A double attack experiment..Part3 (How much auto% is needed)
Apr 11, 2021 8:56:45 GMT
Lurker K, Kazu, and 7 more like this
Post by sstraycat on Apr 11, 2021 8:56:45 GMT
DOUBLE ATTACK EXPERIMENT PART III
Good day, I'm writing this post to both correct and support some points regarding the first two double attack experiments (better late than never i guess xD... also I separated this post to avoid clutter). You might have already heard about how Fairy Wind reduces the autoskill (90-95%?). Was that true? We might have even heard the same stuff. Yet I was shocked to find out from this experiment's actuals how it really works.
In this Experiment Part III, I'd like to:
1. Correct some points from the first 2 experiments
2. Find out how DA% is computed and determine actually how Fairy Wind's Autoskill decrease works, if possible
3. Determine Auto% needed for perfect/100% DA rate (skip to Conclusion for this if u want)
First, I'd like to point out what were wrong about the previous two experiments:
1. Experiment 1 incorrectly assumed a 40% activation @256 CRT, just because it was stated in a wiki. It was wrong to have taken for fact the wiki info about a plateau, and how it computes DA rate using CRT. The wiki post was last modified in 2012, before the rebalance. As such, If another rebalance occurs in the future, then contents of this post might also become obsolete.
2. There was no account for Fairy Wind's Autoskill decrease on both previous experiments.
Funny thing is that even with the bad points above, there will be some form of consistency to be found on the data later by the end of this post.
You'll see that the data taken from these two would still prove significant.
For now, next is to review what the first two experiments brought to light.
1. Experiment 1 disproved some % activation rumors.
It also lead to a question if Autoskill's a contributing factor to DA (a VERY laughable idea at the time), which is why Experiment 2 was created.
A side result was also observed --- the 2.6 multiplier for Hidden Arrow.
2. Experiment 2 showed that Autoskill IS a factor for DA, despite having no available info about that fact at the time.
Why the need for this Part III then? Well in Experiment 2, I got stuck with +30% autoskill (was poor ).
Also left wondering whether perfect or 100% DA is possible.
Only to find out later that yes, 100% double attack is possible (yyeeeeaaaaah!!), and it has its benefits:
- You'd get the best possible DPS
- No more recoil damage
- Your concentration % will not decrease.
So how much Autoskill do we need to put to get 100% DA? If you follow Experiment 2's pattern, it's about 60-67%, but remember that there were wrong/obsolete points on that experiment... so here we are.
This time, I started from the very bottom, CRT=1. The process is the same, each test a total of 1000 attacks, divided between 10 sets of 100 attacks each.
Having 100m training luna this time REALLLLLY helped a lot, else data would've taken weeks/months to gather. Below is the set used:
Weap: Mirror Bow SR ◇Iniliba ◇Sauro II ☆Rate Cut2
Arm: Licht ◇Iniliba ☆Rate Cut3
Add: Sieger ◇Iniliba ☆Rate Cut5
Acc: ☆MagDef5
Spec: NOM ◇Iniliba ☆Rate Cut4
Apos: MaltiG, VulschemG, Ignatz, VatendethG, IraiG. 8G lines
Relic: Horutes, Cottonello,12 x melee+1%
Those were selected just because they have no autoskill bonus. As for equips with auto bonus:
Add: Lava Guard LU ◇Sakimori ◇Sakimori ☆MagDef4 -> gives +9% Auto
Add: Headband NoUP ☆MpCost5 -> gives +5% Auto
Acc: ☆Auto5 -> gives +5% Auto
Please see tables below. Sorry for the jumbled table letters. Took these raw data without regards to any order.
Now I had to label them by manner in which the raw data can be best explained.
