Wednesday, November 09, 2011

Talent Trees

There's been a lot of discussion of the Mists of Pandaria change to talents, essentially the removal of talent trees, and adding in a new system. Most of the discussion has focused on World of Warcraft and what the change means. But I would like to step back and look at the broader, more interesting view.

In the gaming industry, Blizzard raised up talent trees for RPGs. Previously, the main place you saw talent trees were the technology trees of games like Civilization. But with Diablo II and then World of Warcraft, Blizzard made talent trees truly mainstream.

And yet, Blizzard is abandoning the talent tree concept almost entirely. Starcraft II did not use talent trees, Diablo III will not use talent trees, and World of Warcraft will also stop using talent trees.

So that begs the question, what is the fatal flaw of the talent tree concept? What made talent trees appear like a good solution at first, but now appears to have been a blind alley?

Talent Trees and Time

Let's start by looking at what exactly talent trees do. When you first encounter a talent tree, you are presented with a few choices. You pick one of those choices, and that may unlock future choices.

The key thing here is talent trees essentially have a "zone of choice" which is limited to a small fraction of the total tree. As time passes, the zone of choice moves further down the tree. However, it does not usually get much larger or smaller. Almost always, you still limited to choosing from a small number of talents.

What's really important about talent trees is that while you are moving through the tree, the order in which you gain talents is very important. For example, if you are playing Civilization, you can go straight for Literacy, but you give up other talents, like Calvary.

However, this choice is a temporary choice, not a permanent choice. In the end, you usually get all the necessary talents you missed. After you pick up Literacy, you go back and grab the Calvary and declare war on Babylon.

As another example, in RIFT, I'm leveling a rogue. The end goal is Ranger 31/Marksman 18/Assassin 17. The top of each tree gives a specific benefit. Ranger gives a new pet, Marksman gives weapon enchants, and Assassin gives bigger crits. While leveling, the tree I focus on makes a big difference. Do I want the big crits early, can I wait for the pet, etc. But at max level, that doesn't matter anymore. The path to the final build is not important anymore. Only the final build itself matters.

Between those two aspects, we start to see some of the issues with talent trees. The small zone of choice explains why, even though there maybe 30 or more talents, it always seems like you only get a choice between 2 or 3 talents at the end. That's where your zone of choice has been pushed to.  And how the talents were assigned, point by point, ceases to matter.

In a lot of ways, I think the above points demonstrate why talent points were embraced at first. Zone of choice and the order of points mattering are great features when you are looking at leveling. But when time stops mattering, those advantages aren't really advantages any more.

Explicit Choices

In a talent tree, almost all choices are implicit. Every talent matches up to every other talent that can be taken at the same level. There is no idiom for making an exclusive choice within a specific talent tree.

To see what I mean, let's say that you have two talents, Talent A and Talent B. You would like a Holy Paladin to be able to take one of those talents, but not the other. Where do you place the two talents?

In the current WoW talent trees, you place talent A in the second tier of the Protection tree, and you place talent B in the second tier of the Retribution tree. In the old talent trees, you would have to place one talent at the top of one tree, and the other in a different tree just out of range (31/21, etc.)

Kind of honestly, that just doesn't seem right. Talent trees make it very hard to offer exclusive choices. You have to assume that any player can take both talents. For example, a Fury warrior can take both Titan Grip and Single-Minded Fury. It's a bad idea, but nothing in the tree structure prevents it.

The lack of explicit choices means that all talents within reach are compared to each other and ranked accordingly. And this has lead to some awkwardness. For DPS, damage talents always outweigh survivability talents. For tanks, survivability talents always outweigh damage talents. So for a final build, the only choice is between a few talents that you don't really care about, that aren't really important to your role.

In a talent tree, it is extraordinarily hard to make someone choose between two good choices. Most of the time the player will take both, and drop a third, less important choice.

The New System

The new system that Blizzard is using in Mists of Pandaria--and the one they used in Starcraft II--is a series of explicit choices. This system is simpler, and does in the end offer more real choice than a talent tree does.  You have to choose between good talents, instead of being able to take them both.

The other major advantage of this system is that it allows Blizzard to easily offer a choice between a less powerful but lower-skill option, and a more powerful but high-skill option.

For example, let's look at the Tier 3 Paladin talents in MoP:

Sacred Shield
When reduced below 30% health, you gain the Sacred Shield effect. The Sacred Shield absorbs 18005 damage and increases healing received by 20%. Lasts 15 sec. This effect cannot occur more than once every 60 sec.

Ardent Defender
Reduce damage taken by 20% for 10 sec. While Ardent Defender is active, the next attack that would otherwise kill you will instead cause you to be healed for 15% of your maximum health. (Instant, 3 min Cooldown)

Ardent Defender will be the choice of all high-end tanks. But it does require precise timing to use to full effect. Use it too early, use it at the wrong time, or simple not use it when you should have used it and you might die.

Sacred Shield, on the other hand, is a decent option for a less experienced tank. It procs automatically, is one less option you have to think about, and will help with survivability. A less powerful ability that actually gets used is better than a very powerful ability that is never used.

But the point is that this type of choice is very hard to offer in a talent tree. If both talents are in the Protection tree, then every tank will take both talents and drop some threat or utility talent instead.

This is the major significant benefit that the dev team gets. Each talent doesn't get compared to the entire tree, only the other talents in the separate tier. Utility talents match up against utility talents, healing talents against healing talents, and damage talents against damage talents.


