Spring RTS Engine

I was reading over a couple different pieces on game design written by other forum members here recently, I tended to come across this concept alot, balance should be based on "for cost" evaluation.

Thinking about this idea a fair bit, I'm pretty convinced that for cost evaluation is at best an occasionally useful tool, at worst it's a horrendous trap that causes you to overlook more powerful cues in favor of an arbitrary and abstract mathematical calculation.

"For cost" is deceptive, and here is why: If I have a fast melee unit, and a slow ranged unit, and in a 1v1 scenario the melee unit kills the ranged unit, that will not hold true necessarily for 2v2, 3v3 16v16, whatever. Even the slightest variation in range and movement speed will result in a widely variable cost vs cost result. There is no one cost vs cost test that will produce an accurate depiction of every cost vs cost scenario. A krogoth will kill infinity peewees in 1v1 battles, and die to maby 30 if perfectly surrounded and microed against, and that's not just an issue of micro or ideal battle placement, it's an issue of unit role. more on that later though...

Additional to "for cost" being deceptive, it's unrealistic. It's a bit simplified in TA with the 1 resource (sorta) system, but in other RTS games like starcraft2 we see this phenomena more clearly. How much minerals is gas worth? Well, in the early game gas isn't worth very much, it allows you to get out a few key tech structures and units, but focusing too much attention on gas will decrease the immediate DPS and hitpoint values of your army. There are many strong rush builds that rely on getting late gas because if you have an early game mineral heavy army your army is basically just bigger, stronger, tougher than whatever the opponent will have in the early game. In starcraft, gas gets you upgrades and lategame tech, so in the late game more and more it becomes extremely valuable. So... back to the original question, how many minerals is a single unit of gas worth? The answer is that it fully depends on what stage of the game you are in. For cost is unrealistic because the actual cost of a unit shifts throughout the game in a normal RTS, even if the # in resource cost remains the same. There are many more resources than just the in game mineral count, and those resources become more and less valuable as the game goes on.

Adding onto that, for cost is unrealistic because it doesn't account for hidden costs. If I am examining a T3 unit, in TA, there are many many many hidden costs we do not see. How much did it cost to get the factory running in the first place? How much did it cost to survive up to that point? How much cost in variation from the build I was attempting was there? In certain TA variations that used to be played, you could not switch easily between vehical and kbot tech from one teir to the next, it was an early game choice; so certain unit combinations had immense in game cost because they required 2 extremely expensive factories and tech paths, however, their actual unit costs were minimal. In a 1v1 game those combinations would be ridiculously unrealistic, but in a 2v2 game the balance may have been totally different. Hidden cost is virtually impossible to algorithmitize, at least not in one sitting. The hidden cost for one unit may not even be the same as the hidden cost for another unit in the same tech tree built by the same structure: in starcraft 2, if I am relying on a sentry heavy army I have spent alot of gas, and thus the hidden cost of trying to buy colossus is huge, because they are costing me large amounts of my most valued resource; it may be more realistic to build immortals, they cost less gas and let my army be stronger in the unit it is relying on. These structures certainly exist and are valid in all other RTS games as well. Hidden cost can even be built into a unit dependant on what the opponent does. Cost vs Cost is EXTREMELY inadequate at evaluating real game scenarios.

In truth, for cost testing is woefully incomplete as a balancing tool, virtually every for cost test would not produce liniar graphs if unit counts are scaled, small numbers of zerglings will defeat small numbers of marines, but die horrendously to large numbers. This doesn't even account for the myriad of game situations where scenario plays more into effect than raw cost vs cost unit count, marines are better with a wide arch, so they are reliant on terrain, and any unit that controls terrain (infestors, tanks, colossus) will be powerful against them, but this is virtually impossible to accurately test in cost vs cost simulation, let alone more complicated (and therefore more realistic) situations. Many units have high worth without acctually even doing any damage. Flash tanks in OTA were fast and could rip apart your economy, they gave their owner map control, so he could expand freely and you would have to constantly defend, they were more valuable than their outright cost. In starcraft cloak does this for many units. Dark Templars are useless if they are seen, but the simple fact that they have to be seen means that a player cannot move out against them until his tech tree has reached a point where he is comfortable in the amount of detection he has.