The numbers presented here, like in the previous experiments, represent (number of DA procs) / (total attacks)
B A D C
CRT=1, 0% Auto CRT=1, 0% Auto CRT=1, 9% Auto CRT=1, 10% Auto
with Fairy Wind No FW with FW with FW
1 00 / 100 09 / 100 00 / 100 01 / 100
2 00 / 100 09 / 100 00 / 100 00 / 100
3 00 / 100 11 / 100 00 / 100 03 / 100
4 00 / 100 11 / 100 00 / 100 01 / 100
5 00 / 100 10 / 100 00 / 100 02 / 100
6 00 / 100 07 / 100 00 / 100 00 / 100
7 00 / 100 09 / 100 00 / 100 00 / 100
8 00 / 100 08 / 100 00 / 100 02 / 100
9 00 / 100 07 / 100 00 / 100 00 / 100
10 00 / 100 14 / 100 00 / 100 01 / 100
----------------------------------------------------------------------------
Totals : 000 / 1000 095 / 1000 000 / 1000 010 / 1000
E F G
CRT=8, 9% Auto CRT=72, 0% Auto CRT=72, 9% Auto
with Fairy Wind with FW with FW
1 01 / 100 00 / 100 10 / 100
2 02 / 100 00 / 100 06 / 100
3 00 / 100 00 / 100 12 / 100
4 00 / 100 00 / 100 10 / 100
5 01 / 100 00 / 100 07 / 100
6 00 / 100 00 / 100 09 / 100
7 01 / 100 00 / 100 11 / 100
8 01 / 100 00 / 100 08 / 100
9 01 / 100 00 / 100 10 / 100
10 01 / 100 00 / 100 04 / 100
----------------------------------------------------------------------------
Totals : 008 / 1000 000 / 1000 087 / 1000
I H J K
CRT=80, 0% Auto CRT=80, 9% auto CRT=80, 10% Auto CRT=256, 0% Auto
with Fairy Wind with FW with FW with FW
1 02 / 100 14 / 100 11 / 100 20 / 100
2 00 / 100 10 / 100 17 / 100 23 / 100
3 00 / 100 09 / 100 11 / 100 25 / 100
4 00 / 100 13 / 100 07 / 100 20 / 100
5 00 / 100 02 / 100 10 / 100 20 / 100
6 02 / 100 09 / 100 14 / 100 23 / 100
7 00 / 100 08 / 100 13 / 100 24 / 100
8 02 / 100 09 / 100 12 / 100 20 / 100
9 01 / 100 12 / 100 06 / 100 22 / 100
10 02 / 100 11 / 100 09 / 100 20 / 100
----------------------------------------------------------------------------
Totals : 009 / 1000 097 / 1000 110 / 1000 217 / 1000
Looks confusing, but to interpret these data:
A. @crt=1, 0% Auto, No FW, got 095 / 1000 actual => 9.5%
At 1 CRT, the base proc rate for double attack is believed to be 10%.
Got 9.5% actual result. Really close so this ties up with that 10%, provided Fairy Wind is not activated.
B. @crt=1, 0% Auto, with FW, got 000 / 1000 actual => 0%
At 1 CRT, the base DA proc rate 10%, rate becomes 0% if Fairy Wind is activated (really surprised by this coz I was expecting about a 90-95% cut).
At this point that myth had been busted.
C. @crt=1, 10% Auto with FW, got 010 / 1000 actual => 1%
At 1 CRT, with 10% Auto, if you activate Fairy Wind, you'd get 1% chance to proc DA
Hold up... base 10% plus Auto 10%? And FW is activated but it still procs 1% DA??? Go to the next pls...
D. @crt=1, 9% Auto with FW, got 000 / 1000 actual => 0%
At 1 CRT, with 9% Auto instead of 10%, if you activate Fairy Wind, you'd get 0% chance to proc DA
What could C and D tell us? There is always the base 10% chance to proc DA without FW.
But if you use FW, you'd need 10% auto to at least proc 1% DA.
Does this mean FW takes away 19% ? Too early to outright conclude that here so some more tables up for tests...
E. @crt=8, 9% Auto with FW, got 008 / 1000 actual => 0.8%
Using 10% base with 9% Auto, makes 19%. FW supposedly takes away 19%? That takes it down to 0% DA proc rate from item D above.
But here I move CRT. From CRT 1, I incrementally added 7 stats to get to CRT 8 which produced an almost 1% result for DA proc rate.
Would this be 1% DA proc per 8 CRT? And that 19% held up well...
