Type88: Formation and Skins
Today, I'd like to investigate two ideas about Type88's mod:
- The bullet positioning may change for some of her skins, resulting in more/less enemies being pierced.
- Typical RFHG formations have a rifle on positions 1 and 7. Type88's pierce may be (on average) better suited for most enemies on one of these over the other.
The most straightforward and useful way of investigating these is to do some tests, so on we go! We'll be checking three main metrics to see if she does better or worse overall:
- Clear time. Pretty self-explanatory, faster is generally better.
- Total team damage taken. Less is more.
- Damage contribution. This has the potential to get weird if corpsewhipping/overkill/evasion RNG becomes a problem, but otherwise helps tell us how much of the battle Type88 is responsible for. If we see her damage contribution go up significantly with a specific skin or from a specific position, it's probably better for her.
No singular metric is conclusive, they all have their own problems, and are at least somewhat random so we'll need to look at the overall picture rather than any one test result.
Skin#
Going in, I don't expect this to be a massive difference. I used her default mod3 art for my first Polarized Light ranking run, and everything went fine. I used her default art for the second run, and everything also went fine. If you're losing a battle, it's probably not because of the skin one of your dolls is using.
Personal anecdotes aside, let's get to the test results. First set of experiments: all at night, Type88 paired with WA2000. Type88 is still quite good at day, but having her carry more will make it easier to detect differences in her performance.
And two more with Taunt active:
In general, it looks like there is some chance her performance with East Lake Lotus is a little better than the other two skins. Using her mod skin as a reference, here's the average changes in each metric by skin:
| Skin | Clear time | Damage taken | Damage contribution |
|---|---|---|---|
| Mod art | N/A (+0%) | N/A (+0%) | N/A (+0%) |
| East Lake Lotus | -1.5% | -5.9% | +3.2% |
| Default art | -0.1% | -4.0% | +1.9% |
which could suggest her East Lake Lotus skin is marginally better (which lines up with the usual claims I hear). These are very slight differences, and are very susceptible to RNG - nothing like some other sets of tests where many of the box plots had no overlap at all due to the scale of the performance differences. I'm not yet convinced, so let's get a new echelon, new enemies, and see if we observe the same stuff.
| Skin | Clear time | Damage taken | Damage contribution |
|---|---|---|---|
| Mod art | N/A (+0%) | N/A (+0%) | N/A (+0%) |
| East Lake Lotus | -1.2% | +4.1% | -0.4% |
| Default art | +0.3% | +6.8% | +0.3% |
This time, there's no overall winner. More tests and study would be helpful, but so far I see no strong evidence to suggest changing Type88's skin gives a significant performance advantage. The issue has most likely been fixed. Even if there is a functional difference and one of them is better, the difference would have to be absolutely miniscule. It will not matter 99% of the time. Feel free to use the skin you like best.
Type88's pierce hitboxes are not perfect, so from specific angles and enemy positions she may completely miss certain enemies. See this comment for some examples. Each skin tested gave identical results here. See more tests like that over here.
2021-12-21 update: Client v2.08 has released, making the piercing visuals for Type88's skill line up better with what she's actually doing. It looks like nothing was broken hitbox-wise, as testing continues to show no significant performance differences between each skin:
So the idea of her hitboxes being messed up by skins remains a myth. Careful inspection may show miniscule pixel differences in the laser graphic depending on how you look at it, but you need a lot more than that to establish a real combat performance difference. Percentages in table form:
| Skin | Clear time | Damage taken | Damage contribution |
|---|---|---|---|
| Mod art | N/A (+0%) | N/A (+0%) | N/A (+0%) |
| East Lake Lotus | -0.04% | -3.26%* | +0.19% |
| Default art | -0.18% | -4.49%* | -0.34% |
| Christmas skin | +0.52% | -0.76%* | -0.95% |
These are all well within expected experimental variation. This does not suggest any skin will lead to faster clear times/less damage taken.
