AC softcap and low level hardcap discussion

[TytosOfEight]

So if the idea is that a shield, regardless of slot, acts as a shield and pushes AC above the soft cap, the design should be:

Test 1.
Same AC
Non-shield back slot
Shield in secondary vs. no shield
(this should tell you if a shield is pushing above softcap)

Test 2.
Same as above, but with shield in secondary vs. a none-shield AC item in secondary, something like a KDT orb.
(this should tell you if any AC if secondary acts as a “shield” and pushes you above softcap)

Test 3.
Same Ac
Back slot shield (Lodi etc.) vs. non-back slot shield
(this should tell you if items denoted as “shield” work in any lot and it’s not just AC “shield” item in secondary)

Ideally you would keep AC constant across all three tests.

There might be other ways to test it too, but this feels the most straightforward.

[DeathsSilkyMist]

Originally Posted by bcbrown [You must be logged in to view images. Log in or Register.]

I just finished a test against Shiel Glimmerspindel, using a 60 druid. Both sides had 198 worn ac and 61 spell ac for a total of 259. The first side had the lodi shield in the back slot, and the second side had it in the secondary slot. I ended up with 909 hits on the first test and 587 on the second. Back slot had 40 damage/hit, 33.55% min hit, 0.77% max hit. Secondary slot had 39.7 damage/hit, 33.22% min hit, 1.36% max hit.

I was hoping to take 1500-2000 hits per side, but took a peek, saw the hits and intermediate results, and decided to call it there. I'm going to restart the test at slightly lower total AC and compare three treatments: no lodi shield; lodi shield in back slot; lodi shield in secondary slot.

If no-shield and back-slot have the same damage/hit while shield in secondary has lower damage/hit, that means shield AC has an impact on a 60 druid against a 40 mob. If they're all the same that would imply it doesn't have an impact. Either way I'll then do another test with ~20 less worn ac. This'll all probaby be on Sunday; it takes a ton of time so I'm only doing it while watching football.

The preliminary results showed no difference in damage/hit between lodi shield in back slot and lodi shield in secondary slot, which seems to contradict DSM's results. I don't think we can rule out mob-specific AC squelching, but would love for DSM to run some tests at those same AC intervals as he did previously, but against Shiel.

You should compare equal numbers of hits to check AC mitigation.

That is why I am comparing 400 hits to another set of 400 hits, 1400 hits to 1400 hits, etc.

Your second test has like half of the hits of the first, and we saw with my data that 400 hits can be a bit noisy.

You should compare two sets of 1000 hits, for example. The easiest explanation for your results is due to a significantly different amount of hits for each test, and the second test not having enough hit data to rule out noise.

That is why I did 1400 hits for each test on the level 5 shield/no shield experiment.

https://www.project1999.com/forums/s...8&postcount=68

I'll try to get more data at some point, but kinda busy right now with other things.

[bcbrown]

Originally Posted by TytosOfEight [You must be logged in to view images. Log in or Register.]

So if the idea is that a shield, regardless of slot, acts as a shield and pushes AC above the soft cap, the design should be:

Gladiator's Chain Sleeves and Talisman of Benevolence combine for 23 AC, same as Lodi Shield. So my three parses will be:
1) Wear the sleeves, put the Talisman in ranged, empty back and shield slots
2) Empty arms and ranged, shield in back slot
3) Empty arms and ranged, shield in secondary slot

If a shield has an impact when worn in secondary slot, I'd expect 3 to parse lower and 1 & 2 to parse the same. If a shield has an impact in any slot than I'd expect 1 to parse higher and 2 & 3 to parse the same. If all three parse the same, that implies that at least sometimes shield ac has no effect.

I get where you're coming from with three separate experiments each with two parses, but this seems like a lot less work while answering substantially the same question. Only question this won't address is whether non-shield in secondary slot can push above softcap, I think, but I don't have any high-AC orbs anyway.

Originally Posted by DeathsSilkyMist [You must be logged in to view images. Log in or Register.]

You should compare equal numbers of hits to check AC mitigation.

That is why I am comparing 400 hits to another set of 400 hits, 1400 hits to 1400 hits, etc.

Your second test has like half of the hits of the first, and we saw with my data that 400 hits can be a bit noisy.

