The long awaited DEX experiment is finally here! Goal: to check how DEX modifies Skill Critical Rate of Yul Archers, and to answer what is the value of DEX required to reach maximum Skill Critical Rate. Here’s what happened: a Yul Moonlight Sentinel 110 was stripped of all his passive skills except for https://l2wiki.com/Superior_Light_Armor_Mastery_-_Yul_Moonlight_Sentinel . The passive was kept because it modifies Skill Crit Rate with DEX. In theory, this experiment can also be applied to Othells, because their https://l2wiki.com/Superior_Light_Armor_Mastery_-_Othell_Wind_Rider has the same statement. To clarify - Yul did not learn https://l2wiki.com/Superior_Critical_Arrow_-_Yul_Moonlight_Sentinel to make sure the only variable influencing Skill Critical Rate is DEX. The target was an Aeore Cardinal 110 with no passives or buffs that could modify the outcome. We started with 61 DEX (base for Elf), and increased it with items and dyes by 5 each time. On each DEX value, https://l2wiki.com/Quick_Shot_-_Yul_Moonlight_Sentinel was used 500 times. Process was repeated until 114 DEX was reached. After initial analysis of results, DEX values of 121, 125 and 200 were set with GM command and tested. Finally, certain DEX values were re-tested with another 500 hits to decrease statistical error and to support a theory. Raw data can be seen in the table below: Short version - quick answers The current Skill Crit Rate in terms of DEX looks something like this: Certain: -there is a point in increasing DEX above 80/90 -there is a softcap around 107 DEX, most probably it’s 110 DEX, which gives ~35% Skill Critical Rate, while 80 DEX gives only ~22.5% Possible: -it is possible that Skill Critical Rate between 80 and 110 DEX scales better than shown (meaning blue line is more ‘straight’), but it still ends at ~35% Skill Critical Rate; full explanation in Long version -it is possible that for DEX values between 125 and 200 (dashed green line) the curve is not a straight line, but as you can see 200 DEX is around 45% Skill Critical Rate, so it’s highly reasonable it’s a softcap area starting from 110 DEX. Conclusion: If you thought 80 DEX is enough: Bonus: from now on most of my experiments will also be added to http://e-repo.pl/formulas (alternatively 2nd link here) because it allows me to create stat calculators and other tools not available on forums. Feel free to check it for a calculator of this experiment, as well as others. Long version - full of maths and insight Spoiler: Math corner Read the short version first. Dictionary: softcap - an area of the function where values increase really slowly compared to previous area hardcap - an area of function where values don’t increase at all Things to consider: the experiment was based on a random event, therefore extremely accurate values are impossible to obtain with less than 10000 hits for each datapoint the 10000 hits per DEX value were not done due to time - a single point with 500 hits took 20-30 mins to measure, all of the points in the experiment took over 10 hours, 10k hits would require 200 hours instead I’ve used statistical error to know how accurate the results are, and drew conclusions based on that and common knowledge (more on that later) there is a small possibility of human error due to the fact that crits were counted manually using hotkeys and an audio cue from ingame alarm when a crit happened, but missing 1-2 crits changes nothing; example of 1 datapoint (500 hits) here: https://www.youtube.com/watch?v=l5wCS3zy9zg&feature=youtu.be I have started analysis by plotting the raw data to see if I could notice any patterns: I immediately assumed that we won’t be dealing with a single function, but rather two or three, one for 60-80 DEX, one for 80-107/110, and one for 110+. To support that claim I re-tested problematic datapoints (DEX=81, DEX=104, DEX=107, DEX=110) to lower the statistical error, and calculated the statistical error for each point. This operation gave me the following plot: The red and blue lines are boundaries of our function. That means there is 99% chance that the Skill Critical Rate function we are looking for is somewhere between these two lines. Mathematically speaking, it’s equally possible for it to be a straight line instead of the curves that I’ve proposed above. This is where “close to certain” knowledge ends. Everything below this point is “probable” based on what is commonly known among Core archers. I’ve taken few things into consideration: in earlier