Skill Levels
Contents
Introduction#
The main feature that makes each doll interesting (or boring) is their skill. I'd be very cautious of an analysis that spends most of its time talking about the specific stat values of a doll (perhaps with an exception for nitpicking between otherwise identical selfbuffers), as skills carry 99% of their functionality and most dolls are similar enough within their class by stats.
You can invest training data to level up skills, which will generally improve multiple aspects of it:
- Skill duration
- Skill multipliers
- Skill cooldown
Here's a very basic example:
Note that ICD is not affected. Especially as the game goes on, more complicated skills have been introduced with less straightforward improvements:
Skill CD goes from 20 to 16s, erosion's debuffs are doubled and its duration increases, the chance of applying erosion doubles, the 100% erosion rate after the active continues for 2 seconds longer, and the 5-stack effect deals twice as much damage and stuns for 0.9 seconds longer.
Trying to compare the performance of all skill levels of something like this is already giving me a headache, so we'll stick to typical buff/selfbuff skills.
Effect on DPS#
Let's go back to the simpler DPS skills. Adding more skill levels is obviously better, but you don't get that much data from simulations and going to the shop can get expensive. So how much are you really getting out of each skill level? Is that advanced data really worth it?
Here's the cumulative damage output of a certain doll, at all skill levels, compared to no skill use.
They're all equivalent before 4 seconds, start out similarly (as the pre-skill damage is still most of their output), and diverge over time (as in-skill damage starts to dominate). Compared to not using G41's skill, anything is clearly a major improvement. To better look between the skill levels, let's normalize around her SL1 performance instead.
At its peak, a SL8 G41 will have dealt about 21% more damage than a SL1 one. At SL10, she gets up to +29% damage dealt (~7% more than at SL8). A 7% increase in damage is debatable in value, but it's hard to ignore a 20%+ DPS boost. Clearly, skilling DPS dolls is worthwhile - this increase is comparable to the offensive aura of a midrange 5-star fairy.
The lines flatten out a little more over longer time periods:
It's also worth considering that skill levels get more expensive as you go on. For normal dolls,
| Skill level | Basic Data | Intermediate Data | Advanced Data | Energy needed | Gems needed |
|---|---|---|---|---|---|
| 1 | 0 | ||||
| 2 | 100 | 0.457 | 3.0 | ||
| 3 | 200 | 0.913 | 6.1 | ||
| 4 | 300 | 1.370 | 9.1 | ||
| 5 | 120 | 2.424 | 16.2 | ||
| 6 | 200 | 4.040 | 26.9 | ||
| 7 | 300 | 6.061 | 40.4 | ||
| 8 | 400 | 8.081 | 53.9 | ||
| 9 | 200 | 11.321 | 75.5 | ||
| 10 | 300 | 16.981 | 113.2 | ||
| Total | 600 | 1020 | 500 | 51.648 | 344.3 |
And, looking at the uptime-averaged (long-term) DPS, we can do something like this:
| Skill level | Average DPS | % DPS gain/additional gems spent |
|---|---|---|
| 1 | 944 | - |
| 2 | 963 | 0.67% |
| 3 | 986 | 0.39% |
| 4 | 1008 | 0.25% |
| 5 | 1034 | 0.16% |
| 6 | 1064 | 0.11% |
| 7 | 1095 | 0.072% |
| 8 | 1127 | 0.054% |
| 9 | 1176 | 0.058% |
| 10 | 1222 | 0.035% |
Uneven multiplier scaling makes it slightly rocky, but the overall trend is clear. Later skill levels improve your overall damage output by less, relative to the investment of energy/gems. Visually:
Of course, different dolls will have different scalings on their skill level. Creating such charts is a bit outside the scope of a single writeup, so we'll just assume for now that most dolls follow similar trends.
For one last point of reference, here's a typical HG buff skill (25% buff, 8s duration, 12s ICD at SL10):
This is the average DPS buff rather than the in-skill effect.
For example, at SL10 the calculation looks like (1.25 buff * 8s/12s uptime) + (1.0 buff * 4s/12s remaining time) = 16.7% average buff.
And a similar trend is seen when looking at the gain in average DPS increase, relative to gem expense:
y-scaling is rather arbitrary.
