[1.3] ► MJOLNOOB ◄ UBER ATZIRI FARMING - TIPS - ANTI REFLECT - RT or CRIT - SOLARIS or SHAV - MF

"
mayanrm wrote:
Some nice gloves to improve the build ;)




hehe gz ;)

"
MatrixFactor wrote:
Hi thanks for writing up the build. I'm trying a cheap crit version. Looking for tips on gear/tree. In particular what do you think about L1 CWDT vs L10-13ish CWDT?

Gear




I use purity in ele weak maps, and herald everywhere else, it's a huge DPS boost.


L90 Witch Tree

I was also doing some math on crit vs RT. With 75% chance to hit I have 0.75 of the procs I would have with RT. But thanks to crit my discharge does 86% more dmg, and my arc does 50% more dmg. It's also true that RT would have more spell dmg in tree.

I also did some simulations on our probability to discharge before we crit & lose charges. With arc 1st it was 70%, and with discharge 1st it was 80%. So the fact that we lose charges on crit seems to not be a big deal.

MATLAB code

"

NTrials=1e5;

disch_prob = 0.5;
molten_crit_prob = 0.065;
arc_prob = 0.5;
arc_crit_prob = 0.24;

n_disch_first=0;

for t=1:NTrials
keep_hammering = 1;
while keep_hammering
r = rand();
if r<molten_crit_prob
keep_hammering = 0;
break;
end
r = rand();
if r < disch_prob
n_disch_first=n_disch_first+1;
keep_hammering=1;
break;
end
r = rand();
if r<arc_prob
r = rand();
if r < arc_crit_prob
keep_hammering = 0;
break;
end
end

end
end

n_disch_first/NTrials



Overall it seems tedious to compare crit dps to RT, do you know if someone's actually done it or was 2x number just a estimate?

Hi,

About CWDT, i'm actually using a gem lvl 6 and i feel better than earlier when i was using a lvl 16 or 20.

About your mjolner, you linked arc first, then discharge. Since you're using crit variant, try discharge first then arc second, you'll end with more DPS.
Check "TPIS/Mjölner's socketed spell order" section.

You're not using Saffel's, your purities aren't level 20, and you choosed not to take "Unnatural Calm" tree point = low max resists. That's why you won't be able yet to do ele reflect or minus max res map mods without pushing ele flasks.
But you took the "Alchemist" nodes. So i suppose you're a player who doesn't mind playing with flasks all the time ;)

You're missing accuracy on helmet. I have 81% chance to hit with 300ish acc on helmet.

About purity of elements, i prefer to always use it because there are some monsters who curse.
Yes, using HoT is a great DPS boost, but i'm already good with the increased DPS of the crit variant.

I compared crit DPS to RT with feelings only, no maths, when clearing hundreds of Atziri maps.
► Shop ◄ Mirror service • Gear • Maps • Gems • Legacy ■ thread/545555
Myself im using lvl2 CwDT and lvl4 IC. It triggers very very easily but still lasts long enough for my playstyle. Ele reflec maps are fun too as the reflected damage i receive triggers it non-stop making me immune to physical dmg (which is convenient as in those maps my effective leech is pretty low). Now this does eat up some endurance charges so something could be said about losing dps over a immortal call setup this low but for me it feels plenty fast.
meginords with elemental weakness on hit would be nice :)

My Gear
Spoiler
Last edited by lkwpeter666 on Nov 20, 2014, 9:39:05 AM
"
lkwpeter666 wrote:
meginords with elemental weakness on hit would be nice :)

My Gear
Spoiler


When rampage finish I'm going to sell this for a decent offer



Can't wait to play this build, 3 hours left!
So I was doing some more math comparing crit vs RT using the L94 crit and L94 RT trees you provided.

Even with 50/50 malis, 81% chance to hit, and 50% spell damage on gear (for both crit & RT) I was getting 110k single target DPS with the crit build (before considering shock, HoT).

On RT with the same 50% spell damage on gear, and the extra 50% spell damage on your RT tree I was getting the same 110k DPS. Without spell damage on gear RT is more DPS. For example, with my gear (140% voidbringer, 0% spell dmg on gear, 75% chance to hit) crit is 77k DPS and RT is 91k DPS.

