User:Shard/GWBalance

I got into Guild Wars via an ad for a gorgeous looking MMO with no monthly fee, something I hadn't experienced before. I quickly discovered the fun and challenging aspects of competitive play in the game, and even more quickly became familiar with the aspects of the game that most drastically affect balance. Balance is a tricky term. Many people use it in different ways. I define balance as the gap between a the game being too easy and too hard. If players have fun and are challenged at the same time, that game is balanced.

I jumped into pvp with a rich competitive gaming background. No world championship wins, but still, I've always loved showing off my 1337 skillz to anyone and everyone around me. Balance has never been a conscious issue for me - all the games I've played the most until now have all been almost perfectly balanced. Street Fighter, Starcraft, Magic, and DnD to name a few. Guild Wars was the first competitive experience I found which was crippled, but still managed to be both fun and challenging.

This is not for reference. People interested in balance might find this interesting, enlightening, or useless, depending on their personal ideas. If you're expecting a rant with some izzy or anet hatred, you will be disappointed.

This article looks at Guild Wars in its paper form. Each component of the game, be it a skill, an attribute, or a weapon mod, is looked at by itself, with no other factors playing in. The only time this theory uses actual gameplay is in determining the chance you will fight a certain build or skill. In actual gameplay, you will see things very differently than this page sees them, as things like timing, player skill, external map mechanics, and other things come into play.

I owe a lot to Charles Ensign. His seemingly limitless understanding of competitive gameplay still astounds me, and he is probably the biggest influence for me writing many of the things I've created both here and on the fan forums. Reading his guru posts always led me to learn or understand something better. Thanks Ensign.

The Balance Specifics of Guild Wars

This article has been in the works for a very long time, since I've started pvping actually. It is a cumulative collection of my personal thoughts of what GW is and what it could or should be, combined with general balancing guidelines that every competitive game should be built on. That said, this article applies almost exclusively to the pvp aspects of Guild Wars. There is a lot of math involved. I've stayed away from complicated abbreviations in the formulas so that the average person can still understand it.

Basic Math

Guild Wars is a game based on character skills. Equipment makes a small difference, but in the end, a team with all skills and no equipment will always beat a team with good equipment and no skills.

Every skill in the game can be measured by its power level. A skill's power level depends on up to hundreds of things, things like "how much damage does it do?" "Does it shut down another skill?" "Does it shut down a whole character?" All of these things can be measured, most of the time with some error, of course.

Skills that only heal and/or do damage are easy to figure out. Their power level is how much they heal or do damage, minus their costs. Costs can be anything. It can be an energy, adrenaline, or health cost. It can be a skill disabling cost. It can be a time cost. It can cost a skill slot to put on your bar. We can find how much costs are weighed by looking at staples.

Staples are components of a game that contain abilities in their simplest forms. Most RPGs have a healing staple and a damage staple, and Guild Wars is no exception. Luckily, we don't even have to look at those - Guild Wars tells us how health relates to energy through armor insignias. Based on this, we can conclude that 1 energy was meant to be the equivalent of 5 health, though when you actually look at the skills, it turns out to be about double. There are many reasons for this, some being Armor and general Power Creep.

Net advantage is how powerful a skill's effect is minus the value of its combined costs. Therefore, the net advantage of a skill like Otyugh's Cry (before it was buffed) was negative, because it has some costs, but no effect. In Guild Wars, almost every skill has a positive net advantage. This is a good thing - it progresses the game. If all skills had a negative net advantage, nobody would ever use them because you'd always lose more than you gain.

Back to the basics - let's say a skill's net effect advantage goes up 1 point for every point of damage or healing it does. For example, Orison of Healing's net effect advantage would be somewhere around 60 with high healing prayers. Orison of Healing is Guild Wars' healing staple. Vampiric Gaze steals 60 life at 15 blood magic. Its net effect advantage is 120, because it heals for 60 AND does 60 damage. Flare's net effect advantage is 50 because it does 50 damage. You get the idea.

Now for costs. Like I said, costs can be anything. We call these anythings "resources." Your energy, health, and adrenaline bars are resources. Time is a resource. Even your positioning is a resource. The more resources a skill requires to use, the weaker it is. In Guild Wars, you can typically do 10 damage or heal 10 health for each 1 energy you spend. Many skills follow this pattern, including the ones listed above.

A skill's total net advantage, meaning "how much do I get vs how much do I lose" can be found using these rules. However, there are a few rules that add complications to the mix. What about skills that only target yourself? Or an area? What about skills that don't always work, like interrupts or Diversion?

The net advantage of a skill can be found using this equation.