F. @crt=72, 0% Auto with FW, got 000 / 1000 actual => 0%
If its 1% per 8 CRT, maybe take up to 72 CRT? Supposedly 9% proc. No Auto this time. Add the 10% base and proc becomes 19% supposedly.
Cast FW and see the 19% proc rate... go down to 0.
Items C & D WERE telling us something.
G. @crt=72, 9% Auto with FW, got 087 / 1000 actual => 8.7%
Same as in item F above (base 10% + 9%) but this time, add 9% auto. FW takes away 19%, leaving 9%.
Actual result I got was 8.7%, very close to 9%.
H. @crt=80, 9% Auto with FW, got 097 / 1000 actual => 9.7%
Same as item G above, the only difference is 8 more CRT. If 1% per 8 CRT, thats 1% more activation.
The actual proc rate for this item is 9.7% compared to previous 8.7%. 1% more xD
I. @crt=80, 0% Auto with FW, got 009 / 1000 actual => 0.9%
See Item F above. FW took away 19% activation leaving it 0%. CRT = 72 back then, we have 80 CRT here.
From the 8 added CRT stats, we got 1% more activation.
The actual result from this item is 0.9%, very close to 1%. That 19% taken by FW just became a pattern.
J. @crt=80, 10% Auto with FW, got 110 / 1000 actual => 11%
From item I above, increase auto from 0 to 10%, we should also get 10% more activation.
What happened to the actuals? From 0.9% to 11% is 10.1%.
K. @crt=256, 0% Auto with FW, got 217 / 1000 actual => 21.7%
Now this last test was the one I failed to test in Experiment 1. CRT=256.
As you can see actual DA rate @256 CRT is at 21.7%, very far from the expected 40% DA rate.
Result value on that test only got past 40% proc because of my CRT back then.
Speaking of DA rate, what I got from doing these tests was the formula
DA rate = 10% (base) + CRT/8 + Auto%
If fairy wind is activated, that becomes
DA rate = 10% (base) + CRT/8 + Auto% - 19%
If you test the item K above using this formula, DA rate = 10% base + 256/8 + 0% auto - 19% = 23%.
The (computed) 23% compared to 21.7% (actual) is close.
So its like 230 procs and 217 procs both out of 1000 attacks.
I know that the values have mostly just been close so far but a larger sample size should get these values closer.
Now we got a somewhat workable formula, I think? Would like to test this formula versus the data taken two years ago.
From Double Attack Experiment Parts 1 and 2 (arranged by increasing Auto):
Test Set1 - Total 447 double attack procs / 1000 attacks (about 4 procs in 10 attacks), used +0% auto
Test Set3 - Total 565 double attack procs / 1000 attacks (about 5 procs in 10 attacks), used +10% auto
Test Set4 - Total 653 double attack procs / 1000 attacks (about 6 procs in 10 attacks), used +20% auto
Test Set2 - Total 766 double attack procs / 1000 attacks (about 7 procs in 10 attacks), used +30% auto
Checked my CRT back then and it was 439.
Now the fun part, using the past data's auto% we can get expected rates....
DA rate = base + CRT/8 + Auto% - FW
DA rate = 10% + CRT/8 + Auto% - 19%
Test Set1 DA rate = 10% + 439/8 + 0% - 19% = 45. 875%
Test Set3 DA rate = 10% + 439/8 + 10% - 19% = 55. 875%
Test Set4 DA rate = 10% + 439/8 + 20% - 19% = 65. 875%
Test Set2 DA rate = 10% + 439/8 + 30% - 19% = 75. 875%
Comparing these derived values and the actuals from two years ago:
Test Set1 - computed 45. 875% (ie. 458/1000 rounded down) vs. the actual 447 procs / 1000 attacks
Test Set3 - computed 55. 875% (ie. 558/1000 rounded down) vs. the actual 565 procs / 1000 attacks
Test Set4 - computed 65. 875% (ie. 658/1000 rounded down) vs. the actual 653 procs / 1000 attacks
Test Set2 - computed 75. 875% (ie. 758/1000 rounded down) vs. the actual 766 procs / 1000 attacks
Each of which seems to be very close (o.0) Freaky...
CONCLUSIONS:
1. The test tables above show that -19% auto is caused by FW.
Idk really looks rather.. odd? Being -20% seems more like it? Or -20%'s looking more like a safe bet.