Though talent trees were very attractive at first, the lack of explicit choice has proven to be a fatal flaw. The new system is simpler, easier to understand, and has the potential to offers more significant choices than talent trees could have.


  1. I still think Blizzard could leave talent trees in the game during leveling, when they are great.

    At maxlvl you then have the tree filled out (almost) completely and go for the real endgame-choices.

    I don't see no reason to replace the talent trees during leveling with a one-choice-every-15-levels thing. The only reason I can find is that Blizzard doesn't care about the leveling ...

  2. Great Post. I love the new concept. Your example with Paladin T3 illustrates your point very well.

    My slight concerns around the new system revolves around two issues. Granted, I haven't read much about this so I may be in the dark to info others already know about.

    First, how will glyphs work now? It would almost be best if they only effected abilities that come from your class and spec and not talented ones. I say this because if you are comparing three abilities fine, if you have to compare three abilities and how they might be effected by glyphs - that's confusing.

    Second, again I might be missing something here, but if you are a healer it would be nice if every time you got choices they had to do with healing and not other things. I realize that at endgame this may not matter much, but it would make leveling feel more rewarding.

  3. I believe that in the end it won't matter. The field of theorycrafting is such that people will crunch the numbers and will say which talent to take, and people will follow their lead.

    It's too late for Blizz to put the genie back in the bottle, and theorycrafting along with its side effects are here to stay. What might end up happening will be a scenario where guilds will take two Mages into a BG, one with Ring of Frost and one with Cone of Cold. Want to join a rated BG? Well, we've got two Mages covered, so take a Pally but make sure you have Sacred Shield.

  4. I agree that this new system feels better and more meaningful than the current talent trees.

    @Jeremy, I think the design intent is, for the most part, as you describe. Glyphs enhance class and spec abilities, and talents add new abilities. There's a few exceptions, like Fist of Justice, but not too many.

    Like Nils, I'm am a little disappointed that Blizzard seem to have just given up on the leveling game. I guess they had to make a choice about it. Gevlon made a few posts last month about how the end game of WoW is a totally different game to the leveling game, and how this can easily spoil the fun of players. It seems that rather than try to bring the two games closer together, leveling is just being marginalised further. Expect plenty more heirlooms in MoP.

  5. I'm not saying that talent trees are the best system for leveling. Just that trees are better during leveling than at endgame.

    Most of the time, trees are just a whole bunch of tiny choices that don't really affect gameplay. You make very few important choices.

    Like the Rift example. 66 talent points, and really only about 10 of them truly matter. Everything else is essentially "increases damage by 0.5%".

    It's quite possible that the new system will turn out to be better for leveling *and* endgame.

  6. Completely agree with Rohan here. The only thing you lose while leveling the is the ability to push a button at level up. The literal choice was 100% meaningless in TBC as you can see from talent trees I edited here.

    Whether Blizzard let the leveling game go to weeds is an entirely separate issue from the talent system.


    Blizzard answered the "but there will always be cookie-cutters!" charge in the latest Q&A:

    A: Since so many of the talents focus on survivability, movement, and utility we are skeptical that there will ever be a talent build that is the perfect build for every PvE fight in the game. It is likely that as players learn specific encounters, each spec finds an ideal set of talents for that encounter. Those will be the "cookie cutter" builds. However, that will mean that players are interacting with the system and picking a unique set of customizations on a frequent basis. This is a vast improvement over a system that is solved once by a dps spreadhseet and then everyone copies that build once and ignores their talents for the rest of the expansion. In addition, there will be likely disagreement over which talents are best for which encounters.

  7. > If both talents are in the Protection tree, then
    > every tank will take both talents and drop some
    > threat or utility talent instead.

    This could probably be solved by something like what The Witchers used with its bronze, silver and gold points. Certain talents require gold points which you only have very few.

    > The small zone of choice explains why

    The small zone of choice was introduced by the Cata system which locks you into one tree until you have every useful talent.

    The vanilla system didn't have this problem. You had a much bigger zone of choice and the result were many "hybrid builds" which didn't put 31 points in one tree.

    I think the Cata system had to die because it was never a talent *tree*, it were only *gated* talents. They could have put all the effects of the talent system as passive talents on the class trainer.

  8. I can see Sacred Shield being a superior choice in a fight using the Chimaeron mechanic, I suspect we haven't seen the last of it.

  9. Great post. I'm surprised Talent Trees won't make it into Diablo III. In a game where you're not as focused on optimizing your role in a group (like Diablo), Talent Trees offered tons of opportunity to customize a character. There are still pages and pages dedicated to dozens of builds for each Diablo II class.

    Anyway, the real killer is the cap on the number of talent points which prevents any real customization since everybody needs to optimize role fulfillment. Occasionally you can flip a talent or a glyph for a specific fight mechanic (see Rohan's Holy Wrath Glyph post). Seems like a good move for WoW.

  10. @Kring, the zone of choice is small compared to the entire talent trees in all talent trees. The small zone is good thing, it means that at any given time your menu of choices is smaller than the total choices available. It sort of "rations" out the choices as time passes.

    It's a certainly larger zone in the older trees. but given that most of the talents in the other trees are for other specs, it's not really that big of an advantage.

    Maybe for the pure dps classes, but given that pure builds tend to be driven by the last points, to reach a certain level in each tree.

  11. @Azuriel-- I don't see how the blue poster can say that constantly changing your talent tree on a per-fight basis is an improvement, since there will still be an optimal talent tree agreed upon.

    I just think that the law of unintended consequences will be at play here.