Dexterity's impact on critical hits

[CDF]

Practically speaking, I imagine 99% (just made it up) of players couldn't (and shouldn't) care less, but I am just curious to see how it pans out as far as hit rate/crit rate being discrete or not and this method requires no effort except to make sure you hit for 0 dmg.

[Masamune]

this method requires no effort except to make sure you hit for 0 dmg.

That's actually harder than it sound for persons elvaan and/or having low DEX gear opportunities :s

EDIT: updated with CDF's typos + Pchan dDEX6 additional data + removed Searain data @ my earlier post http://www.bluegartr.com/forum/3203925-post234.html

Launching dDEX=14 this night.

[pchan]

what would be the stats of Om'hpemde ? They can be level 77.

[Masamune]

using my own excel tool for stats @ lvl73-77 (listed levels for Om'Hpemdes @ wiki), i obtain the following stats:

- Om'Hpemde lvl73:
VIT=76
AGI=70
DEF=298
EVA=287

- Om'Hpemde lvl74:
VIT=77
AGI=70
DEF=303
EVA=292

- Om'Hpemde lvl75:
VIT=77
AGI=70
DEF=308
EVA=297

- Om'Hpemde lvl76:
VIT=80
AGI=72
DEF=314
EVA=302

- Om'Hpemde lvl77:
VIT=80
AGI=72
DEF=319
EVA=307

i didnot verify with checks though so take care :s

which would mean in terms of max STR for 0 dmg :

dSTR range         fSTR approximation              max STR to get 0 dmg on:
                                   Ul'Hpemde lvl 68 69 70 71 72 73 74 75 76 77
12 or greater      (dSTR + 4) / 4 <=-1           63 64 65 67 67 68 69 69 72 72
6 to 11            (dSTR + 6) / 4 <=-1           61 62 63 65 65 66 67 67 70 70
1 to 5             (dSTR + 7) / 4 <=-1           60 61 62 64 64 65 66 66 69 69
-2 to 0            (dSTR + 8) / 4 <=-1           59 60 61 63 63 64 65 65 68 68
-7 to -3           (dSTR + 9) / 4 <=-1           58 59 60 62 62 63 64 64 67 67
-15 to -8          (dSTR + 10)/ 4 <=-1           57 58 59 61 61 62 63 63 66 66
-21 to -16         (dSTR + 12)/ 4 <=-1           55 56 57 59 59 60 61 61 64 64
-22 or less        (dSTR + 13)/ 4 <=-1           54 55 56 58 58 59 60 60 63 63

[Yugl]

How far along is this test and are there any safe formulas (Or conclusions) thus far? It looks like so far we have a steady growth in critical hits until dDEX=40 where the hit rate dramatically increases.

[pchan]

How far along is this test and are there any safe formulas (Or conclusions) thus far? It looks like so far we have a steady growth in critical hits until dDEX=40 where the hit rate dramatically increases.

dDEX=0 => 9% crit
dDEX=39=>13% crit

on "average", 1% per 10 dex... which is totally underwhelming.

[Shamaya]

It's somewhat important, but already.. people can make a pretty general idea of things. It's probably already safe to use rough linear piecewise calcs in personal estimations. A friend of mine was telling me earlier today I should WS in Byakko's over heca+1. Assuming I'm in the 'high tier', that could be something like 3acc and 6% crit vs 22 attack. If in the lower tier completely, that's something like 3acc and 1% crit instead. So this information does have practical application.

Bout to go to bed/work and do dDex42^^

[Masamune]

Woke up seeing for dDEX14: 1658/14613=11.35%.

Definitely a jump starting either @ dDEX=12(=10%) or 13(=??%).

[Shamaya]

Not a whole number? Nooooo.

Did dDex42 while sleeping / at work,
http://img.photobucket.com/albums/v7...Rate42dDex.png
This is so far still consistent with the "1/12th" or 0.833%crit/dDex model for the upper tier. But we see where consistency is or isn't getting us so far w/ the lower tier. Given that this so far appears semi-consistent, I'll do my next data point at around where the cap is supposed to be. It seemed from the colibri data, to me at least, that it was at 51 rather than 50. So far my samples on the fishies would predict 52, but *shrug*, we'll see. So I guess I'll do 50 -> 51 next. At this rate we should be finishing this relatively soon.

