Item & stat system

[Marek]

How do you make a rpg stat system interesting?

The simplest approach is a single stat: call it something like "power". Items give you +1, +2, +3 power, and so forth, and the more power you have higher chance of defeating an enemy.

This is supposing that a battle is a simple random roll, with a percentage chance of success: either you win or you lose, and your items and stats help your chances to win.

But with a single "power" stat, items are one-dimensional: the gameplay decision is trivial for players. Always pick the items with the highest power -- there's no interesting strategic decision there.

What do you add to make items & stats involve interesting gameplay decisions?

[Ecce]

The obvious next step is to add additional stats which make items multi-dimensional. If Power scales damage, have "To Hit" scale accuracy, and "Dexterity" scale the chance of a critical hit. Now you have a three dimensional item. The more stats you add the more dimensions you have, but you have to balance that with not making the item system too difficult to understand or for the player to strategize. Too many stats can also make actual game balance hard from a content creation perspective as it becomes difficult to make sure that a point in one stat isn't significantly better than in another.

[Nox]

However gameplay > balace (http://www.eldergame.com/2010/11/how-to-balance-an-mmo-and-how-to-stop/)

Take a look at Diablo 2, that's a game that has a real shiny item system
1) enough of stats (and supporting character system) to allow a good diversification while not being a mess. Coupled with good GUI it is quite easy to get the info (not necessarily easy to decide though!). Stats often share some philosophy, like pikes/poles have low min dmg and high max dmg, swords and axes are more balanced but with shorter range etc.
2) special properties -> it is very common for an item to have various special properties rolled, these are either passive bonuses to character or the item; or adds some effect to the item/character's actions; or allows a player to use special skill (item has X charges and then has to replenish by reparing)
3) special items -> not all items are the same:
 a) worse/better quality commons, magic, rare, set items, uniques, crafted items, socketable items;
     normal x exceptional x elite versions;
     class-specific items
 b) making some item special by allowing a unique special properties on them, not findable anywhere else (circlets)
4) items modifications -> socketing gems (effect by gem type and item type), jewels (randomly rolled special properties), runes, rune words!

The special properties are often tied to the item type, so boots can get walk/run speed bonus while shield rather gets damage reduction... but it's not 100% set that it's not possible to have speed bonus on armor.

Unique and set items are named and have their properties tailored by the name and type and stuff, often even only by reading those specs you can imagine a little story behind.

Interesting item names, even for basic items

[Marek]

What if the multiple dimensions of items (such as damage, accuracy, speed, etc) end up just being a simple calculation away from a one-dimensional scale of power?

For example, players can easily calculate the damage-per-second of an item based on damage per hit, accuracy percentage of hits, and speed in hits per second. So it becomes one-dimensional item again.

In PBBGs which lack real-time combat decisions to provide a tactical aspect, how do you give the player an interesting mechanic?

[Nox]

That's a great question

Some thoughts:
Special modificators aren't one-dimensionable, for example when they will affect the character itself of his other item(s).

You can also apply effects past the combat - in NR, you can choose a loud weapon, it can have a great power but then you'll be luring so much presence to yourself
Or (and these are more commonly used)
- a weapon with only a few rounds in a magazine
- or one with rare ammo
- or one easily breaking down
- or one with short or long range

A weapon can have certain melee defense measures - simply adding resitance or allowing block or perhaps special blocking abilities

http://www.sirlin.net/articles/rock-paper-scissors-in-strategy-games.html (but read the rest too )
Some stats can be dependant on a not-so-predictable current situation of the player - let's say damage stays but hit chance drops with tiredness/low hp if you have a heavy weapon ("this weapon's generally better, but will i be able to keep the health up?")
It shouldn't be "situation" generally, because in PBBG player often does not really have a chance to know someone will attack him and adjust etc.

[Marek]

Thanks for the ideas Nox!

Your mention of power vs risk got me thinking: the element of risk and unpredictability could be simplified to the idea of variance.

Two weapons can have the same mean damage ("power") but could differ in variance. The higher the variance, the more its damage rolls are spread out, while keeping the same mean. So a high variance weapon would produce more critical hits, at the price of more critical misses.

Now, here's my favorite part of this idea: the mean and the variance happen to be the two parameters of a normal distribution. I like this because it means damage can be modeled very elegantly.

The question remains: is higher variance a good thing or a bad thing? Statistically, it makes no difference in a large data sample. But I wonder if there could be an interesting level of gameplay decisions hiding underneath...

[Nox]

I heard - but this was in case of realtime games (perhaps DotA and/or the D2) - that for PvE you'd rather like steady predictable reliable damage while for PvP you'd rather like more variance.

It kinda makes sense - usually the AI will not really adjust or be affect by what's happening (and predictability will make the player be able to make decisions with lower risk) - while for players this creates unpredicability+surprise (which creates emotions)* - they dunno whether the next attack will be again 5% of their HP or will it jump to 15%? Should they risk staying at low HP?
It also has a "shock value" - when the opponent suddenly receives a huge amount of damage, it can throw him off.
So it has these benefits while the mean remains the same.

Will probably be dependant on the player's mentality, although I guess people are a bit more influenced by the max value then a real mean.

I can't say this 2-parameter differenciation will be enough to create a system you (and your players) would like, but surely it can be tried

*) therefore balance-wise I would make the considered mean higher for weapons with greater variance ... or something like that and then balance by PvE (real mean) x PvP ratio in the game, so the shock weapons will be a bit toned down... although 'm not 100% sure about this

[Marek]

Well, the issue with variance is that it's symmetric, so there's both the shock value of a critical hit, and the shock of a very poor hit.

The issue here is that if you're fighting an enemy which is considerably weaker than you, then critical hits aren't as important to you (because you can defeat the weaker enemy without them) but critical misses, if they are critical enough, could make you lose. So it's disadvantageous to bring a high-variance weapon to a fight with a weak enemy, because you're just opening yourself up to a chance of defeat, albeit rare.

Basically the higher the variance, the more you open a window for something unlikely to happen. With lower variance, that window is entirely closed. So to defeat a monster much stronger with you, consider two scenarios: you make 100 attempts with low variance, and fail every time, or you make 100 attempts with a higher variance, and on 1 attempt  you fail really, really bad (but it doesn't matter because all failures are alike) and on 1 attempt you get a crazy critical that makes you win.

Why at all constrain things to only a simple normal distribution with two parameters? Because RPG combat (real-time or turn-based) doesn't translate well to plain PBBGs, so instead of trying to force it in, I want to explore the idea of just stripping it down to the underlying model. If you ask me, there's no point in making 10 random rolls to determine an outcome if you can just derive the random distribution underneath, and do it in one roll.