Advantage = (Net Effect Advantage - Net Effect Cost) * Net Viability

Viability is the magic variable that changes with the more interesting skills. A skill's viability can be found using this equation:

Net Viability = Average Usability / Average Chance to be Countered

For skills like Orison, this is (always)/(always). Damage is a counter for healing, and healing is a counter for damage, and 100% of teams bring both of those. Let's look at a skill like Remove Hex. Not every team brings hexes. Even if they bring a few, like the occasional water snare or diversion mesmer, you have no guarantee that you will be able to use Remove Hex, or if it will even accomplish what you want it to. The best way to measure its usefulness is to multiply the percent of teams it is useful against times the percent chance someone is hexed while the skill is recharged, or in simpler words, how often it will actually do something against that particular team. Let's say 50% of teams in gvg are running some sort of hexes that are threats (as opposed to parasitic bond spamming), and half of that 50% is running full-on hexway, while the other half is running a single hex like Migraine. Remove hex will work against 50 of the 100 teams you might face tonight. Of that 50, 25 of them are hexway, in which case remove hex is always useful (while recharged). 25 more teams only have Migraine, and no other hexes. Assuming you have perfect timing, remove hex will work against BOTH of the types of teams running hexes. Overall, remove hex is 50% viable tonight. Pretty good, considering you don't need to spend attribute points on it.

Let's look at a more complicated example. What's the viability of Restore Condition?

I'll skip to the math. First, I assume 1 energy can heal 10 health, and 1 second is the base time cost of all skills.

Net Advantage = ((50*conditions removed)-(5 energy)) * Net Viability

Where Net Viability = (90% of teams running any conditions) * (90% of those teams have no way to counter RC)

So the total net advantage is

Net Advantage = ((50*conditions removed)-(5 energy)) * 0.81

Substituting 50 health for 5 energy (because they're the same thing) and doing some algebra

Net Advantage = 40.5 * (conditions removed-1)

So RC's Total net advantage is positive as long as it's removing more than 1 condition, and of course, plus whatever values the conditions themselves are to remove. RCing a Deep Wound, for example, is better than RCing blind. That last variable is something that just happens. Player skill goes into it. Good RC monks will only use it on people with many conditions, unless they have plenty of stock energy. However, there is another cost we haven't taken into account: RC is an elite skill.

Being elite is a value on the skill. We don't know what it is because the game never tells us. However, we can figure it out. Nonelite skills only have one metagame drawback - they take up one skill slot. Elite skills have an additional drawback. In addition to taking up one skill slot, they also take up your elite slot, an imaginary skill slot that could be explained as a 9th skill. In this regard, elite skills by default as half as good as regular skills. This is why their effects seem almost twice as good as non-elites (in most cases).

In any balanced game, you want to have a power level as a staple. Let's say that power level (net advantage) is 50 for Guild Wars, if we use the above value system. RCs becomes:

20.25*(conditions removed-1) + value of conditions removed.

The condition values can be plugged in if you know them. Removing only a deep wound from a character with 600 total health means RC's power level is 20.25*(0)+100, or just 100. Again, you can't know beforehand what RC will be used on, so you use the average chance of use formula to guess it.

For skills whose usability can be guessed, you can actually say "Skill A is X times better than skill B."

Effect Uptime

The above method for finding the power level of one shot skills is easy. Unfortunately, it's not always that simple. Some effects have durations. Obviously, a Life Siphon with a 10 second duration is better than a Life Siphon with a 5 second duration.

An effect's Uptime is simply its duration divided by its recharge (remember to take cast times into effect). The uptime for Order of the Vampire is (5 second duration-2 second cast)/5 second recharge, or 60%. Other effects may change this value, such as enchanting weapons, Nature's Renewal, or Tranquility.

Duration effects have the unpredictable side effect of being removed before the duration runs out (only enchantments, hexes, and conditions). Again, we can't know exactly how long each effect will last before it happens, so we make assumptions based on the metagame.

For durational effects, multiply their total net advantage with the following:

(Duration of the effect) * (Average Recharge of its viable counters)

_________________________________________________________

(Effect's recharge)^2

It took me a long time to derive that, so I'd rather not go into it. Let's just understand it with examples. This value will be 0 if either the duration OR the recharge of its viable counters is 0. Makes sense, right? RC has 0 second recharge and 0 second cast time, so all conditions are useless.

Let's say two skills with identical effects have different recharges, but the same uptime. For example, let's say Healing Breeze has a 10 second recharge and a 10 second duration, while Super Healing Breeze has a 20 second recharge and a 20 second duration. Equal uptimes. However, the 20 second one has twice the chance of being removed before it recharges, which makes it half as good. Let's plug in some numbers, and assume enchantment removals recharge in 10 seconds, on average.

