As perhaps a few of you know by now, my current project is a large scale turn based strategy game. One of the features I've been pondering [and wished to recieve feedback and criticism of] is the ability to customize the race of peoples which the player controls. Similar customizable races are fairly common in space 4x games. Fantasy 4x games seem to focus more on the "wizard" player customization. Commonly used is a point system, where the game has a set number of points that the player can spend on advantages for their people. Some games also have disadvantages which "give" points. These systems lead to balancing nightmares. Commonly a few advantages are "underpriced" and thus everyone seems to take them. Some advantages are not very powerful, or quite limited, so nobody takes them. These situations lead to a system that does not provide many choices to the player because many are useless choices. The main problem I faced though was not directly the problem of previous point systems themselves. Commonly, there would be a group of advantages which are complimentary. For example, having an advantage that causes the people to mine 50% more minerals, and an advantage that causes minerals to be sold at 50% than "normal. Alone, these advantages are less effective than when combined. This leads to one of the balancing problems. These two picks are now "underpriced" when combined. What I am aiming to do [and would like feedback/criticism on] is a dynamic point system for racial customization. In the above example, taking one of the advantages would cause the other [and indeed, any related advantages] to become more costly, because the advantage has become that much more of an advantage due to the first.

It's a good idea, but how would you ensure that you've taken all possible combinations into account?
Some of the combinations might not be as obvious
The problem is that it really is useless if the system only catches 99% of these underpriced combo's. If there's one left that's still underpriced, people will use that, and you might as well not bother restricting the other 99%

You're talking about diminishing returns which could be a good thing.

There is no reason why one skill should be more versatile than the other though. Take a long, hard look at your game design and see where those other skills could be a factor. Are you really exploring all the possibilites of that skill? Expand on your game and create situations where the player has few or no other choice but to have a strong rating with one particular skill.

The challenge, like you said, is balance. If there are too many combinations of one particular skillset, then what happens when another skill is needed but unavailable? Your players will have to realize this.

I find it easier to think in terms of economics (supply and demand), which ties in with your idea of diminishing returns. In the end though, it boils down to ensuring that you're making full use of all the available skills as creatively as possible. Sure, one guy could be an overpowered miner, but make lumberjacks/carpenters equally important in the game so players won't ignore them completely.

An excellent game that implements this system very well is Morrowind. Check it out if you haven't.

While I see the good in your customizable races, I think that the advantages you are referring can be attributed to technology. Perhaps a research ability is more what you are looking for. In some games, you get numbers tied to each branch of research (economics, mining, stealth, etc.), and each point gives a raw percentage bonus to your base output. Of course, the higher you get in your research, the more raw research points you would need for another technological advancement.

Instead of these technological boons at the start (which anyone should be able to match given 100-1000 years), perhap you should look more at the characteristics of your race. A stat method is in order, akin to the str/agi/dex formulas of rpgs but given a more "general" outlook and racially distinct traints (growth rate, habital environments, etc.). For example, your society may place a big emphasis on working hard, and thus produce an x% bonus to all work related activities.

Using a stat method (for example, using a sliding bar for each statistic with only a certain total "points"), you can make even more distinct races than just by using advantages alone. And, by making higher attributes more costly to increase, you encourage well-rounded races and discourage overspecialization.

Also keep in mind that 50% is a lot of a bonus. It might not be for something that is really small, but for something extremely vital to any game (resources), 50% is a _huge_ bonus.

I don't think you're ever going to be able to escape balancing via playtesting because you simply won't be able to cover every base AND have a game of sufficient complexity.

However, would it help at all to balance every bonus with a deficit? You would still face the problem of pricing everything, which can only be determined once detailed interactions and strategies are revealed through play. But it might cushion the effect somewhat because you'd always be bringing them back to a zero level. So if you get mining +x% you automatically have to take something else (fleet size, reaction adjustments, etc.) as a loss.

Alternately, you could strictly class picks and allow only so many picks from one class: So mining at x% is in the same class as selling minerals at x%. You would have to be diligent about possible cross-class benefits here, though.

Rather than trying to do this now I'd seriously consider doing your best few passes then washing this all out in the QA phase. Unless you develop a rigorous statistical model I just don't otherwise see how you're going to cover all bases.

Hrm, perhaps I was not clear...

1. Certainly it will still require much balancing, and can still fall victim to the common pitfalls of customization. Play testing will tell wether things are useful or not, under priced, over priced, etc.

