#Fleet build
1 messages ยท Page 1 of 1 (latest)
w15 fleet
lets assume i have all the ships but forgot to lvl them
cause i have all UR
Mob: Musashi, Unicorn, Aquila, Laffey 2, San Diego, Shimanto
Boss: Alsace, Shinano, Amagi, Guam, Jeanne, Plymouth
Support: Implacable, Illustrious, Independence
a few ways to optimize it if you really wanted to but it should be fine
the support fleet can have cat and fritz
they are giga cracked
For AA equipment you start out with:
Roombas on Backline (twin 113mm)
Sexybofors on your AA carry (Sextuple 40mm)
Pastabofors on the rest of your vanguard (twin 90mm, pr4)
If you were to craft a rainbow 76mm gun, you'd replace the sexybofors first
ALO GPH specifcally reccomends STAAGs on main fleet first, though
everything else is fairly standard
ig torp buldges aren't too standard, but they help a lot in w15
gearing shouldn't be an issue for someone who has implac
right 
can even make 6 ur aa
2 years of opsi is probably enough for like 14-18 UR equips if they did every month
and i doubt they did every month
i also did opsi last year because i started last year 
same lol
but i can guarantee u it will be way easier than my attempt
since i did a speedrun for it
can u show how many designs do u have ?
waht if some ships are underleveld like alsace and shimanto
what design
the blueprints?
leveling up ships shouldn't be a question for someone who is going to do w15 
yeah they are called designs
in the depot first tab
you get them to 120 minimum before you even try
๐ฉ
you could consider using hte new asw plane
i forgot how much it costs though
wdym by that? w15 has subs?
yes
although i didn't use the cheat code
doing boss mob boss with 2 mountains made guam sustain at 0.1% hp laffey at 5% off tank and shimanto at 10% in the middle lvl 125 
fyi the largest jump in power is from +12 to +13 so you don't want to split gold plates
except for auxes, sometimes
these are the gold designs
what about the ur ones ?
he's good for UR ones
he has 2 wvyerns and a 100mm
im pretty sure with what he screenshotted he could copy cheeky's screenshot exactly
although if you have eldridge i don't see the use in Jeanne
iff you have her retrofit
well i could, but isee some equipments that i can craft at gearlab
fuck
yeah it is hard for me to compare anyone's performance because 1 i didn't use the cheat code and 2 my fps were under 2 in the giga cv node
for the reccord, Eldridge should be Laffey 2 there
eldridge i dont know why i skipped her retro thing
forgot to change her out
biggets blunder
is that boss mob boss
?
in theory it should be able to go either way
For support fleet I've found the most success mixing flapjacks and either purple fws or gold fws in this pattern:
Slowest cv: all fws
Middle cv: 1 fw slot, 1 flap
Fastest cv: all flaps
it's kinda arbitrary though
wdym fastest cv? in reload?
yeah
you just want to desync the different carriers so that the planes that do slip by have some of their hp taken out
suchidata says that method is terrible
just full skyrocket or full hellcat
full hellcat if you have more wyverns
suchidata had issues, and this isnt cn method
wait what about the new one 
10 trials per with the box plots doesn't give a statistically significant results
ill just go all flapjacks then
that's what most people do tbh
wait why not just use my other wyvern to unicorn?
if he has enough he can use i think
he is in electronic poverty
well you need some meat on the boss fleet
oooh ok ok
breguet also is very strong 
suchidata tested 20 different setups with 10 trials per setup.
It essentially was unintentional p-hacking
i think it is safe to use the new plane even with the bugged speed 
I haven't seen anything definitely placing it better than purple fw atm tbh
anyways essentially by testing 20 different setups he essentially just slashed the confidence of the results by like a factor of 10-20
what
i have a lot of twin 113, can i replace the twin 90mm?
preferably not
imo what made the new fighter promising was the announced flight speed but after discovering that it was bugged i think i still heard that the new one had the advantage 
you'll have a lot of range but also a pretty long reload time if you use Roombas on everyone else
I'm not even gonna ask
hmmmm
If you wanted to apply error correction to the confidence interval for a setup testing 20 different intercept fleets with the a HB correction you'd divide the smallest p-value by 20, second smallest by 19, etc..
it basically makes the whole test useless tbh
I'm gonna be honest, what you just said was complete gibberish to me
statistics 
still doesn't exactly make the tests invalid or useless
basically there was a chocolate bar study in like 2014 or smth where they concluded choclate was healthy with a p<.05 for some metric
but they studied like 18 different metrics
same principle
uhhh
I'm not quite sure I follow but all the tests are comparing the same results with varying set ups
when it comes to support fleet interceptions what is or are the random events that might interfere with the data
? i think based on this the tester might put a suitable number of the total tests
you're not comparing against different factors here
yes, but you still would use the same error correction
ok for the record, my knowledge of statistics, data collection, etc is basically non existent
anyways suchi's tests were honestly fine but you can't use them to make any direct comparisons
between the setups

h u eh
this wasn't about comparing vs the CN results though, not directly anyway
sorry about clogging thread
im leveling up my ships and cameback to this ๐
what i can assure u is this though
if u will use fw with under 2fps u won't do any damage and the enemy planes will hit ur backline in the giga cv node 
i didn't test any other setups personally
i tanked all the planes in the 2 cv nodes
imo musasi might die
the a priori thinking behind a mix is that damage uptime is worthwhile b/c crash damage is equal from 1% remaining hp to 30% so a higher uptime for support gains value
i dont like the cn reqs though
it seemed like the reasoning was shaky
Does this mean eHP3 is the greatest criminal of p-hacking
i assume the number of sims you ran gave more than enough confidence