10*10 / 10^2 = 100/100 = 100%

versus

20*10 / 20^2 = 200/400 = 50%

The 10s version is exactly twice as strong. (Keep in mind this does not take other costs into consideration. If these two skills cost the same, the 10s version would be twice as expensive to keep up 100% of the time, so they would actually be equal overall).

Durational effects and their counters are one of the most commonly seen things in pvp. Durational effects without counters (like shouts or weapon spells) can ignore this equation. Instead, treat it as a one-shot effect by calculating the total advantage over its duration. Weapon of Warding, for example, heals for 60-80 health, plus whatever damage you stop with blocking.

Punishment

Also knows as Negative Reinforcement, Punishment is based on the following idea: The stronger your opponents get, the stronger you should get. Skills like Smite Hex, Cure Hex, Reversal of Fortune, Prebuff Wail of Doom, Diversion, and Bull's Strike all use Punishment, and that's why they're so inherently powerful.

Let's start with a simple example. Reversal of Fortune is often considered one of the best skills in the game because it has the capability of causing up to 160 life advantage, just for 5 energy and almost no cast time. It gets more powerful as your oppponent's offense gets more powerful, and that's why it's ok for it to be strong. If you rof someone getting wanded, it's only going to prevent and heal 20 damage. If you use it right before someone takes a Lightning Orb to the face, you not only negated the damage, that person's health just spiked up!

Punishment is my personal favorite balance mechanism, there is just so much you can do with it. Skills which use punishment ON TOP OF being overpowered break the game. An example of this is Xinrae's Weapon. No matter what happens, it's going to cause 160 life advantage AT LEAST. On top of the 160 virtual damage it just did, it's also going to reduce the damage of the next thing that hits you down to about 30. Let's say someone uses Victorious Sweep on you, gets a doublesundering critical, and gets +30 damage bonus from Strength of Honor and a Conjure. It will probably come out to be somewhere around 200 damage, which Xinrae's will cut down to a net effect advantage of 160+(200-30) = 330. Rof can almost do that too, but let's see the difference between them. Remember, we're only finding out the effects, not the costs.

Rof's effect advantage range = (damage your opponents do)*2, maximum 160 = 0-160.

Xinrae's effect advantage range = (80+80)+damage your opponents do, maximum damage taken: 30. = 160+(infinity-30).

Obviously, nothing in GW can do infinity damage (anymore), you'd have to use calculus integrals to find out that last value (which would be a limit), but it would be somewhere around 60, so a rough estimate of Xinrae's average effect advantage is 240, compared to rof's 80. Being elite is the only difference between the two, and Xinrae's, even after its power level gets cut in half, it still much better than rof.

Attributes and Skill Slots

In the introduction I stated that Guild Wars uses a crippled balance mechanism. That mechanism is Attributes.

In most RPGs, you get your skills powered up with skill points. One skill point = one skill goes up a level. This is not the case for Guild Wars. 1 "skill point" in Guild Wars (actually attribute rank) can increase your skills by any number from zero to 8. If you're an ele running all fire skills and a res sig, each point you put in fire magic is increasing your total skill rank by 7. Since you have more than enough attribute points to max out 2 separate atts, you can obtain maximum attribute efficiency by having exactly two different attributes among the skills on your bar. Most builds do this. Touch rangers, domination mesmers, monks, single-attribute eles, most warrior builds, some of the most popular builds in history, all powerful because they maximize their attribute power. Primary attributes and their compensations throw an even bigger problem into the mix.

To try to make up for the imbalance caused by attribute lines, Guild Wars has a very strict skill limit of 8 plus one elite slot. This severely limits how far attribute points can go. If the skill limit was 16, each attribute point could max out a whole skill, on average. This makes your skill slots a resource, not only while preparing for battle, but even during battle.

Blackout is a favorite skill of many mesmers. You take someone out of the game for 5 seconds by touching them? Sign me up! Fortunately for balance, there is a severe drawback - you can't use skills either. Blackout doesn't do or prevent damage, nor does it have any real tangible effect. However, it changes the resources of two people on opposite teams, and that's something we can count.

Blackout's Net Advantage = (Enemy loses 8 skills and all adrenaline for 6 seconds) - (You lose 7 skills, 10 energy, and all adrenaline for (1+5) seconds)

This skill looks terrible. Its net advantage is...they lose 1 skill for 10 energy. What kind of POS skill is this? Blackout disables skills. But skills don't have a mathematical value of (1 skill), they have a mathematical value equal to their net advantage, or their power level. You can make optimal use of blackout when:

1. You have no other skills recharged (so you're not losing any of your other skills), or the skills you have recharged are weak or unneeded.
2. Your enemy has more powerful skills available than you do.