The main difference is in chance to shock. For each molten strike projectile crit has (with my gear) 32.5% chance to shock whereas RT has only 10% chance to shock (0.5*20% on quality arc). After adding shock into the equation RT still came out on top!!! RT was 121k DPS and crit was 114k DPS. After adding Herald of Thunder (which should help crit) RT was 143k DPS and crit was 139k DPS. (Fun fact arc before discharge made crit DPS 136k but RT DPS 144k, so just as you predicted, nice!)

So shock makes crit comparable to RT. Alternatively getting around 80% chance to hit makes crit superior to RT when you can both shock and use HoT. For example in Atziri we can't shock, so with crit we're gonig to be doing less DPS with crit than RT.

After doing this math I will still play crit for leveling because I think surgeon's flasks are OP. However if I ever switch to farming atziri full time I will likely respec into RT.

Feel free to check my code for mistakes or wrong assumptions I made, and if you have matlab run it for yourself.

Another cool thing I found is that RT actually benefits from power charges more than crit does. Also it's better to have the same number of power/endurance charges than an imbalance. So 4/4 is better than 5/3.

matlab code for the simulation

NTrials=1e6;

global_crit_tree = 0;
spell_crit_tree = 1.95;
spell_crit_gear = 1.4;
global_crit_gear = 0.3;

spell_crit_multi_tree = 1.0;
global_crit_multi_gear = 0.0;

Shock_length = 10; %number of projectiles that hit in 300ms

NPmax = 4;
NEmax = 3;
NFmax = 3;

HoT = 1;

Disch_base_light_avg = (31+586)/2; %Level 21 gems
Disch_base_fire_avg = (219+329)/2;
Disch_base_cold_avg = (180+269)/2;
Arc_base_light = (42+797)/2+181/2*HoT*1.26; %innerforced HoT

crit=0;

if crit ==1 %Crit Section
hit_chance = 0.75;
disch_prob = 0.5;
molten_crit_base = 0.05;
arc_prob = 0.5;
arc_crit_base = 0.05;
disch_crit_base = 0.08;
spell_dmg_tree = 0.78+0.16;
spell_dmg_gear = 0.0;
else %RT Section
hit_chance = 1;
disch_prob = 0.5;
molten_crit_base = 0.0;
arc_prob = 0.5;
arc_crit_base = 0.0;
disch_crit_base = 0;
spell_dmg_tree = 0.5+0.78+0.16;
spell_dmg_gear = 0.0;

end

NPow_avg=0;
NEnd_avg=0;
NFre_avg=0;

NDisch = 0;

NPow=0;
NEnd=0;
NFre=0;

Num_reach_pow_max=0;
Num_reach_end_max=0;

shock_chance_avg=0;
shock = 1;
shock_end = 0;

Dmg_done = 0;

for t=1:NTrials
if t > shock_end
shock = 1;
end
r = rand();
if r<hit_chance %if we miss with this projectile we missed. Is this per projectile or per attack?
r = rand();
if r<molten_crit_base*(1+global_crit_tree+global_crit_gear+0.5*NPow) % the probability of crit with NPow charges
NEnd=min(NPow,NEmax); %on a crit power charges are consumed and voll's converts them to end charges
NPow=0;
else
NPow=min(NPow+1,NPmax);%on a non-crit romira's gives us a power charge
end

r = rand();
if r<arc_prob
r = rand();
shock_chance_avg = shock_chance_avg+arc_crit_base*(1+spell_crit_tree+global_crit_tree+spell_crit_gear+global_crit_gear+0.5*NPow)+0.2;
if r < arc_crit_base*(1+spell_crit_tree+global_crit_tree+spell_crit_gear+global_crit_gear+0.5*NPow)
NEnd=min(NPow,NEmax); %on a crit power charges are consumed and voll's converts them to end charges
NPow=0;%loose all power charges because we crit with arc
Dmg_done = Dmg_done + Arc_base_light* (1+spell_dmg_tree+spell_dmg_gear+0.15*HoT)*(1.5)*(1+spell_crit_multi_tree+global_crit_multi_gear)*shock;%assume arc doesn't chain back
else
NPow=min(NPow+1,NPmax);
Dmg_done = Dmg_done + Arc_base_light* (1+spell_dmg_tree+spell_dmg_gear+0.15*HoT)*shock;
end
r = rand();
if r < arc_crit_base*(1+spell_crit_tree+global_crit_tree+spell_crit_gear+global_crit_gear+0.5*NPow)+0.2
shock = 1.5;
shock_end = t+Shock_length;
end
end