I'd like to say -20 but that's just my bias idk coz I didn't expect to get such an odd value.
Cannot deny -19% was what came out of these tests starting from Table C, and so it was used in the formula.
2. As per the test data above, the formula for double attack rate while using Fairy Wind is
DA rate = (base) + (CRT/8) + (auto% from all gears) - (auto% deducted by FW)
DA rate = 10% + CRT/8 + Auto% - 19%
3. As it is the auto needed to get perfect/100% DA is... (If Lv420 char can get 466 CRT)
DA = base + CRT/8 + Auto% - FW
(100% DA here)
100% = 10% + 466/8 + Auto% -19%
Auto% = 100% - 10% - 466/8 + 19%
Auto% = 50.75%
With lower max CRT, the more Auto% you'd need to get 100% DA.
(Edited Dec 2021) POST NOTES:
1. Idk if that 50.75% needs to be rounded somewhere along the way (yet we're all betting 51%'s safe yes?).
I am also unfamiliar exactly how much Auto% does purple line give, given values for RL and BL.
That was why the tests above used 8G lines to get rid of that unknown.
2. Wandering Clothes (previously known to us as Exile Garb or Outflow Clothes) has been released late November.
This enables STR users (be it normal or spec bow) to use less Auto%.
Eg. If your STR is 464 it would be like an equivalent of +29% chance to get double attacks.
Half of 464 is 232 and that is converted to CRT. Double attack chance procs have CRT/8 in the formula so 232/8 is 29.
Another example, if your STR is 480, then 480/2 is 240 converted to CRT. CRT/8 or 240/8 is 30, giving you free 30% DA proc.
So If you have STR in your build, instead of using your resources to get 29-30% auto
(which does the same job of boosting double attack chance and is expensive), just get Wandering.
3. Equipping something that decreases autoskill, like Slebinian Badge, will eat up the DA proc rate provided by CRT and even erase the base.
Try to equip that item and see your DA procs go away. Autoskill debuffs would do the same.
On the contrary, an auto buff like the one from Ench's Draw skill increases DA rate.
4. It was fun that even data from two years ago came into place. If a rebalance happens again, would all this be changed?
Good day, I'm writing this post to both correct and support some points regarding the first two double attack experiments (better late than never i guess xD... also I separated this post to avoid clutter). You might have already heard about how Fairy Wind reduces the autoskill (90-95%?). Was that true? We might have even heard the same stuff. Yet I was shocked to find out from this experiment's actuals how it really works.
In this Experiment Part III, I'd like to:
1. Correct some points from the first 2 experiments
2. Find out how DA% is computed and determine actually how Fairy Wind's Autoskill decrease works, if possible
3. Determine Auto% needed for perfect/100% DA rate (skip to Conclusion for this if u want)
First, I'd like to point out what were wrong about the previous two experiments:
1. Experiment 1 incorrectly assumed a 40% activation @256 CRT, just because it was stated in a wiki. It was wrong to have taken for fact the wiki info about a plateau, and how it computes DA rate using CRT. The wiki post was last modified in 2012, before the rebalance. As such, If another rebalance occurs in the future, then contents of this post might also become obsolete.
2. There was no account for Fairy Wind's Autoskill decrease on both previous experiments.
Funny thing is that even with the bad points above, there will be some form of consistency to be found on the data later by the end of this post.
You'll see that the data taken from these two would still prove significant.
For now, next is to review what the first two experiments brought to light.
1. Experiment 1 disproved some % activation rumors.
It also lead to a question if Autoskill's a contributing factor to DA (a VERY laughable idea at the time), which is why Experiment 2 was created.
A side result was also observed --- the 2.6 multiplier for Hidden Arrow.
2. Experiment 2 showed that Autoskill IS a factor for DA, despite having no available info about that fact at the time.
Why the need for this Part III then? Well in Experiment 2, I got stuck with +30% autoskill (was poor ).
Also left wondering whether perfect or 100% DA is possible.
Only to find out later that yes, 100% double attack is possible (yyeeeeaaaaah!!), and it has its benefits:
- You'd get the best possible DPS
- No more recoil damage
- Your concentration % will not decrease.