If you're shutting down someone with 8 Wail of Dooms on their bar (mathematically the best skill in the game) while you have nothing else recharged, Blackout's power level goes into the thousands!

Any skill that disables or prevents certain skills from being used actually takes the other skill's power level into consideration when determining its own. This is why skills like Diversion and Distracting Shot are so powerful, even though they don't do any damage.

Randomness

Game designers and competitive players cringe when they see the word "Random," especially in a computer-based game. Technically, computer randoms are not really random, they are pregenerated at some specific time and a routine cycles through numbers that already exist, so it's more unpredictable than it is random. Computer randomness, though, isn't what makes us cringe.

Randomness and player choice are complete opposites of each other. In a game of choice, you want to keep random occurrences to a minimum. The less randomness you have in your game, the more players have to rely on their own skill, and vice versa. Before we get to the specifics, there is one subtle detail about randomness you have to know. 50% is the most random you can get (when choosing between two outcomes), and everything else is a deviation from that.

Let's look at three separate mechanics that exist in Guild Wars, all of which utilize measurable randomness to some extent.

Blind - "You have a 90% chance to miss with attacks." This is identical to "You hit 10% of the time." Blind is a powerful counter to physicals, for obvious reasons. Technically, this stops 90% of the damage (90% chance of 100% damage reduction = 90%) you put out. While this is a powerful effect, the randomness aspect (which is actually 10%) is pretty small, and in the big picture, it doesn't cause that many problems. Yeah, pushing an Eviscerate through blind to kill something is always nice, but it happens so rarely that it's barely worth mentioning, and the chance of it happening is so low, most players save their spikes for when the blind is removed or ends.

Aptitude not Attitude - "You have a 20% chance to cast skills of this item's attribute twice as fast." Blind was 10%, let's look at a 20%. Rather than looking at the overall spell casting speed you get (which is dumb in this case), let's look at the direct result of this inscription. People don't use it to cast more spells per minute; they use it to bypass interrupts. 20% chance of being immune to interrupts (assuming you're not using this to fast-cast meteor shower) is an extremely powerful thing to bring with you, especially to high level pvp where interrupts are plentiful. Considering most rangers get 5 interrupts every ~18 seconds, having amnesty from them once in that period of time can change the tide of battle. Here's the thing - you can't control when it happens. You can't choose to have interrupt immunity on your snare, or your guardian. There is actually nothing you can do about it. You have three choices because of this mod:

1. Hope you get lucky.
2. Weapon swap and lower your chance to FC to 0%
3. Don't cast anything.

Obviously, options 2 and 3 are not viable, so the number of choices you have is 0 (a choice between one option isn't a choice). For the person interrupting you, they have identical choices:

1. Hope they don't get lucky. (interrupt)
2. Don't interrupt them.

For key skills, 2 isn't a viable option to make, so again, the number of choices on the interrupter is 0 as well.

Guardian - "For 5 seconds, target ally has a 50% chance to block attacks." This skill is evil. It is a pure random (50%) effect that sometimes makes you immune to damage, and sometimes does nothing. Just like fast casting mods, this lowers the amount of choices players can make (in this case, how many your enemy can make). However, they're a little different from the fast casting mod:

1. Keep attacking, hope you hit (sacrifice ~50% of your damage)
2. Change targets.

Unless your target is almost dead, 1 isn't a viable option. This almost never happens between teams of skilled players, because monks with 1 health cast heals, not prots that fail half the time.

So far, we've discovered that randomness subtracts from player choice, but why exactly is this bad? Well, because if a computer game has no choices, it's not a game. It's a movie. Games are most exciting when players have a lot more to do. Randomness is a detriment to competitive games. In very small amounts, it can be manageable. What if guardian lasted 25 (5 times as long) seconds and gave you a 10% (1/5) chance to block? Statistically, its long-term outcome would be identical to the current form, but it would be far more fair.

Last Words

If the staple power level for every skill is 50, EVERY skill's average power level should be 50. Maybe one skill does 20 less damage, but heals 20 health. Maybe one skill does twice as much damage, but only works half the time. Those are fair tradeoffs. Removing a player from the game for the equivalent cost of 6 energy is not balanced, and it never will be.

Guild Wars is like physics. There are rules that must be followed. When one of these rules becomes broken, so does the game. The popular defense for the game balancer is "You can't balance 1000 skills." The simple reply: Yes I can. Every skill in this game is based on the same rules, the same costs, the same restrictions. It's like turning a base 10 math equation into a base 8 math equation, or into hexidecimal. Is it brainless? Of course not. Is it easy given time and knowledge? Yes. There is absolutely no reason why any skill in GW should be 2, 4, 10, even 120 times better than another one.