r = rand();
if r < disch_prob
base_dmg = (Disch_base_light_avg*NPow+181/2*1.26*HoT)* (1+spell_dmg_tree+spell_dmg_gear+0.15*HoT)+...
(Disch_base_fire_avg*NEnd + Disch_base_cold_avg*NFre)* (1+spell_dmg_tree+spell_dmg_gear);
crit_chance = disch_crit_base*(1+spell_crit_tree+global_crit_tree+spell_crit_gear+global_crit_gear+0.5*NPow);
avg_disch_dmg = base_dmg*(1-crit_chance)+base_dmg*crit_chance*(1.5)*(1+spell_crit_multi_tree+global_crit_multi_gear);
Dmg_done = Dmg_done+avg_disch_dmg*shock;
NDisch=NDisch+1;
NPow_avg=NPow_avg+NPow;
NEnd_avg=NEnd_avg+NEnd;
NFre_avg=NFre_avg+NFre;
if NPow==NPmax
Num_reach_pow_max=Num_reach_pow_max+1;
end
if NEnd==NEmax
Num_reach_end_max=Num_reach_end_max+1;
end

if NPow>0
shock_chance_avg = shock_chance_avg + crit_chance;
r = rand();
if r< crit_chance
shock = 1.5;
shock_end = t+Shock_length;
end
end

NEnd=min(NPow,NEmax);
NPow=0;
NFre=0; %assume that each time we discharge we kill a monster and get a frenzy charge due to blood rage, this is true for most bosses because of adds spawning
end

end

end

fprintf(1,'Probability of reaching %d PC = %.3f, %d EC = %.3f\n',NPmax,Num_reach_pow_max/NDisch,NEmax,Num_reach_end_max/NDisch);
fprintf(1,'Avg power charges = %f endur= %f frenzy =%f\n',NPow_avg/NDisch,NEnd_avg/NDisch,NFre_avg/NDisch);
fprintf(1,'Chance to shock = %.3f\n',shock_chance_avg/NTrials);

DPS = Dmg_done/(NTrials)*42 %with my APS assuming all 42 molten strike projectiles hit the target (not in terms of accuracy, but in terms of hitbox)

All my builds /view-thread/1430399

T14 'real' clearspeed challenge /1642265
Last edited by MatrixFactor on Nov 20, 2014, 6:47:13 PM
I finally leveled up to 95 with my Scion (RT variant) running about 20 uber maps and counting 3 deaths.
I'll do the same thing with my Witch (Crit variant)... I'll let you know.

About the double flameblast, i died today while pressing rumi+ruby flasks. Best way to fight her is to avoid it if you don't have enough portals remaining.

@MatrixFactor
I don't understand the maths you did, sorry.
As i told you, i feel the DPS. I've done hundreds of normal Atziri maps (maybe 1 thousand now) and maybe like 60-80 uber maps, with the gear i linked in the guide.
This is my opinion (to each his own): theory is good to get a global vision, but practice is better.
That's why i didn't make any DPS calculations in this guide. People can feel the DPS by watching the videos and comparing with their builds.

@mad4thrash
GL with the build. I bet you'll enjoy it ;)
► Shop ◄ Mirror service • Gear • Maps • Gems • Legacy ■ thread/545555
Oh yes I'm enjoying it!

Gear




Tree (lvl 91 Scion)

http://www.pathofexile.com/passive-skill-tree/AAAAAgAAAW8MfQ18DY0OPBB_EVAUcRVQFy8ZLhslHNwdFB_HIvQksCW8Jwsn7Sj6Kjgspi1HLYstqC9vNAo1uTa6Nug3ZjeDOlJCo0ZxRwZJUUuuUEJXyVivWfNca1yKX2pmnmwLbRlul3C7cql4DXvXfNl-WX_GgKSCx4MJg7aESIbRhymLeow2j0aP-pBVkNaTJ5uhn8uio6brpwinK6lurKqxkLQMtDi31riTwBrBxcHzw23GntDQ2RPZW9rd21nb59-_4Yji9-UZ6QLqGOq66-7sGPDV8pz3Mve-99f46_rS

Why I waste some points for lust for carnage notable?