Basically, I'm arguing here that cost vs cost testing has very little authority into the balance of units against each other and relative to the game. I think it should be implicately understood by game designers of all spring projects that cost for cost testing is fine if you take it for what it is, but what it is not is a good gauge of the way things should be viewed. I've heard a fair few people purporting this tool in various places, and I'm not going so far as to say that it doesn't have it's uses, but I want to strongly encourage spring game designers not to overvalue cost vs cost test results. Most units should not be very close to each other in most cost vs cost tests, balance comes from the realisticness of getting a certain sized army of a certain type of unit in a real game scenario, the hidden cost of moving that army around the map (slow units cost more to be away from the base than fast units and soforth) and the many many hidden costs associated by terrain and positional advantages and disadvantages. There are just so many hidden costs and considerations that in one cost vs cost test you can't ever really claim to have learned anything substantial. Ultimately, over and over again, unit role breaks for cost testing and makes it useless.

So where to go from here? I'm not really sure, get into games and play the hell out of them I guess? Find replays from other players and go over them with a fine tooth comb to look for things you weren't expecting relative to your design? I think unit statistics are somewhat telling, who does damage to what in real games played? I think ultimately balance is a bit subjective to begin with, I kinda feel like it's best to tweak away from everything being useful and instead aim to tweak things such that the current state of things is fun. It's probably not fun to face the same strategy over and over again in a game, but it may be less fun if there are 4 alternates that are barely different. If one unit is crushing its counters due to real game situations, buff counters maby? nerf unit? depending on what other interactions you expect to encounter. Balance is not an easy task, but I'm convinced that cost vs cost testing isn't even close to the ultimate answer.

Actually sounds quite familier. :D

I also quite agree that it is a wild goose chase to go after a 'perfect balance'. Even chess has not been 'solved' yet, and games nowadays are much much more complicated, so there is no chance in ever achieving complete perfect balance, at least with the resources available to us currently.

Balance itself also tends to be rather overrated. Ultimately I see at as just a means to an end, that end being having a fun game. After all, in the end, its a video game and most of us play to relax, even if we like to relax competitively.

I concur swift. Pretty much why I arbitrarily assign value to roles or abilities and then just adjust as I see them exploited or neglected. Fang once said that balance is an illusion and that he often would say he made changes that he didn't really make just to see people react making a unit stronger or weaker due to their confidence in a unit.

I think that watching games over and over and over is a huge part of balance. Even so like you say, team dynamics make things very strange. I have seen it in gundam, I see it in team based games and things like LoL.

your post is a good read but neglects discussing meta game. "oh shit they have X, we lost" followed by a player sandbaging his team. It also doesn't really discuss when players take on roles in a team. A really sick team of gundam players will have coordinated airstrikes, raids and mass troop(t61, zaku1, magella) charges. When you get say a strong artillery player who uses a spam player as a shield and has an ally providing air cover for the artillery it becomes impossible to stop.

It's a good point smoth. Certainly team games drastically change the dynamic of a game, especially at a high level of play. I've been watching alot of starcraft 2 pro events recently though so I'm thinking a little more in context of 1v1 games, but definitely within the context of a team multiple players can have different roles that can significantly modify the way the game plays out and therefore how it's balanced. OTA with the strong build paths is a good example, but even just any game, twice as much player attention and micromanagement time on the team can have significant results in how different units feel in balance and how balance works.

Balance is a distraction from more important goals. The only time it even matters is when armies field different units. Chess is a poor example of bad balance since the only distinction between players is white moves first.

When you think about about whether something is balanced what you really mean is the game would suck if every player used that unit or faction to the exclusion of more interesting units. It's all about the entertainment factor. I see no harm in having an imbalanced (cost wise) unit provided it is a fun unit and other viable and entertaining strategies exist.

I honestly feel the reason balance is discussed so much is because it is the easiest of game elements to change and debate. It's certainly easier than building new models or fixing bugs. You'll often see mutators in modding communities that only exist to change balance through the tweaking of stats. While some may improve the game for everyone the majority simply favour a style of gameplay and thereby reinforce that balance is actually an opinion, not a tangible objective.

Eventually balance is related to the workings of the engine itself.