So how much Autoskill do we need to put to get 100% DA? If you follow Experiment 2's pattern, it's about 60-67%, but remember that there were wrong/obsolete points on that experiment... so here we are.
This time, I started from the very bottom, CRT=1. The process is the same, each test a total of 1000 attacks, divided between 10 sets of 100 attacks each.
Having 100m training luna this time REALLLLLY helped a lot, else data would've taken weeks/months to gather. Below is the set used:
Weap: Mirror Bow SR ◇Iniliba ◇Sauro II ☆Rate Cut2
Arm: Licht ◇Iniliba ☆Rate Cut3
Add: Sieger ◇Iniliba ☆Rate Cut5
Acc: ☆MagDef5
Spec: NOM ◇Iniliba ☆Rate Cut4
Apos: MaltiG, VulschemG, Ignatz, VatendethG, IraiG. 8G lines
Relic: Horutes, Cottonello,12 x melee+1%
Those were selected just because they have no autoskill bonus. As for equips with auto bonus:
Add: Lava Guard LU ◇Sakimori ◇Sakimori ☆MagDef4 -> gives +9% Auto
Add: Headband NoUP ☆MpCost5 -> gives +5% Auto
Acc: ☆Auto5 -> gives +5% Auto
Please see tables below. Sorry for the jumbled table letters. Took these raw data without regards to any order.
Now I had to label them by manner in which the raw data can be best explained.
The numbers presented here, like in the previous experiments, represent (number of DA procs) / (total attacks)
B A D C
CRT=1, 0% Auto CRT=1, 0% Auto CRT=1, 9% Auto CRT=1, 10% Auto
with Fairy Wind No FW with FW with FW
1 00 / 100 09 / 100 00 / 100 01 / 100
2 00 / 100 09 / 100 00 / 100 00 / 100
3 00 / 100 11 / 100 00 / 100 03 / 100
4 00 / 100 11 / 100 00 / 100 01 / 100
5 00 / 100 10 / 100 00 / 100 02 / 100
6 00 / 100 07 / 100 00 / 100 00 / 100
7 00 / 100 09 / 100 00 / 100 00 / 100
8 00 / 100 08 / 100 00 / 100 02 / 100
9 00 / 100 07 / 100 00 / 100 00 / 100
10 00 / 100 14 / 100 00 / 100 01 / 100
----------------------------------------------------------------------------
Totals : 000 / 1000 095 / 1000 000 / 1000 010 / 1000
E F G
CRT=8, 9% Auto CRT=72, 0% Auto CRT=72, 9% Auto
with Fairy Wind with FW with FW
1 01 / 100 00 / 100 10 / 100
2 02 / 100 00 / 100 06 / 100
3 00 / 100 00 / 100 12 / 100
4 00 / 100 00 / 100 10 / 100
5 01 / 100 00 / 100 07 / 100
6 00 / 100 00 / 100 09 / 100
7 01 / 100 00 / 100 11 / 100
8 01 / 100 00 / 100 08 / 100
9 01 / 100 00 / 100 10 / 100
10 01 / 100 00 / 100 04 / 100
----------------------------------------------------------------------------
Totals : 008 / 1000 000 / 1000 087 / 1000
I H J K
CRT=80, 0% Auto CRT=80, 9% auto CRT=80, 10% Auto CRT=256, 0% Auto
with Fairy Wind with FW with FW with FW
1 02 / 100 14 / 100 11 / 100 20 / 100
2 00 / 100 10 / 100 17 / 100 23 / 100
3 00 / 100 09 / 100 11 / 100 25 / 100
4 00 / 100 13 / 100 07 / 100 20 / 100
5 00 / 100 02 / 100 10 / 100 20 / 100
6 02 / 100 09 / 100 14 / 100 23 / 100
7 00 / 100 08 / 100 13 / 100 24 / 100
8 02 / 100 09 / 100 12 / 100 20 / 100
9 01 / 100 12 / 100 06 / 100 22 / 100
10 02 / 100 11 / 100 09 / 100 20 / 100
----------------------------------------------------------------------------
Totals : 009 / 1000 097 / 1000 110 / 1000 217 / 1000
Looks confusing, but to interpret these data:
A. @crt=1, 0% Auto, No FW, got 095 / 1000 actual => 9.5%
At 1 CRT, the base proc rate for double attack is believed to be 10%.