I think that doing dynamic pricing will allow for another tool to use during that playtesting phase.

2. It is not so much diminishing returns as cumulative cost with cumulative benefit...

In current/past systems, something like this would occur:

Base ability: 4.
Advantage 1 (+2) - costs 2 "points"
Advantage 2 (50%) - costs 2 "points"

adjusted ability: 9 for 4 "points"

But advantage 1 gives 2 ability for 2 points, and advantage 2 gives 2 ability for 2 points. Put them together though, and you can get 5 ability for 4 points.

Now things are unbalanced, as the combination gives more than the points say it should.

Since the benefit is cumulative, the cost should be as well. Advantage 2 should cost 3 "points" if taken in conjunction with Advantage 1.

The point system should adjust for that so that the points more accurately reflect the benefit gained.

[of course, this is a fairly simple example. Things in the real game will unfortunately be a bit less clear cut]


3. No percentages. In fact, I can almost guarantee there won't be any "bonus to mining" style advantages. They either don't differentiate the races sufficiently to make customization worthwhile, or they are totally powergamed to the exclusion of other alternatives.

I actually don't want the race picks to be researchable either. I'd like to promote the mixing of races to get advantages/skills unavailable to you initially for quite a few reasons. [though that's a discussion for another thread...]


Argh, but it is late, and I cannot think too well, I shall check in the morning.

I think the problem you're describing can instead increase replayability in a game. I know you're trying to avoid drastically over- or under-powered choices, but the player probably won't know which choices those are until after playing the game. Even then it will take a few plays through to discover/invent some killer combinations. Then they can play at a higher difficultly level with their better knowledge of how the game works.

Also, if you did manage to remove over/under-powered choices, you could end up with the opposite problem where the decision makes no discernible difference. If you had trading bonuses, such as +5% selling price to one of minerals, oil, metal, food and so on, does it really matter which you pick? Only if some won't be traded and others will, i.e. food is too important to ever sell. However if you can build mines for minerals and oil platforms for oil, then picking one bonus over the other doesn't change gameplay much.

One possibility I thought up to avoid "the obvious choice", but in researched items, was to slightly change the stats of several items each new game. Not randomly but designer-specified, such as +10% damage or -15% fire rate for a laser cannon. Each game would pick a small number of items and activate their +ve or -ve adjustment, so the tech tree would still have some surprises to increase replayabilty, while remaining recognisable from the previous game. I'm not sure how this would translate to race abilities because they are visible at the start of the game, but obviously the value of mineral bonuses depend on how abundant minerals are in the universe and their distribution.

Fulby

Ah, to address the replayability issue:

Something fairly major I forgot to add, the game is currently designed as multiplayer only. Humans unfortunately are not very sparing when it comes to taking advantage of other players making sub-optimal choices...

But having random maps and human enemies has always been enough replayability for me. Is it enough for others?

In that case people are bound to find certain combinations which are the best choices. If possible, game reporting would be useful so you could collect statistics on most of the game's played, and see if particular race setups tend to win a lot more than others.

As an aside, one of the things that I thought kept Counterstrike fresh during the early versions was that the balance changed fairly often, letting the player try out new tactics and choices to find the new optimal play method. It does cause friction though when people loose their favourite tactic that let them win so much.

Fulby

I played Stars!2 and it had a wizard to select the customized race stats. Unfortunately there were a couple of unbalanced items on the list.

One grievious error was a race main stat called claims adjuster. This trait allowed the race to automatically teraform any world they landed on. ON the surface this may not seem like a large advantage when complared with the other possibilities. However, when this stat was mixed with Total terraforming, and 75% research cost for biology, the advantage was huge. The player could esentially start with a small range of planets they would be able to colonize, As they progressed up the tech tree, they could colonize more and more, essentially becoming a hyper expansion race with none of the disavantages associated.

I think that dynamic costs would be one way to solve this. However that would require you to analyze a lot of game play to find out just how much an advantage like that was worth. I believe there is a much simpler and much more elegant solution:

Free Market Race picks
Use a free market sytem to determine the price per pick. That way the players will automatically generate how much each pick is worth. There could be some abuse issues such as repeatedly picking a trait and then dropping the game to raise the price, but overall i think it is a fair solution. This would require a server to determine pickrates before each customization but overall it should be fair.

One method to ensure a fair battle would be to force people to recustomize their race based on current rates at the beginning of every game.(freeze the rates when the game begins to ensure all are on an even field)

I hope this gives you something to consider.