You should compare two sets of 1000 hits, for example. The easiest explanation for your results is due to a significantly different amount of hits for each test, and the second test not having enough hit data to rule out noise.

Not gonna lie, it's kinda funny hearing you ask me for longer parses when my first post in this thread was asking you for longer parses and you defending 400-hit parses. But yes, in case I wasn't clear enough I don't think my results had enough hits on either side to be conclusive. Suggestive but not conclusive, which is why I'm going to restart the experiment with a better design and a third option.

But there's no need for the parses to be exactly the same length. 1800 hits vs 1700 hits is just as legitimate as 1700 hits vs 1700 hits. You just have to normalize the total damage into damage/hit and then it's comparable.

[DeathsSilkyMist]

Originally Posted by bcbrown [You must be logged in to view images. Log in or Register.]

Not gonna lie, it's kinda funny hearing you ask me for longer parses when my first post in this thread was asking you for longer parses and you defending 400-hit parses. But yes, in case I wasn't clear enough I don't think my results had enough hits on either side to be conclusive. Suggestive but not conclusive, which is why I'm going to restart the experiment with a better design and a third option.

But there's no need for the parses to be exactly the same length. 1800 hits vs 1700 hits is just as legitimate as 1700 hits vs 1700 hits. You just have to normalize the total damage into damage/hit and then it's comparable.

400 hit parses are fine if you use them in the correct context. This isn't a black and white issue with only one answer.

I use smaller parses across a large range initially to check possible patterns, and then I hone in on areas of interest with larger parses.

This is because going from 0 AC to 300 AC should be a large enough difference to where noise is less likely to affect the outcome, as a simple example. My initial post had a difference of 123 AC at level 5. That is a very large gap.

Clearly you use smaller parses as well, so please do not pretend that you believe smaller parses have no merit. Since you complained about smaller parses, one would think you would lead by example and supply larger parses.

And yes, you need the same number of hits for each set. That is the only accurate way to compare total damage taken, and how many hits are on the lower half of the damage values vs. the upper half. There's no reason to use different hit amounts. If you have one set with 1800 and one set with 1700, just take 100 off of the 1800 test.

[bcbrown]

You don't need to compare total damage taken if you compute damage/hit. Likewise you can convert min-hits and max-hits (and all the rest) to percentages. There's no reason to throw away some of the signal just to match an arbitrary number. But I don't want to fight with you about it. We both agree the experiment I posted last night wasn't conclusive.

[DeathsSilkyMist]

Originally Posted by bcbrown [You must be logged in to view images. Log in or Register.]

You don't need to compare total damage taken if you compute damage/hit. Likewise you can convert min-hits and max-hits (and all the rest) to percentages. There's no reason to throw away some of the signal just to match an arbitrary number. But I don't want to fight with you about it. We both agree the experiment I posted last night wasn't conclusive.

You want to reduce the odds of something affecting the the results. Theres really no reason to have sets with different hit amounts.

I request that your data sets in the future are the same number of hits. Or you can supply your logs.

[TytosOfEight]

Originally Posted by bcbrown [You must be logged in to view images. Log in or Register.]

Gladiator's Chain Sleeves and Talisman of Benevolence combine for 23 AC, same as Lodi Shield. So my three parses will be:
1) Wear the sleeves, put the Talisman in ranged, empty back and shield slots
2) Empty arms and ranged, shield in back slot
3) Empty arms and ranged, shield in secondary slot

If a shield has an impact when worn in secondary slot, I'd expect 3 to parse lower and 1 & 2 to parse the same. If a shield has an impact in any slot than I'd expect 1 to parse higher and 2 & 3 to parse the same. If all three parse the same, that implies that at least sometimes shield ac has no effect.

I get where you're coming from with three separate experiments each with two parses, but this seems like a lot less work while answering substantially the same question. Only question this won't address is whether non-shield in secondary slot can push above softcap, I think, but I don't have any high-AC orbs anyway.

This seems pretty solid and good, practical compromise. Looking forward to seeing the results of it.

Not gonna lie, it's kinda funny hearing you ask me for longer parses when my first post in this thread was asking you for longer parses and you defending 400-hit parses. But yes, in case I wasn't clear enough I don't think my results had enough hits on either side to be conclusive. Suggestive but not conclusive, which is why I'm going to restart the experiment with a better design and a third option.

