Dex/Parry tiers - definitive testing methodology needed

[Coramac]

Well, thanks for the input. To update, I am 30+ hrs in and still going.

I am doing 5k hit parses to start. If I see an anomaly or possibly jump in parry rate, I either retest or let the parse go to 10k (or 15k on some tests).

Include all results. Going to 10k doesn't really do much over 5k. You need to do approximately 100k tests to bring the results under a +-1% margin of error for a single data point. To determine thresholds, it doesn't have to be that precise. We can use a series of points and probability ranges to determine what the actual parry rate is or is highly likely at a given point. Make sure that your parry stat is the same for all tests. It's much easier that way, but if you did change it that may be something that can be accounted for. It was a problem with Ninjiitsu's data.

[maicom]

Very cool. Just finished my 5th crafted accessory looking forward to your results.

[Ruckus]

I wasn't able to parse today due to moved raid times, but here is some of the data so far:

512 parry, 209 dex - 5302 hits, 1346 parries (25.39%)
512 parry, 218 dex - 5077 hits, 1252 parries (24.66%)
512 parry, 220 dex - 5151 hits, 1256 parries (24.38%)
512 parry, 225 dex - 5166 hits, 1328 parries (25.70%)
512 parry, 230 dex - 5733 hits, 1500 parries (26.16%)
512 parry, 230 dex - 10758 hits, 2722 parries (25.30%)
512 parry, 235 dex - 5338 hits, 1302 parries (24.39%)
512 parry, 240 dex - 5730 hits, 1450 parries (25.31%)
512 parry, 241 dex - 11735 hits, 3103 parries (26.44%)
512 parry, 242 dex - 5221 hits, 1371 parries (26.26%)
512 parry, 243 dex - 10071 hits, 2571 parries (25.53%)
512 parry, 244 dex - 16154 hits, 4182 parries (25.89%)
512 parry, 245 dex - 10510 hits, 2807 parries (26.70%)
512 parry, 246 dex - 5268 hits, 1406 parries (26.69%)

Testing above 246 and below 209 sometime this week. Will post results.

[Maqaqa]

Im not the smartest warrior out there but i cant see any DEX "wall" or "roof" or w/ever we want to call it.

But i wonder if DEX gets checked against the target's STR (and parry against accuracy).

[Coramac]

I wasn't able to parse today due to moved raid times, but here is some of the data so far:

Do you by chance have the crits for each test? They don't show up as regular hits. May not be needed, might be helpful.

[Ruckus]

Do you by chance have the crits for each test? They don't show up as regular hits. May not be needed, might be helpful.

Unfortunately not. I am parsing all tonight. I plan to do several 1k trials at each dex point per your previous post. I will record the crits this time.

Any other suggestions before I kick it off?

[Gamemako]

Everything in this game that has been analyzed has worked as a step function.

Uh, nothing does that we know of except parry reduction and damage spread. Skill speed does not. Crit rate does not. All damage functions (STR, WD, DTR effects on damage) do not. HP does not, MP does not, healing amount does not, etc, etc, etc. Parry reduction appears to do so, increasing by strength tier; and damage spread appears to do so, increasing as a function of ceiling(damage/20). That's it. These both seem to be rounded for display reasons.

Yes, I do see evidence of step function here. There appears to be a significant break with the testing around 240ish.

A step increase from what to what? That value is less than 3 other points and nearly identical to another, recorded from the very first data point (i.e. it's lower than the one recorded at 209). There is no subsequent step, as you see from 243-244 tests. There is nothing there. The only thing you see is the grouping becoming tighter due to doubled sample size.

Any other suggestions before I kick it off?

I know it takes a long time, but this data makes it rather clear that you'll need 10,000 hits per trial... I would not expect nearly such a wide variance with the number of trials already done, but there you have it.

//EDIT: Also, please parse baseline -- 341 parry and 193 dex. If you don't, I'll set it up tomorrow and run overnight.

[Ruckus]

//EDIT: Also, please parse baseline -- 341 parry and 193 dex. If you don't, I'll set it up tomorrow and run overnight.

Doing it right now. Thanks for the input.

//EDIT: It will be 341 parry and 192 dex, but I figure same difference.

[Coramac]

I wasn't able to parse today due to moved raid times, but here is some of the data so far:

512 parry, 209 dex - 5302 hits, 1346 parries (25.39%)
512 parry, 218 dex - 5077 hits, 1252 parries (24.66%)
512 parry, 220 dex - 5151 hits, 1256 parries (24.38%)
512 parry, 225 dex - 5166 hits, 1328 parries (25.70%)
512 parry, 230 dex - 5733 hits, 1500 parries (26.16%)
512 parry, 230 dex - 10758 hits, 2722 parries (25.30%)
512 parry, 235 dex - 5338 hits, 1302 parries (24.39%)
512 parry, 240 dex - 5730 hits, 1450 parries (25.31%)
512 parry, 241 dex - 11735 hits, 3103 parries (26.44%)
512 parry, 242 dex - 5221 hits, 1371 parries (26.26%)
512 parry, 243 dex - 10071 hits, 2571 parries (25.53%)
512 parry, 244 dex - 16154 hits, 4182 parries (25.89%)
512 parry, 245 dex - 10510 hits, 2807 parries (26.70%)
512 parry, 246 dex - 5268 hits, 1406 parries (26.69%)

Testing above 246 and below 209 sometime this week. Will post results.

Could you retest:
512 parry, 230 dex
512 parry, 235 dex
512 parry, 245 dex
512 parry, 246 dex

I'm showing 25% parry through 240 dex and a step up to 26% after. I'm a little concerned at the number of points there are outside 1 stdev. While step does fit better than a linear model, it does have a WECO violation, albeit extremely small. I suspect we're missing something or that there is some error in the parser.

[Gamemako]

Pretty messy dataset. Can dump linear regression on it based on deviation from minimum value,

f(x) = 0.400x + 24.01
r^2 = 0.357

I'm wondering if it's a linear increase in DPS relative. I noticed that my block strength with the same shield decreased as I leveled through Toto-Rak, which suggests to me that it might be all normalized by the base level value. If so, the 512 parry would suggest a 15.99% base parry rate with base dex and base parry. Just a hypothesis -- should be easy enough to test using a weathered axe and no +dex or +parry gear.