A RPS structure, as already noted, leads to unwholesome gambling in army creation (as opposed to thinking strategically of a balanced roster) in the common case of incomplete information.
We can reduce the problem by making this gambling not convenient or by reducing the amount of incomplete information.

Saving and reloading games, or playing with random strangers on the net, might be an incentive for unnecessary risks, but in real life Warhammer matches and tournaments armies tend to be rather balanced because
1) A balanced army is likely to have moderate advantages or disadvantages against other balanced opponents, ensuring a fair and interesting match, while a match between overspecialized armies is likely to be won or lost too easily and without a true test of skill.
2) A balanced army is likely to be flexible enough to have an edge against most overspecialized opponents and, in the hands of a good player, beat them almost as reliably as their respective dominant strategies.

Regarding incomplete information, players can buy units in small increments after seeing most or all of the opponent's army so far. This ensures that anything specialized finds some counter in the opponent's army.
For example, if you buy tanks I can buy snipers with rockets and fast units that aren't easy to hit with the tanks, then you can buy cheap infantry to delay and harass my special units, then I can buy large area of effect weapons to kill infantry effectively, etc.
This kind of informed army building could be accomplished in alternating turns, starting with a player that makes an arbitrary purchase then buying enough units to exceed the already committed points of the opponent, or simultaneously, spending some minimum number of points each round. In both cases, the granularity of purchases should be small: not a "squad", which could represent a large portion of one's army, but individual men and maybe even extra weapons and equipment for previous units.

Quote:

The Uber Sniper unit I described does not need a specific counter unit at all. It's basically strong vs. high tier (high cost, low number) units, and weak vs. low tier (low cost, high numbers) units. It's also strong vs. ranged and weak vs. melee. Note that this is not really RPS: it beats and is beaten by entire general categories of units, without needing to know much about what those other units do to each other. All we need to know is that each faction has something in each of those categories, and we know the sniper is useful, without being invincible.

This is exactly why RPS is not the right terminology for this example, or the game. What's stopping a tank or a howitzer unit from firing a shell in the direction of the sniper. Whats stopping a whole squad of units armed with machine guns from emptying their clips in the general direction of the sniper. What's stopping your fast attack squad from charging his squad and getting into melee range. Since the sniper is geared with the minimal of armor, it should die very quickly. On the other hand, this sniper can get off a few shots before dieing, possibly at a tank or commander unit. So, the sniper unit doesn't necessary have ONE rival, it can have many.

What this starts to do is force the players to use strategy instead of raw power. It adds to the complexity of the game by having circumstantial advantages and disadvantages to a role. Snipers, when out in the open, are easy targets, therefore the best thing to do is to escort them to some cover. Have them in a mechanized transport vehicle, for example. Then use that vehicle as cover for your melee heavy units so that they can get into range. That is a strategy that can be effective when executed correctly, or it can be toppled if the opponent is smart enough.

Quote:

I don't think they're exclusive, I think RPS can be a useful design pattern within a role-based system. It's useful to be familiar with it, but I don't believe it should lead your unit design.

I agree, RPS cannot be avoided in this case because, due to the point system, all units will have vulnerabilities that can be exploited by at least a handful of other types of units. Therefore, it will come down to strategy to decide a victory. A player will try to hide vulnerabilities while trying to exploit the vulnerabilities of his opponent. But this isn't RPS and like I said earlier, its RPS on steroids.

Quote:

Note that this is not really RPS: it beats and is beaten by entire general categories of units, without needing to know much about what those other units do to each other.

You are essentially saying that the RPS system is therefore not applied to individual unit types, but top groups of units.

RPS is an abstract analysis. So it does not matter to it whether the choice is selecting a specific unit type, selecting an upgrade, using a unit in a specific way (placing archers in a swamp for example), or whatever. So yes, RPS does not even attempt to analyse a unit as to how they defeat an enemy, only that one choice (or even sequences of choices) beats another (or multiple) and is itself beaten by another (or again multiple) choice.

the reason you are dislike RPS is that you are using it wrong. I bet I would dislike screwdrivers too if I tried to use them like hammer.

Quote:

With the uber sniper mentioned earlier, you could try and concoct some kind of giant, sprawling diagram that represented all 5 factions' eleven units and how they interact to try and understand where everything fits in. It would probably take months to create, and any small changes in the unit statistics could quite possibly change it significantly.

The thing is, whether or not you create that diagram, the relationship between the units still conforms to that. If you don't have need for the analysis, then don't do the analysis (that is a no brainer). If however, you have need (balance issues) then it is a good idea to use the analysis tools you ahve to solve your problem.