The effectivness of a unit is embedded in it's stats
(including any engine quirks that are revealed by a certain combination of a unit's proporties); projctile stats, hitsphere and any other unit characteristic.
When we say unit effectivness IMO it basicly means the amount of time a unit is used per game in an ideal state in which every unit is used to its full potential.
And this is on the entire plathora of maps available.

Ultimately the amount of use each unit gets within a certain set of units depends(again if they are all balanced) also on the total amount of units in the set.

We dont know to what effeciency players will be able to use their units and increase their economy, this might be possible in theory to calculate but is probably very complex(average speed of comands mouse clicks per minute mouse movement etc).
If a unit is very expensive compared to available resources or very cheap also matters and decides what proporties of the unit will dominate in game.

Finnaly there are engine bugs and unforseen behaviours that were not intended when the engine was created or rather, not expected and a truly in depth analysis of gameplay and balance would actually examine the engine workings as well from some outside perspective , and try and deliver a complete report on possible unit behaviours based on how the engine is built(meaning you'd have to fully inspect and anylyze the engine and connect it all back to the final creation of the desired balance and gameplay).


The practical approach used by all game devs is something in the middle where they attempt to contorl the engine and unit stats and prethink the gameplay and balance while also understanding that this route is limited and they will have to change things as they move forward based on their analysis of new gameplay evidence and user feedback.

To fully prefigure the gameplay based on code (i.e. making balance and gameplay changes without playing at all and only looking at the code) one would probably have to create a program that would analyse the engine and make the connection between the code, human abilities and limits, available maps and gameplay and economy rythm.

When you will make code changes/change behaviours of units or simulation the program would immidietly show you how this will effect unit interactions and balance...

added

This is not touching fun...
For a program SC2 gameplay and TA gameplay is not more or less fun...
I guess in theory its possible to draw a few trees and algorythms of different gameplay styles that would be considred fun by players but this is probalby an even harder thing to do and would also probably be pretty general since you can practicly do anything in games these days.

Gota wrote:To fully prefigure the gameplay based on code (i.e. making balance and gameplay changes without playing at all and only looking at the code) one would probably have to create a program that would analyse the engine and make the connection between the code, human abilities and limits, available maps and gameplay and economy rythm.

If all maps were flat, symmetrical and all player computers equal MAYBE you can get close to to argue any calculation based balance is folly.

I agree with most of what you've said but for cost does have uses. As with any balance tool you have to know the limitations of any method and make sure it doesn't get in the way of something fun.

These analytical methods are pretty good at turning the idea of a unit with a set of interactions into a real unit. You must have a good idea of what you want a unit to do and use things like 'for cost' to make sure the unit is doing close to what you think it should do. If creating a new unit it is good to do some sanity checks with things like DPS/cost and HP/cost for the role of the unit you wish to craft. Bur you cannot use these methods to design a game for you, only to provide a starting point for the fine detail which can be later refined with playtesting.

Recently I did a bit of rebalancing of ZK AA and gunships with which I used for cost a lot. The scale of combat issue is pretty easy to workaround by testing different sized battles, it is not much of a problem. I have to admit the gunship example is one of the best for this method of balance and I would be much more dubios of the results in a ground combat test. Gunships in Spring don't often gain much from manual control, the AA turrets are static and terrain is almost irrelevant. But for cost only really works when you have a good idea of the exact relationship between 2 units and of course the numbers were only rough and required testing in real games.

Many unit types and roles are very hard to analyse. Things like scouting, map control, fast response, artillery, special abilities ect.. are very hard to quantify so must be tested.

Although it is possible to do some analysis on these units I do have an example where a basic stats check would have been useful. Kingraptor added an artillery unit that had less range. far less DPS/cost (and no significant burstfire) and far less HP/cost than another artillery unit with very similar projectile physics and manuverability. In this case it was easy to see that the added unit should not have been added which could have been picked up with a sainity check on a few basic numbers.

Although now you've made me want to create an automated testing system because I can picture how nice the 3d graphs would look. I like 3d graphs. 1 axis would have the cost of unit A, the other axis has the cost of unit B and the 3rd axis has which proportion of the origional cost is left over for either A or B. The graphs would be completely impractical for anything useful but they would visualise things like the skirmisher effect and swarming really really well. Unit types may be distingiushable based on the characteristic shape.

