Balancing AC and HP (For Paladins/ShadowKnights)

[Jimjam]

It could be his AC is way too low or way too high for the shield to make a difference.

[pharmakos]

sakuragi -- what were your raw item AC totals during those tests?

[kaev]

Originally Posted by Splorf22 [You must be logged in to view images. Log in or Register.]

great post

I'd be very cautious about drawing any conclusion from this small amount of data. Given the behavior I've seen from the p99 pseudo-random number generator while crafting, I'd want to see a much larger sample as the basis for analysis.

Also curious about distribution of the hits. On live, ~ 20% of all hits were (dmg*2)+(dmgbonus), with additional peaks at min and max, the remainder were fairly evenly distributed. And the min and max peaks were affected by extreme values of ATK vs. AC (i.e. low level mob would get hardly any max and a lot of min.)

[koros]

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

It could be his AC is way too low or way too high for the shield to make a difference.

It could be, although if that's the case it should still have a marginal effect. 40 raw AC is a LOT, and the cliff golem is an appropriate leveled mob for a 60 war to be tanking.

[koros]

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

I'd be very cautious about drawing any conclusion from this small amount of data. Given the behavior I've seen from the p99 pseudo-random number generator while crafting, I'd want to see a much larger sample as the basis for analysis.

Also curious about distribution of the hits. On live, ~ 20% of all hits were (dmg*2)+(dmgbonus), with additional peaks at min and max, the remainder were fairly evenly distributed. And the min and max peaks were affected by extreme values of ATK vs. AC (i.e. low level mob would get hardly any max and a lot of min.)

It's too small to be definitive, but 150 hits should be a representative sample, *most of the time.

[Splorf22]

I think it is reasonable, even likely, that the shield performed worse due to insufficient sample size. You can see the 95 CI bars are quite large. That's why I was very careful with my conclusions. My suspicion is that the 40AC is good for 1-2% mitigation which the test is simply not measuring. But if I were Kanras I would definitely be rechecking the code for a possible bug.

Sak was at like 170/210 item AC for this test which I think is a pretty reasonable range for most people I think (a full set of indicolite is 155AC according to the wiki). One unknown factor in this whole business is the Iksar AC bonus and how its screwing with things.

Oh, and one last thing. I wrote a little sim code. Even if AC works HP is still by far the way to go IMO. Here are the simulated results for Sakuragi with various amounts of HP against the Spiroc Lord with evasive disc up (I took the numbers from two parses). So at 4500 HP, in order to have a 90% of surviving against the Lord for 100 seconds, he would need a complete heal every 14.5 seconds. AC0 means 15 and 5% chances for max and min hits; AC1 means 14 and 6% chances for max and min hits; AC2 means 13 and 7% chances and so on. I hope all that makes sense.

So according to these sims, even reducing the max hit chance from 15% to 5% (AC 10), which would probably require several hundred item AC at least, is only worth about 600 raw HP. Those results make a lot of sense to me, because even at AC10 you are basically flipping one hit from max to min. If you are getting a CH every 20-25 hits, its quite possible that you could just get unlucky once or twice, whereas HP provide more space for the hits to return to the norm. You can in fact see that each HP is worth more than the previous one. (4500->5000 is worth an additional 2 seconds per CH, but 0->4500 is worth only 11 seconds not 18).

Of course that analysis assumes AC is working . . . [You must be logged in to view images. Log in or Register.]

caveat lector: i may have had a bug in my code

Increasing HP (4500): 90%: 14.5, 99%: 11.5, 99.9%: 10.5
Increasing HP (4600): 90%: 15.0, 99%: 12.0, 99.9%: 11.5
Increasing HP (4700): 90%: 15.0, 99%: 12.0, 99.9%: 12.0
Increasing HP (4800): 90%: 15.5, 99%: 13.0, 99.9%: 11.5
Increasing HP (4900): 90%: 16.0, 99%: 13.0, 99.9%: 11.5
Increasing HP (5000): 90%: 16.5, 99%: 13.5, 99.9%: 12.0
Increasing HP (5100): 90%: 17.5, 99%: 14.0, 99.9%: 12.0
Increasing HP (5200): 90%: 17.5, 99%: 15.0, 99.9%: 12.5
Increasing HP (5300): 90%: 18.0, 99%: 15.0, 99.9%: 12.5
Increasing HP (5400): 90%: 18.5, 99%: 15.0, 99.9%: 14.0
Increasing HP (5500): 90%: 19.0, 99%: 15.5, 99.9%: 14.5
Increasing HP (5600): 90%: 19.0, 99%: 16.5, 99.9%: 14.5
Increasing HP (5700): 90%: 19.5, 99%: 17.0, 99.9%: 14.5
Increasing HP (5800): 90%: 20.5, 99%: 17.0, 99.9%: 15.0
Increasing HP (5900): 90%: 21.0, 99%: 17.5, 99.9%: 15.0
Increasing HP (6000): 90%: 21.5, 99%: 17.5, 99.9%: 15.0
Increasing AC (0): 90%: 14.5, 99%: 12.0, 99.9%: 11.0
Increasing AC (1): 90%: 14.5, 99%: 12.0, 99.9%: 10.5
Increasing AC (2): 90%: 15.0, 99%: 12.0, 99.9%: 10.0
Increasing AC (3): 90%: 15.0, 99%: 12.0, 99.9%: 10.0
Increasing AC (4): 90%: 15.0, 99%: 12.5, 99.9%: 10.0
Increasing AC (5): 90%: 15.0, 99%: 12.0, 99.9%: 10.0
Increasing AC (6): 90%: 15.5, 99%: 13.0, 99.9%: 12.0
Increasing AC (7): 90%: 15.5, 99%: 13.0, 99.9%: 11.5
Increasing AC (8): 90%: 16.0, 99%: 13.5, 99.9%: 11.5
Increasing AC (9): 90%: 16.5, 99%: 13.5, 99.9%: 12.0
Increasing AC (10): 90%: 17.0, 99%: 14.0, 99.9%: 11.5

[Jimjam]

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

It could be, although if that's the case it should still have a marginal effect. 40 raw AC is a LOT, and the cliff golem is an appropriate leveled mob for a 60 war to be tanking.

Then I guess the AC cap is lower than we thought and as we would expect there is no overcap bonus for shields (unlike live)?

Edit: actually, perhaps p99 has been programmed to correct for the shield AC overcap bonus but the correction slightly overcompensates, hence the shield parse being worse? /shrug

[koros]

Also... you should parse yourself tanking either naked or with the same worn ac as another 60 war with the same agi. That could yield some interesting information about the iksar ac bonus.

[pallius]

Shield bonuses came later in the game. Not classic.

[Sazzabi]

Here is the numbers for my parses on low dark blue mobs, I am a level 51 Troll Sk

Mobs - Nro freeport guards (Guard Brendyl, Guard Stoutman, Guard Fintran), all level 40, max hit - 100, min hit - 17

The 963ac and 837ac parses were around 1 hour total tanking time each. The 916ac and 681ac were about 30 minutes each.


244 raw ac (963ac), with shield, 100agi
avg hit 41,16, hit% 47.20

164 raw ac (837ac), with shield, 100agi
avg hit 45.07, hit% 48.26



216 raw ac (916ac) no shield, 100agi
avg hit 42.13, hit% 48.89

63 raw ac (681ac), no shield, 100agi
avg hit 57.13, hit% 45.91

Based on this I am leaning towards the hardcap on this server not being exactly what it is in Kavhok's post.