The other thing is that it is easy to expand a RPS system diagram than it is to create it at full complexity from scratch. So if you start off with a 3 way RPS system, then you can add options onto it and expand it without much effort and still retain the balance that it brings.

As RPS system can easily be represented in a spread sheet, then making changes to unit stats can easily be handled automatically and this therefore eliminates these problems.

Doing it by hand would be hard, but then we are creating computer games and therefore I assume that one would have a computer handy :D .

Quote:

What about the Spotter, or the Tank? Neither of these roles exist to actually kill things by themselves; in the case of the former, it's job is simply to give Line of Sight to some powerful ranged unit (like perhaps, the Sniper) sitting miles away across the map. In the case of the Tank, it's job is simply to keep the enemy occupied while flimsier, but more damaging units can get on with the business of killing stuff undistracted. They're both support units that can make a big difference to the effectiveness of your army, yet neither of them specifically need to 'beat' anything at all.

Well the Tank unit would likely be good against the Spotter unit as the spotter unit as you stated is weak. So the Tank beats the spotter.

Now if the Sniper can kill the spotter, then the Spotter is in trouble as it is beaten by the Sniper. So this means that Spotters, although useful are not necessarily a good choice. However, as RPS is about choices (and series of choices) then a Sniper and a Spotter is the Strategic choice that is represented by a single node on an RPS diagram.

It would be better to model the Spotter, not as a Unit, but as an Upgrade for the Sniper (although one that can be removed).

Also RPS is about interactions between units. If a unit can only be acted on, then there is no interaction. A Spotter by its self can act on a Sniper (the sniper range is too great) so at best the Sniper beats the Spotter, but as the Spotter can act on the Sniper, it is not really part of the RPS diagram.

This is what I was talking about why you are using the RPS analysis incorrectly. You are trying to use it to analyse all choices, even ones that don't have interactions, but as RPS is a model of interactions, if you try to use it to model non interactions, then it will not work.

It would be a bit like trying to use the Role analysis method to analyse something that does not ahve a role. It wouldn't work.

For example: In Starcraft each unit displays a little animation depending on its unit type. Why unit role dose this animation window have compared to other units?

The whole question is pointless as the window does not have a unit role. Just as if there is something that does not interact, then trying to model it with interactions is equally pointless.

Don't use a Screwdriver like a Hammer.

Quote:

Except it doesn't tell us much about balance either, unless our RPS system is boringly symmetrical.

Nope.

RPS systems don't have to be symmetrical. I gave one example earlier so let me repeat it again for you:

Quote:

A beats B, C and D B beats C and E C beats D and E D beats B and E E beats A

This is not symmetrical and yet it is still an RPS system.

That's that criticism disproved.

Quote:

As soon as you get into more complex mechanics and start fiddling with DPS figures, cooldown rates and unit costs, build times etc. RPS becomes completely useless for balance, and you really need to start playtesting.

With RPS systems, you don't only just have the hard counters. You can also factor in costs as well.

Role analysis does not allow you to factor in cost and strengths for balancing. Therefore RPS is superior to Role analysis here.

Instead of just A beat B beat C, you can also use a chart like this:

   A | B | CA| 0 | 1 |-1B|-1 | 0 | 1 C| 1 |-1 | 0

As you can see this chart describes the RPS system but it has hard counters.

Now what if you change the chart like this:

   A | B | CA| 0 | 6 |-3B|-6 | 0 | 6 C| 3 |-6 | 0

This is still an RPS, but now we can see that it is no longer as simple as before. Using a bit of maths, we can work out that the relative unit strengths would result in both A and C being used twice as often as B.

This can easily be calculated in a spread sheet and this makes using this kind of analysis very convenient. It in fact explicitly tells us what the balance will be.

That is another concern disputed.

Quote:

It's basically strong vs. high tier (high cost, low number) units, and weak vs. low tier (low cost, high numbers) units. It's also strong vs. ranged and weak vs. melee.

Actually it look remarkable like RPS. From this I would say it is part of either a symmetrical 5 way system, or part of a larger asymmetrical system.

So

Sniper beats High cost/low number units and ranged units but is beaten by low cost high numbers or melee.

or

A beats C
B beats C
C beats D and E

It is a section out of an RPS. And so this doe snot actually make a good argument against RPS analysis.

The High Cost Low Number and the Low Cost High Number could even be removed from the specify unit analysis as these are specifying larger strategies. Either Hero (High cost low number) or Swarm (low cost high number) strategies as you could always be using Low Cost units in low numbers (Spotters?).