I have a unit, it is artillery, it has high dps low aoe and fires in an arc. The problem is the shots do not hit 50% of the time.

balance it using your method.

player builds 20 of them. They now can absolutely dominate because while they may miss, others will hit, and the shots that hit nearby cover when a player tries to dodge. so with each additional artillery you get increased aoe. Furthermore they are guaranteed to hit because even if the player dodges, the misses hit because they land where the player was going. The inaccuracy now a benefit. This is what swift is trying to describe.

You cannot cost balance based on these types of synergy type bonuses unless you make them cost more for each successive piece of artillery.

I think you can...just need more math and complex equations.
20 artillery is not alone in the game...They are not built in a vacuum.
They are built within a certain time frame when you play a 1v1 on a certain map at a certain skill level.
It's the same with many other units...in TA if you have 1 flash or 20 their aoe growes as well.

Gota wrote:I think you can...just need more math and complex equations.
20 artillery is not alone in the game...They are not built in a vacuum.
They are built within a certain time frame when you play a 1v1 on a certain map at a certain skill level.
It's the same with many other units...in TA if you have 1 flash or 20 their aoe growes as well.

Even if you use ingame cost as the evaluation criterion, it's a long range unit, so the graph won't be linear, and with the arch shot it will break for anything that moves faster than the shot lob accounts for. Artillery are an especially complex unit role, they should be very powerful in for cost testing because they tend to just out right miss if the opponent rushes them. They are hard countered generally simply by the nature of their behavior.

I found the whole value-ation very distracting from the core issue, and even the pros are not imune to that. I remember sidmeier (civ-oh, wait, if you dont know that, forgot about this) saying: "To test a units balance, double the value in question, play a testgame, then half the value in question, after that, you should know were to go." So even he sometimes is caught by the rpg trap, that values, are what makes the game go round. Which is wrong.

What makes games good, is to force players, with every unit, every situation to do interesting decisions, that he can win. Example?
Its wrong to say, i want a unit that is strong for close combat only, because there is no dilema, no drama involved, no interesting decision, that holds something new every time you play the game.

Now, if i have a unit that instantanious kills One Oppossing Unit, but while doing so, looses the hitpoints it has taken from its Oponent, this create very interesting decisions. "Is that a kamikaze mission against a bigger Opponent? Can i retreat with my weakened Units after they hit to regenerate? How do i micro them from shooting twice, and that way disolving themselves?"

Loads of dilemas, drama and challenging stuff ensues.

Tl,dr; Chronotrooper from Red Alert Good designdesicion (values or not) Five Variations of the same Unit bad design desicision.

PicassoCT wrote: "To test a units balance, double the value in question, play a testgame, then half the value in question, after that, you should know were to go."

binary search

smoth wrote:I have a unit, it is artillery, it has high dps low aoe and fires in an arc. The problem is the shots do not hit 50% of the time.

balance it using your method.

I'm not saying I can. As I said some analytical methods can be useful in some situations and you have to know when they are useless. The applier of these methods needs to know the game quite well because no calculations will make good balance for you. It is just sometimes possible to test specific things in closed testing.

AH! I misunderstood

TA has (sorta) 1 resource system in what way? You have to balance metal, energy, buildpower, it's much more complex and harder to manage perfectly than SC2 economy.

Johannes wrote:TA has (sorta) 1 resource system in what way? You have to balance metal, energy, buildpower, it's much more complex and harder to manage perfectly than SC2 economy.

I agree and its also harder to grasp.

For KP I balanced spam units on a 1:1 scale but I didn't want them to be perfectly balanced all the time:

1 unit:
1 bug = 1 bit
* > packet > exploit

30 unit:
30 bug < 30 bit
30 exploit < 30 bit
15 bug + 15 exploit > 30 bit
* > packet

Why is that last one balanced? Because Network has superior force projection abilities and can strike a location with its full army easily while Hacker and System often have dozens of units in transit to the frontlines, in a situation with multiple frontlines they have to split their force between the frontlines while the Network can move the entire force to one or the other near instantly.

I think KP is especially interesting because build time is really the only resource and there are so few units to interact with each other.