experiments, done by both Innova and the players, 80 DEX was a softcap and 90 DEX was a hardcap in player experiments on this update, increase from 75 to 80 DEX gave much more extra critical rate than increase from 80 to 85; I’ve had the same feeling in early updates, values of DEX above 80 were almost unreachable, and 90 was only possible with 3x DEX dyes, so there was no point to have any formula there, thus a hardcap new items were later added to the game which in turn allowed players to reach much higher DEX values whatever the formula is, more DEX should give more or the same Skill Critical Rate Using the points above, I had the following reasoning. First, the Koreans created some formula that was working until 80-90 DEX and then was capped Later when new DEX items became available, they wanted to add scaling for higher values of DEX, but the most reasonable way to do that without touching players below 80 DEX, was to create another formula for players with more than 80 DEX That formula kept the 80->90 DEX increase low, close to the softcap before the revision After 107 DEX the values grow quite quickly, so they had to add a softcap in case a player reaches extreme values of DEX (with ingame items only, and 3x5 DEX dyes, I reached 114 on an Elf) before they get a chance to revise it The experiment showed that softcap starts at 107 DEX but it’s much more probable it’s 110 DEX This made me simply fit three functions to the data to get a basic model. It is not a perfect 100% accurate model, but it uses the assumptions above. The coefficients are rounded for easier display. And this is how it looks like compared to the experiment data: The values between 125 and 200 were not tested, but I drew a straight line through points with DEX=107,110,114,121,125, and then I continued it and it simply ended really close to datapoint for DEX=200, so it’s quite possible it really is a straight line. There is no reason for testing that for now, as these values are unreachable, and if my theory holds, once they are reachable, the Koreans will change the formula or add another stage.
Actually, one of the most usefull and properly tested experiments, congratulations on making this happens and thanks for your time, I'm pretty sure this will change a lot of ppl's builds.
Very nice Job, but we all know the Dex/Str we can have is limited, so what people rly need is to find out the Sweetspot between Str and Dex, for example with 150 Str and 90 Dex u crit for 27% to 150k dmg, with 100 Str and 140 Dex u crit for 42% to 100k Dmg, or something like that xD Dont wanna talk down ur Work, its an very nice Experiment, kudos to that, and i hope the STR Experiment will follow, i think u got the idea what i meant, critting to 100% with 20% of DMG is still worse then Critting to 30% with 100% dmg xD dont know how else to explain
I ask for characters, items, skills, levels, and when it is required certain special commands (like changing some base stat, item enchant, level, or disabling random damage) are applied to me before or during an experiment. Other than that, I'm on my own, and whatever tools I use to help me cannot break the EULA.
Can you remake the experiment with: 1. No buffs (I guess that's what you made, only the passive) 2. Pom 3. Abundance + pom 4. Abundance + pom + 101 skill To see how the scale, if it's really 27% x 1.25 (pom), or if it multiplies in a different way. This way we can see how to treat modifiers on the damage vs skill crit chance fight..
Very good experiment but in my opinion it would be better with all the passives and easily acquired skill crit stats like from AP or maybe from the weapon SAs + possible buffs POM/Abundance actives to have a better idea about the actual practical usefulness of each DEX point towards skill crit rate. But still overall very useful.
Would be useful to have a complete list of items which boost DEX and how much. I don't know "how much" because I know only low/medium gear TATOO SHIRT HAIR ACCESSORY LONGING/INSANITY CLOAK ARMOR VENIR BRACELET AGATHION LIVINDOR EXALTED PASSIVE EMERALD other?
TATOO = DYES...or at least that's what I meant. Anyway nothing cheap it seems. So maybe better go on focus skill enchanting while can't reach 107 dex..maybe
Cap p.skill critical rate like m.skill crítical rate and all win It is a cap too unfair for my opinion. That the feoh need debuff and can only criticize 1 in 3 And yet the yul ... to infinity and beyond. Jun 24, 2019 3... 2.... 1.... FIGHT PD: very good work anyway.