Got 9.5% actual result. Really close so this ties up with that 10%, provided Fairy Wind is not activated.
B. @crt=1, 0% Auto, with FW, got 000 / 1000 actual => 0%
At 1 CRT, the base DA proc rate 10%, rate becomes 0% if Fairy Wind is activated (really surprised by this coz I was expecting about a 90-95% cut).
At this point that myth had been busted.
C. @crt=1, 10% Auto with FW, got 010 / 1000 actual => 1%
At 1 CRT, with 10% Auto, if you activate Fairy Wind, you'd get 1% chance to proc DA
Hold up... base 10% plus Auto 10%? And FW is activated but it still procs 1% DA??? Go to the next pls...
D. @crt=1, 9% Auto with FW, got 000 / 1000 actual => 0%
At 1 CRT, with 9% Auto instead of 10%, if you activate Fairy Wind, you'd get 0% chance to proc DA
What could C and D tell us? There is always the base 10% chance to proc DA without FW.
But if you use FW, you'd need 10% auto to at least proc 1% DA.
Does this mean FW takes away 19% ? Too early to outright conclude that here so some more tables up for tests...
E. @crt=8, 9% Auto with FW, got 008 / 1000 actual => 0.8%
Using 10% base with 9% Auto, makes 19%. FW supposedly takes away 19%? That takes it down to 0% DA proc rate from item D above.
But here I move CRT. From CRT 1, I incrementally added 7 stats to get to CRT 8 which produced an almost 1% result for DA proc rate.
Would this be 1% DA proc per 8 CRT? And that 19% held up well...
F. @crt=72, 0% Auto with FW, got 000 / 1000 actual => 0%
If its 1% per 8 CRT, maybe take up to 72 CRT? Supposedly 9% proc. No Auto this time. Add the 10% base and proc becomes 19% supposedly.
Cast FW and see the 19% proc rate... go down to 0.
Items C & D WERE telling us something.
G. @crt=72, 9% Auto with FW, got 087 / 1000 actual => 8.7%
Same as in item F above (base 10% + 9%) but this time, add 9% auto. FW takes away 19%, leaving 9%.
Actual result I got was 8.7%, very close to 9%.
H. @crt=80, 9% Auto with FW, got 097 / 1000 actual => 9.7%
Same as item G above, the only difference is 8 more CRT. If 1% per 8 CRT, thats 1% more activation.
The actual proc rate for this item is 9.7% compared to previous 8.7%. 1% more xD
I. @crt=80, 0% Auto with FW, got 009 / 1000 actual => 0.9%
See Item F above. FW took away 19% activation leaving it 0%. CRT = 72 back then, we have 80 CRT here.
From the 8 added CRT stats, we got 1% more activation.
The actual result from this item is 0.9%, very close to 1%. That 19% taken by FW just became a pattern.
J. @crt=80, 10% Auto with FW, got 110 / 1000 actual => 11%
From item I above, increase auto from 0 to 10%, we should also get 10% more activation.
What happened to the actuals? From 0.9% to 11% is 10.1%.
K. @crt=256, 0% Auto with FW, got 217 / 1000 actual => 21.7%
Now this last test was the one I failed to test in Experiment 1. CRT=256.
As you can see actual DA rate @256 CRT is at 21.7%, very far from the expected 40% DA rate.
Result value on that test only got past 40% proc because of my CRT back then.
Speaking of DA rate, what I got from doing these tests was the formula
DA rate = 10% (base) + CRT/8 + Auto%
If fairy wind is activated, that becomes
DA rate = 10% (base) + CRT/8 + Auto% - 19%
If you test the item K above using this formula, DA rate = 10% base + 256/8 + 0% auto - 19% = 23%.
The (computed) 23% compared to 21.7% (actual) is close.
So its like 230 procs and 217 procs both out of 1000 attacks.
I know that the values have mostly just been close so far but a larger sample size should get these values closer.
