Designing Skill Trees


Continuing the discussion from Defender’s Quest 2 - Weaselmancer and cast preview:

Took me a moment to find this fabled feature of yours. ^^

My problem with the atomic A-or-B choices is that they don’t often fix the underlying problem we were having with the old Skill Trees, that is, options weren’t balanced. Even in Bastion, there was always one more correct choice for me, and often it boiled down to what I valued more, which often was what would make my experience with using the weapon better. Reload and ammo count upgrades were a major preference for the ranged weapons so that the reload cooldown would break my combat flow less. Removing bullet scatter was another one like that.

And this preference-based optimization never changed, so barring special achievement requirements, I never bothered with it ever again.

But this worked for Bastion because it was basically an effect of leveling up your weapon. You made it more powerful or you made it more like you wished it was, and level up done. If you regret the choice or something, you can go back and change that particular level, because it wasn’t a talent tree so much as a build optimization. The builds being your two weapon choices and one skill choice.

The atomic skill tree in Wings of Liberty worked much the same way. It was catering your units to your play style. There was always one more correct choice based on power or play style preference, and once picked, you never wanted to go back unless you decided you screwed up or it wasn’t what you wanted for some reason. And the three-way atomic skill tree in WoW bit me, because it was basically asking me to make a ‘play style’ vs ‘most correct’ vs ‘least correct’ choice. There was the talent choice I wanted to pick, the talent choice all the guides said I should pick, and then the choice nobody picked. This was more annoying in general, so I’m not entirely sure it was a gameplay improvement when I had to change my choices every boss fight or so to optimize my build against said boss. It became a situation of knowing what I needed to pick beforehand, and picking it whether I wanted to or not. That’s not fun in the end. That’s doing your homework, checking the appropriate boxes, and then either doing what you were going to do with the blue shield rather than the green shield or finding yourself with a character that’s less fun to play as.

Which is to say that the basic atomic skill tree falls well short of what I think most people want/expect from a skill tree. Especially in a tower defense game, where it’s integral to the feel of the tower becoming more powerful over time.

So coming from my Blizzard/Bastion/TD/etc experience, let me suggest something a bit different that somewhat incorporates the idea of choosing distinct tower upgrade paths. Plus some variations.

Everything costs 1 point to get, and I think an interesting idea that would make this work better is if each character got Skill points and then passive points. 3-4 Skill points total, so there would always be 1 or 2 skills a character didn’t have, then a limited number of passive points to emphasize certain skills or abilities or synergies. Passives could be a mix between stat bonuses (Attack/Defense generics) or effect specific bonuses (Bleed power/HoT applied on Heal) or skill specific bonuses (Skills 1 and 2 gain increased armor penetration/Ice Skills get +1 range).

If you go for the atomic passive choices variant, you could probably do it like Bastion, where they have two or four or six ideas for the way a skill can develop, so the player can choose a mismash of these development paths they want at each tier.

I think this simplifies things down akin to what a simple atomic tree offers without completely sacrificing the diversity of interesting choices and possibilities that a more standard tree offers. A compromise between ease of balance and fewer, more potentially interesting choices on one hand and the desire for more diverse and tunable outcomes on the other.


Why not just make skill B scale with a previous/weaker skill of the same type? This keeps that simple skill tree where you just want a skill without the others, but still has the feeling of a branching skill tree.

For example, there are 2 skills, Fireball is A and Explosion is B, and they are ordered like Lars proposed (Fireball is unlocked before Explosion, but the player chooses skill C before Explosion). If the player chooses Explosion right away, they’ll get, say 100 damage in a range of 3 tiles per explosion (let’s assume the character throws a bomb). And by upgrading, they can increase that to 500 in 4 tiles. Now, assuming Explosion is maxed and Fireball is gotten, then Explosion’s power can increase slightly, and add a little burning effect (and even a visual effect! :O) or something similar to Fireball.

And that’s it. I’m not that good at writing long walls of text.


Because there isn’t a skill with a previous/weaker skill of the same type?

Each skill is its own tree branch, that’s the entire point of my setup. You have your primary interesting choices that are the skills, and then the secondary interesting choices that modify the primary choice in some way. The shield bash skill unlocks options to be tankier. The power strike skill unlocks options to defeat armored opponents. The slash skill unlocks improved bleed effects or slow effect options to make the character more DPS oriented. You have synergies between skills mixed in too to further emphasis one specific aspect of a character.