*Damage taken in particular is an unreliable metric when looking at percentage differences - it only takes an unlucky run or two to add on 200 damage taken with one configuration. See also: my tests with Rico where, despite autoskill doing better by damage contribution/clear time almost across the board, there were no significant differences in the team's damage taken.
Extra: if you'd like to see what test results look like when there certainly is no difference (and compare those to this), check out this page where I compare G41 to G41. You may notice the charts look quite similar to these.
Echelon position#
It's possible to handwave all this and just say "test things out for your specific fight." You could also argue that all skills should always be on manual, because there are some cases where you may want to slightly delay some activation and manual skills lets you perfectly control those. Those aren't terribly useful recommendations even if technically accurate, so let's see if there are any general trends and ideas to take away.
As for the experiment. Step 1: find a rifle that has identical tiles so we don't have that as a confounding variable when we swap their places. Options are Gepard M1, SRS, TAC-50, Mk 12, M82A1, KSVK, Falcon, M200, and Zas M76. I really don't care about a 3% difference in RF tiles, so I'll be pairing her with SRS for these fights. Mk 12 could work but she restricts HG choice a bit more as a half ROF selfbuffer.
Step 2: take a bunch of fights that seem at least moderately relevant to lategame content and see what happens.
- Pos7 may be better in both cases, primarily from clear times. The other metrics could be consistent with this, but not by enough to be meaningful.
To avoid unfair comparisons against melee units, I'll switch out Stechkin for Px4 Storm in the next tests. With a Calico/P22 frontline, melee units like Zombies and Guards could walk up to Type88 on 7 and make her switch to melee mode. This wasn't a problem against the Scouts/Strelets, but came up enough against the next few to be worth changing.
- Pos1 may be better against the Guard/Brute and Prowler/Vespid groups
- Performance looks similar against the zombie horde
For another non-standard comp, I'll use Type88/Mk 12/P22/Five-Seven/Px4 Storm for the next set. This makes CDR tiles exactly equal, covers Type88 from both positions, and Mk 12's partial ROF buff will keep their shots far apart.
- Pos7 seems to have a significant advantage against the Guard/Striker group
- Pos1 may be better against the Guard/Jaeger and Scout/Brutes
- It looks very equal against the Brutes
On average, having Type88 on position 7 resulted in 0.8% faster clears, 1.7% more damage taken, and 0.02% higher damage contribution. These numbers don't mean much and only really tell us that there is not some massive skew in favor of one position. In terms of absolute differences, performance on average varied by 5.8% (clear time)/17.6% (damage taken)/10.5% (damage contribution). So it'd be reasonable to expect an oath (+10% FP/acc at mod3) to contribute more to Type88's performance than her echelon position.
Conclusion#
As far as I can see, concerns about Type88's skins are much ado about nothing. Use whichever one you like best.
Much more interesting is her formation placement. From the test results, I'd put the score somewhere about 4-3-2 (favoring pos1-pos7-equal). 9 enemies hardly represent the whole of lategame content so take that with a grain of salt. Unfortunately, I don't see enough evidence to make a general recommendation on where she should be placed in an echelon. As there do appear to be real performance differences, this is something you have to test on a case-by-case basis when dealing with particularly high-requirement battles. If it's not critical to optimize the fight, I'd go with whatever gives better tile coverage. The differences are ultimately on the order of 5-10% and are not likely to make or break a fight.
Some possible topics for further investigation:
- Putting Type88 on position 4. This gets a bit harder to test when trying to keep tiles equal between positions.
- Intentionally triggering Type88's melee mode. This gets more into the range of active in-battle controls, making it much harder to analyze. All I know right now is it was handy on one of the triple Typhon deathstacks in PL.
- Collecting even more data on Type88's positioning would be useful. It currently appears to be a somewhat even split but this won't necessarily hold for all relevant enemies.
- Kord positioning. Her piercing acts similarly, but MGSG echelons use very different formations and go against different types of enemies.
More like this
Comparisons for general use
(the talent)