Now we got a somewhat workable formula, I think? Would like to test this formula versus the data taken two years ago.
From Double Attack Experiment Parts 1 and 2 (arranged by increasing Auto):
Test Set1 - Total 447 double attack procs / 1000 attacks (about 4 procs in 10 attacks), used +0% auto
Test Set3 - Total 565 double attack procs / 1000 attacks (about 5 procs in 10 attacks), used +10% auto
Test Set4 - Total 653 double attack procs / 1000 attacks (about 6 procs in 10 attacks), used +20% auto
Test Set2 - Total 766 double attack procs / 1000 attacks (about 7 procs in 10 attacks), used +30% auto
Checked my CRT back then and it was 439.
Now the fun part, using the past data's auto% we can get expected rates....
DA rate = base + CRT/8 + Auto% - FW
DA rate = 10% + CRT/8 + Auto% - 19%
Test Set1 DA rate = 10% + 439/8 + 0% - 19% = 45. 875%
Test Set3 DA rate = 10% + 439/8 + 10% - 19% = 55. 875%
Test Set4 DA rate = 10% + 439/8 + 20% - 19% = 65. 875%
Test Set2 DA rate = 10% + 439/8 + 30% - 19% = 75. 875%
Comparing these derived values and the actuals from two years ago:
Test Set1 - computed 45. 875% (ie. 458/1000 rounded down) vs. the actual 447 procs / 1000 attacks
Test Set3 - computed 55. 875% (ie. 558/1000 rounded down) vs. the actual 565 procs / 1000 attacks
Test Set4 - computed 65. 875% (ie. 658/1000 rounded down) vs. the actual 653 procs / 1000 attacks
Test Set2 - computed 75. 875% (ie. 758/1000 rounded down) vs. the actual 766 procs / 1000 attacks
Each of which seems to be very close (o.0) Freaky...
CONCLUSIONS:
1. The test tables above show that -19% auto is caused by FW.
Idk really looks rather.. odd? Being -20% seems more like it? Or -20%'s looking more like a safe bet.
I'd like to say -20 but that's just my bias idk coz I didn't expect to get such an odd value.
Cannot deny -19% was what came out of these tests starting from Table C, and so it was used in the formula.
2. As per the test data above, the formula for double attack rate while using Fairy Wind is
DA rate = (base) + (CRT/8) + (auto% from all gears) - (auto% deducted by FW)
DA rate = 10% + CRT/8 + Auto% - 19%
3. As it is the auto needed to get perfect/100% DA is... (If Lv420 char can get 466 CRT)
DA = base + CRT/8 + Auto% - FW
(100% DA here)
100% = 10% + 466/8 + Auto% -19%
Auto% = 100% - 10% - 466/8 + 19%
Auto% = 50.75%
With lower max CRT, the more Auto% you'd need to get 100% DA.
(Edited Dec 2021) POST NOTES:
1. Idk if that 50.75% needs to be rounded somewhere along the way (yet we're all betting 51%'s safe yes?).
I am also unfamiliar exactly how much Auto% does purple line give, given values for RL and BL.
That was why the tests above used 8G lines to get rid of that unknown.
2. Wandering Clothes (previously known to us as Exile Garb or Outflow Clothes) has been released late November.
This enables STR users (be it normal or spec bow) to use less Auto%.
Eg. If your STR is 464 it would be like an equivalent of +29% chance to get double attacks.
Half of 464 is 232 and that is converted to CRT. Double attack chance procs have CRT/8 in the formula so 232/8 is 29.
Another example, if your STR is 480, then 480/2 is 240 converted to CRT. CRT/8 or 240/8 is 30, giving you free 30% DA proc.
So If you have STR in your build, instead of using your resources to get 29-30% auto
(which does the same job of boosting double attack chance and is expensive), just get Wandering.
3. Equipping something that decreases autoskill, like Slebinian Badge, will eat up the DA proc rate provided by CRT and even erase the base.
Try to equip that item and see your DA procs go away. Autoskill debuffs would do the same.
On the contrary, an auto buff like the one from Ench's Draw skill increases DA rate.
4. It was fun that even data from two years ago came into place. If a rebalance happens again, would all this be changed?