The binary choices are different in that they’re like ‘Fireball Skill + improved DoTs’ or ‘Fireball Skill + improved explosion radii’ then the next tier is ‘Explosion Skill + improved Dots’ or ‘Explosion Skill + improved explosion radii’.


I like the idea of each skill branching in a different way though, but I don’t think that skills should vastly change the stats, as they can easily make a character overpowered by having some skills mixed together. Unless the developers take this into account. Huh, that’s actually a good idea!


I’m cool with waiting to see what sort of fleshed out system Lars actually comes up with for skills and A-B selections. It could be that A is better against singular enemies and B is better against groups, making it so that they’re both good depending on circumstance. It’s not a very groundbreaking system, still, but at least there’s some choices to be made there. :stuck_out_tongue:

The game is far and away from being done and next to nothing combat-wise has been revealed, I don’t have enough info at the moment to critique anything.


Well, Lars kinda did throw out their current idea for a skill tree - a sequence of binary choices of which powerup do you want for a character.

Maybe I should mention one other problem with the atomic weapon upgrades from Bastion that were (or seemed) balanced - choosing one or the other didn’t really matter. I remember the hammer in specific for this, where it asked you if you wanted a flat damage increase or a crit damage increase. It seemed balanced, so the question was ‘do you want smoother damage output or spikey damage output?’ I decided smoother damage output except when going for the achievement that required a crazy critical strike. Other than that, the choice was immediately forgotten because it didn’t seem to matter in the end. It was like, kill a medium target in two hits always or sometimes pop a medium target in one hit and sometimes do it in two and more times do it in three? Kinda all evens out in the blur, doesn’t it?

So the only options that mattered were ones that changed how a weapon behaved or were unbalanced where the numbers said one was the clear winner. So the only options that mattered were ones that changed how the weapon behaved.

In the case of a TD game, the atomic skill tree would only be meaningful to the extent that the options offered a clear mechanical role difference. This creates a challenge in exactly how many of those distinct choices across thirteen characters are actually interesting ones (as opposed to 10% health vs 5% damage). But then the process for the player becomes ‘Do I need role x?’, if so, check off all ‘role x’ choices. Any remaining options then become the interesting choices for the player. If a character has three roles, and each role has an equal number of options, then only 1/3rd of the atomic tree is going to consist of interesting choices and the other 2/3rds is something of a chore during a role swap. There can be some gradient to that, as in, ‘I only need this character to be 80% role X so 20% can be shifted to role Y and Z’, but still. If you’re crunching the numbers, may as well have two drop downs - Primary role and secondary role (preset to ‘damage’ if not primary role) - and save your player the trouble.

Well, that might be a little harsh but probably a given. Where the atomic skill tree could shine is in options like ‘blind effect vs slow effect’, which cycles back to exactly how many of such choices can you put in thirteen talent trees?


I quite like the idea of a few major choices (maybe we could call it specialization/build/whatever), with individual skill points assigned to various “modifiers” of the specialization.

Example: Each character could choose a specialization at 1st level and another specialization at xth level. The skills could be arranged in a circle, allowing you to choose only the neighbours of the primary choice, or choose the primary choice again, unlocking more powerful subskills or increasing level caps of subskills.

Berzerker could have the following specializations:

  1. rage
  2. speed
  3. survival
  4. bleeder

And you could only have neighbours - rage would not allow survival as secondary and bleeder would not allow speed.

Possibly subskills could have different caps for primary and secondary specialization (i.e. primary 10 points, secondary 5 points, repeated primary 14 points)

Knight: tank, armor-stripper, stun/knockback and raw power
Archer: Bleed, poison, criticals and attack density
Ice mage: area attacks, piercing attacks, slow and/or freeze
Healer: Healer, buffer, debuffer
Dragon: Poison, Fire, Roar, Nom

The number of primary skills wouldn’t necessarily have to be exactly 4. A character with 3 choices would simply have no impossible combinations while a character with 5 choices would have a lot of them.

This would limit the number of “builds” or rather character roles to a reasonable level, and the individual skill subtrees could then be atomic or not, somewhere with skill synergy, somewhere plain…