But there's no need for the parses to be exactly the same length. 1800 hits vs 1700 hits is just as legitimate as 1700 hits vs 1700 hits. You just have to normalize the total damage into damage/hit and then it's comparable.

[bcbrown]

Originally Posted by DeathsSilkyMist [You must be logged in to view images. Log in or Register.]

You want to reduce the odds of something affecting the the results. Theres really no reason to have sets with different hit amounts.

I request that your data sets in the future are the same number of hits. Or you can supply your logs.

I don't want to fight over it and this seems pretty important to you so sure, I'll include a same-hitcount analysis of any experiments in this thread. I'm glad you're paying attention to methodology!

I'm more interested in discussing the potential implications if this experiment I'll run on Sunday ends up with equivalent results for all three parses.

Looking again at DSM's parses with 60 shaman:

level 60 Shaman 157 worn AC vs level 50 mob, 400 hits: 86.22
level 60 Shaman 177 worn AC vs level 50 mob, 400 hits: 83.24
level 60 Shaman 200 worn AC vs level 50 mob, 400 hits: 80.33
level 60 Shaman 217 worn AC vs level 50 mob, 400 hits: 75.57
level 60 Shaman 250 worn AC vs level 50 mob, 400 hits: 73.44
level 60 Shaman 300 worn AC vs level 50 mob, 400 hits: 73.61
level 60 Shaman 386 worn AC vs level 50 mob, 400 hits: 74.52
level 60 Shaman 411 worn AC (25 AC from shield) vs level 50 mob, 400 hits: 68.37

Shaman 60 207 AC vs level 50 mob, 800 hits: 78.18
Shaman 60 217 AC vs level 50 mob, 800 hits: 77.67
Shaman 60 227 AC vs level 50 mob, 800 hits: 76.51
Shaman 60 300 AC vs level 50 mob, 800 hits: 73.27
Shaman 60 386 AC vs level 50 mob, 800 hits: 73.28

The 800-hit parses don't seem to provide any evidence one way or the other, but that last 400-hit parse with 25 shield AC sure suggests shield AC can have an impact.

23 AC Test, 400 hits: 6.60
40 AC Test, 400 hits: 5.05
45 AC Test, 400 hits: 4.54
50 AC Test, 400 hits: 4.35
55 AC Test, 400 hits: 4.32
61 AC Test, 400 hits: 4.27
178 AC Test, 400 hits: 4.32

55 AC Test, 12 of this AC is from a shield, 400 hits: 4.03
190 AC, 12 of this AC is from a shield, 400 hits: 4.12

178 AC, 12 of this AC is from a shield, 1400 hits: 4.11
178 AC, No shield, 1400 hits: 4.31

Again, pretty good evidence that shield AC can have an impact. Looking at the 55 AC tests with and without shield AC shows 4.32 vs 4.03. 178 AC with and without shows 4.31 vs 4.11. The one anomaly that jumps out is that 55 AC with a shield has lower damage/hit than 178 AC with a shield, 4.03 vs 4.11. I don't think that's enough to invalidate anything.

So what could account for DSM finding evidence of shield AC mattering while I found evidence of shield AC not mattering? I think the most optimistic and least likely possibility is that shield AC applies even to a shield worn in the back slot, and when I rerun the experiment with a better methodology that's what I will find. I think the most likely possibility is that for (at least some) mobs it's possible to squelch their damage at a worn AC lower than any softcap such that shield AC is simply irrelevant. I notice that DSM's tests are lvl5 vs lvl5 and lvl60 vs lvl50, while mine was lvl60 vs lvl40, a bigger level difference.

The least likely but most hilarious possibility is that whenever I play EQ shield AC doesn't have an impact but whenever DSM plays EQ it does have an impact.

[Cecily]

When there's a mind watching something, it changes the results. So it follows that you testing would change the results but his don't change.

[DeathsSilkyMist]

Originally Posted by Cecily [You must be logged in to view images. Log in or Register.]

When there's a mind watching something, it changes the results. So it follows that you testing would change the results but his don't change.

I have plenty of videos supporting my various tests over the years. If you want to accuse me of being a master at video editing to the point where I can fake videos perfectly, that is honestly a compliment. That would be difficult to do.

I can record myself doing these tests if you want, but that won't change the results. Most people do not supply videos with their parses/logs, so you are accusing Bcbrown of the same thing you accuse me of.