Special Effects#
A few dolls get to notable breakpoints, gain an extra effect, or get more/less value than you'd expect at certain skill levels. For an incomplete list:
SL1 dolls
| Doll/Fairy/Entity | Justification |
|---|---|
| FO-12 | What FO-12 does: on reaching 100 "Activity" stacks, she instantly kills one non-elite enemy with the highest FP stat. What leveling FO-12's skill does: slightly increase the magnitude of some FP buff after that, and change literally nothing else. Keep her at SL1. |
| M26-MASS | M26 is cool because of her passive that ignores any damage of 5 points or lower. This works straight out of the box. Skill levels only improve the multiplier on her damage-dealing active, which isn't a big deal. You can leave her at SL1. |
| PPK | PPK goes from a 5% FP/25% CR buff at SL1 to 9% FP/34% CR at SL9 and 10% FP/35% CR at SL10. The crit rate is usually excessive and the FP gain is marginal, so spending more than a little basic data on PPK is probably a waste. |
| Scarecrow | Leveling Scarecrow's second skill only increases the damage of its landmine-like effect. This is unlikely to be relevant in any ranking scenario, but you may wish to keep it at SL1 just in case (as it would reduce the points you get from killing anything that walks into it). |
| Reinforcement Fairy | The scaling on this skill is beyond appalling - its main function of healing one link is unchanged, and its EVA buff barely budges from 3% at SL1 to 10% at SL10. Do not invest data in this (or anything, really). |
SL9 dolls
| Doll/Fairy/Entity | Justification |
|---|---|
| G11 | Goes from 2->3 hits per shot at SL9. |
| Micro Uzi mod | SL9 adds an additional tick of burn damage to her molotov. |
| RPK-203 | Reduces her self ROF debuff from 20%->10%. |
| VSK | Goes from 2->3 hits per shot at SL9. |
SL10 dolls
| Doll/Fairy/Entity | Justification |
|---|---|
| Parachute fairy | Infinite range at SL10. Also, the 40% stat penalty makes debuffed fights hard enough. Don't make it even worse. |
| Illumination | Active grants 1 extra node of vision at SL3 and a second extra node at SL10. |
| Carcano M91/38 | Guaranteed proc of skill shot at SL10. Eternal staple in the hardest content. |
| NTW mod | NTW mod needs to kill each enemy to continue her chain shots, so maxing her skills makes her more likely to chain. |
| Python | Active gets a 100% chance to trigger at SL10, improving consistency. |
| Ranking | Dolls you intend to use in ranking should go to SL10, to whatever extent possible. You may end up in a fight that you're just inches away from being able to win, and some extra skill levels could save you an hour of resetting. |
| RMB-93 mod | RMB's mod skill goes from 3 points of damage reduction per stack to 4 points at SL10 - or 9 to 12 points at full stacks. It's a big jump, and this damage reduction is the whole point of her mod. (Her skill 1, on the other hand, is quite safe to ignore as it's just basic damage/knockback.) |
Special Considerations
| Doll/Fairy/Entity | Justification |
|---|---|
| AK-74M | Rather than the usual modifiers, leveling AK-74M's skill unlocks bonuses for different AKs. These are: AK-74M and AK-47 (SL1), AK-74U (SL4), AK-12 (SL6), AK-15 (SL8), AK-Alfa (SL10). If all the dolls in your echelon are covered by a certain skill level, you can pause there. |
| Hunting Impulse MGs | If you don't need the accuracy, Hunting Impulse MGs will generally empty their magazine at 5 seconds (more if there's walk time) so leveling their skill past ~SL3-5 (3s ICD+~2-3s duration) is less useful. Examples: LWMMG, Mk48, M1919A4. |
| M200/SSG3000 | Gets an extra skillshot with the last few levels (with some variation based on CDR tiles, spreadsheet), making them more worthwhile. It's also the main way to improve her in-skill ROF, as typical ROF buffs don't affect that delay. |
| Px4 Storm | High skill levels are extra helpful on Px4 Storm for reducing her skill level penalty. See also: Px4 analysis, DPS charts at each SL. |
| Sniper Fairy | Leveling Sniper's skill makes it shoot significantly faster, increases the damage cap, improves its HP scaling, and adds more vulnerability. In other words, every skill level makes it significantly better. Encounters you'd bring a Sniper to are also probably ones you'd like to end quickly. Raise as much as possible. Example DPS values, assuming a high-HP enemy. |
| Taunt/Twin Fairy | The taunt/twin dummy may be the only thing keeping scary SWAP SF away from your poor dolls. Raise as high as possible. |
Note that these are skill level recommendations if you intend to use the doll. This is not a general recommendation of dolls to raise. Dolls listed as "SL9" can still be brought to SL10 if you use them enough.
Conclusion#
For once, common wisdom is mostly on the mark. As such, there's not much to add here. Just to be safe, though:
- Skills are the key factor that makes dolls worthwhile.
- Leveling skills makes them do this better.
- The DPS improvement at higher skill levels typically does not scale that well with their increased data cost, reducing their value. In most cases, you can stop around SL7-8. It's a bit case-by-case, and there's no universal "best" stopping point.
- You should still take the dolls you use most to SL10, plus some other special cases. You'll get at least some advanced data from events/campaigns, and more DPS is better.
More like this
Myths, math, buffs, and more
Force manual vs Autoskill