them. Naturally, when the game is changed, their previous experiences count for less.

[elliott20]

well fundamentally, the best we can do to short cut the analysis phase, other than recording whole games and observing veteran gamer actions, is to do a theoretical game on paper, using statistics like Tino said, and use that as our basis. This means we are basically cutting things down to the following:

1. strategic goal, like I said before, where you try to figure out what is your early, mid, late game fighting assets. That is, what things do you need before you can safely move against another player?

2. resources needed to achieve this strategy

3. how does it perform against other strategies?

Because the strategic goal itself is what drives this process (that is, the end result army / spells/ enchantment combo), knowing what these strats are is still going to be the first step. I'm willing to bet that as soon as we do that, a lot of big ticket strategies immediately disappear in multiplayer.

hell, I'm sure razing will become far more common just due to the fact that it's the fastest way to clear the area of city with races you don't want, fame be damned.

[Virm]

Well, as a temporary alternative to both existing propositions of running random samples with the AI, or recording the statistics from player-driven games (this category would include the partially guided AI simulations), we could actually do this as a hard numbers game. All aspects of the game are programmatically defined already, and the vast majority of these values are widely known or easily discovered. This leaves us with a finite number of variables to consider, most of which have predefined values, and the remainder of which are bounded probabilities. The only genuine unknowns are the intelligences behind the motions, but even this uncertainty may be mitigated if we calculate based upon using ideal encounters and situations. Once these figures have been compiled, a (long) series of trials may be performed with the results recorded to a database for statistical study. After perhaps 100 or 200 repetitions of each possible scenario, we would arrive at a comprehensive statistical chart of both the raw capability of each magic base/race/unit mix/etc. as well as a reliable estimate of the relative power of each combination in simple strategic use. This would make a reasonable means of comparison by mean power distribution, and could be used for rough adjustment of power balance. It would also make obvious any simple "win all" setups that could occur and thus make correction of such much simpler.

Again, this makes a perfect (and automated) means for "rough" balancing. Anything more would indeed require the logging of games played by human players and the results of any given setup and encounter. Any meaningful results from such logging would likely require many hundreds or thousands of games to be played and logged, otherwise results may be overly skewed by the happenstance of random numbers generated for a small number of occurrences. However the previous automated method would serve to reduce any extraordinary combinations that may be "game-breaking".