-I don't want to spend thousand chrome on my chest
-I'm lazy as hell and I don't want to spam life flask 24/7
-4 lfgoh seems enough and I can spam molten strike forever in 1vs1 situation

I played only maps for now but when I'll hit 92 I'm going to start some trainig with normal Atziri
"
Funsy wrote:

@MatrixFactor
I don't understand the maths you did, sorry.


The math is a simulation. I simulate what happens for each projectile of molten strike. First I roll chance to hit (75% for crit vs 100% for RT). Second I roll chance this projectile crits and we lose power charges (convert to end charges), or it doesn't crit and we gain +1 power charge. Then I roll chance for discharge to proc, 50%, and roll chance for it to crit. Then I roll chance for arc proc, and chance for it to crit (convert power charges to end with crit build). That's it, I repeat this 1 million times and collect the statistics.

"
Funsy wrote:

This is my opinion (to each his own): theory is good to get a global vision, but practice is better.
That's why i didn't make any DPS calculations in this guide.

I agree that practical intution is important in cases where the math is too hard to do without making unfounded assumptions. This is not the case in a computer game. By definition everything here is discrete and can be computed. In this specific case all we are doing is single target DPS.

I think the reason why you believe crit is double DPS is that you notice the big hits you get with crit, but you fail to notice the absence of smaller hits that add up to help RT DPS.

"
Funsy wrote:

People can feel the DPS by watching the videos and comparing with their builds.

I watched the latest RT Expensive uber run (12min) and the Crit Expensive uber run (17min). Note that the RT run was 5min shorter, but that can be explained by the fact that you skipped some mobs in the RT run. What is convincing for me is that with RT you killed trio in 40sec, and with crit you killed trio in 38sec. Clearly crit doesn't have anywhere near double DPS even in practice.

Anyways I redid my calculations (I forgot about pain attunement before). These numbers are with HoT turned on, and RF turned off. With 4PC/4EC, 50/50 malis, and 80% chance to hit crit gets 208k DPS. With 4PC/4EC, and an extra 72% spell/ele dmg on tree compared to the crit build (taking the last 2 nodes at shadow start) RT gets 212k DPS. With voidbringer, and 75% chance to hit crit gets 186k DPS, significantly less than RT.

So with good gear (+acc helm, legacy malis) crit and RT get equivelnt DPS. Against targets that cannot be shocked RT is 157k DPS and crit is 147k DPS.

So if we go crit we have more difficult itemization (+acc on helm, no good ES gloves).

It seems to me the tradeoff between crit and RT is whether surgeon flasks are worth the more difficult itemization. This is where your practical knowledge comes into play. If crit and RT are the same DPS, what are the tradeoffs, and which should we choose?

--------------

What do you think about using lightning strike to clear the blue mobs at uber? Keep discharge and arc, but just swap in lightning strike instead of molten?

What do you think about replacing arc with concentrated effect for bosses?

+End charge from oak seems good since it increases our DPS and our IC length.
All my builds /view-thread/1430399

T14 'real' clearspeed challenge /1642265
Last edited by MatrixFactor on Nov 21, 2014, 4:03:41 PM
I suppose i'll have to upload another video with crit variant...
I'm know using crit variant to get exp in uber map as i did with RT variant, at least i hope so :).

You compared the worst fight: trio. There's something to know about crit build: it comsumes power charges when crit and results in less endu charges sometimes, no like RT and its constant 3 endu charges up. That's why i always have to be aware of it when fighting trio since i rely on IC during whole fight.
Check out Vaal's fight and see how fast his life bar comes down compared to RT's videos. I don't rely on IC too much during this fight.

About LS, i mentioned it on TIPS/uber map section.

About conc effect, i'll test it on normal atziri first, with RT char, since i want to get exp with crit char. If it's safe using it, i'll test in uber with crit char. I'll let you know in a couple of days.

EDIT
About +1 endu charge, maybe ;). I give this test to others :)
► Shop ◄ Mirror service • Gear • Maps • Gems • Legacy ■ thread/545555
Last edited by Funsy on Nov 21, 2014, 4:23:21 PM
I used trio to compare DPS since it's more consistent: no going underground like vaals and no feed phase like atziri.

Anyways in your latest RT video you kill vaals in 16sec and in your latest crit video in 25sec (since one of them goes underground). Again, you're killing faster with RT, it's impossible that crit has more DPS, much less double DPS.
All my builds /view-thread/1430399

T14 'real' clearspeed challenge /1642265

Report Forum Post

Report Account:

Report Type

Additional Info