Though after the model looks set, double-checking on a level correction would be best. On Om's of course.

[Masamune]

er.... doing dDEX50 is way too early imo < <

dDEX47or48 sounds farthest to test before testing above.
Also in your case, there is a jump at maybe each dDEX increase, so you have hardest part of the dirty work :/

I can understand you tempted to skip points, but that's bad imo :s (especially in that region)

Added yur point to current data, thank you deeply.

[Shamaya]

np. We don't /have/ to do every point. We can, if we want to. But it really depends on how confident we want to be and what we're satisfied with. I highly doubt any game testing in ff yet would have seen samples and care close to as much as these. If you go back to CDF's post on investigating a linear relationship, he explains good reasons why it's ok to not test at each point of deviance (or dDex in this case). But I see where you are coming from, wanting to test very carefully. However I do believe it's time to glance at the cap and then work back from there, and I think 50 as opposed to 48 is a fine starting point. Nothing is keeping me from doing 48 or others later.

[pchan]

dDEX=15 : 1212/11234=10.79% +/- 0.57%

Looks likes dDEX=14 and 15 are at 11%. Doing 13 next.

[Masamune]

ok Pchan and Sham.

I'll continue dDEX12 data : now @ 1849/19275=9.59%+-0.416%.

But for Sham, CDF were talking about the region BEFORE dDEX40, since there are a few points having same crit%, which isnot the case at all for the region AFTER dDEX40, that's why i still think needs to test more carefully the region dDEX=[41;50]. But is np if you(and others) prefer to do it backward from dDEX50, as long we get data for EACH of these points.

[Shamaya]

Nah you got me wrong. Ugh, how do I explain this. Basically I'm referring to his concept, that even if you parse many different points, you can combine them as if they were the same sample. This lets you look at the linear relationship which will "hold true" in some meaning of the word "true," even if the real formula is not linear. If that makes sense.

But don't worry, if you want to find the "real" very precise formula, with high confidence, I'm sure we'll do that. I find that I'm often happy with highly accurate though possibly imperfect results. But we might as well one up ourselves.

[CDF]

More dDEX = 24 (DEX = 91) on a level 72 Ul'hpemde (combined results from the previous trial):

Crit rate: 2443/20218 => (11.63%, 12.54%)

Hit rate: 33052/37580 => (87.6%, 88.3%)

I already said the predicted hit rate from direct calculation is 88%, but I forgot if that was a truncated value. I'll have to check later but I'm busy.

[Masamune]

I'm gonna test dDEX=21 this night.

If dDEX/7+9% is correct for 1st and 2nd jumps, it's predicting 3rd jump occuring between 20 and 21.
The JP data shows data for dDEX20: 165/1491=11.06% +-1.6%, but we have nothing for dDEX=21.

[pchan]

dDEX=13 : 978/9454=10.34% +/- 0.61%
I'll continue it tonight but it seems to be 10% crit, which means the second tier happens at dDEX=14, although I doublt it happens every 7 dex since dDEX=39 gave 13% IIRC.

[Masamune]

true im like wtf when i plugged crit%=ddex/7 + 9 and saw the jump @ 14% before dDEX39 (occuring @ dDEX35).

also, there s another jump to test around dDEX28-30 (which i can't test).

EDIT: dDEX21 test launched. see yall tomorrow.

[Masamune]

Got a thought about WS crit based testing:

Would it yield exploitable results to autoWS for high samples using this same method + auto-meditate(since 0dmg earns 0TP) + relic base weapons?

[Foldypaws]

Got a thought about WS crit based testing:

Would it yield exploitable results to autoWS for high samples using this same method + auto-meditate(since 0dmg earns 0TP) + relic base weapons?

No way of telling if a WS crit if it did 0 damage, there's no log info.