Experiment 175 Level difference influence on damage dealt

Goal: to verify damage bonus/fall-off granted by level difference between characters

Reason of experiment: Orfen update made new character levels possible, creating more situations where characters of different levels fight each other.

During the experiment a Tyrr Duelist was hitting an Aeore Cardinal 105 with https://l2wiki.com/Sonic_Flash_-_Tyrr_Duelist . Duelist’s level and weapon enchant value were modified during the experiment, to keep a constant (or at least close to constant) P.Atk.

Spreadsheet with raw data.

It was assumed that [base damage] occurs when characters have the same level. The plot of raw experiment data can be seen below:

The x axis represents level difference between attacker and defender. The y axis shows what portion of base damage is dealt for that particular level difference.

Spoiler: Math Corner

The data above was approximated by a linear function

Result of approximation:

As it was noted before, the P.Atk was close to constant during the experiment. As it turns out, points below the red line are where P.Atk was too low, and points above the red line are where it was too high. This allows us to confirm that the approximation is correct.

It has been concluded that each level of difference between characters (with same stats on each level) gives 0.87% damage bonus or penalty, depending if the attacker’s level is higher or lower than the defender.

The bonus can be calculated for any level difference using the formula:


giving a value we use to multiply base damage.

Practical example:
Damage when characters have the same level: 10000
Damage when attacker has 3 levels more than defender:

Damage when attacker is 10 levels behind the defender:

Conclusion:
The values are not spectacular, but paired with the fact that P.Atk and M.Atk also scale with level, the total damage boost reached by gaining new level is higher than just the bonus from level difference.

 

Thx bro !

 

it was too hard to make table with:
+1lvl - -10% dmg
-1lvl - +10% dmg
?
or smth like this.. no offence. good that you doing those tests. but it's 2nd time where i barely understand something..

 

thx. much better.

 

Best experimenter Eu

 

Don't even mind him, he doesn't care he just here to hate.

 

I should have a solution for all the "it's not clear" issues after the next experiment.

 

Another experiment where s some random numbers just to show how cool you are xD
Insteed of give just a results, you show everyone how you do math like anyone cares.... oh wait your clanmates care, at least they pretend to because thanks of fake friendship and telling you how good you are, you will make their experiment in fist hand and after this they will not care anymore about you super useless math.
Last two experiments just show how attention you need to show how "smart" you are. Insteed of gave just a results that ppl looking for.

0.87% dmg diffrence between 1 level. Wondering from where are rumors its almost 10% many ppl spread around this theory about 1=10% but seems its diffrence. Well at least good you do something AttentionButGoodExperimentator < Your New Nick.

 

You did the experiment completely backwards. You take tyrr's constant level and increase/decrease level of defendant without learning passives...Of course both chars completely naked. I highly doubt they changed much since last experiment was done (on old ncwest forums) a few years ago which indicated a quite higher numbers than you have here. And knowing l2 the numbers are hardly linear.

 

yeah, totally matter which level changes when we test difference, 50iq there buddy
also if your mentioned ncwest test was made in the way you described, it was simply broken. If you just modify level and keep everything else intact - it will also scale your attack stats which affect dmg on top of level difference.

 

so finally some decent tests, not some "we open 100 random *** chest and what is inside that".. and somehow it is too much for some people to understand it.
you deserve all the *** what innova gives you after this, for real.
if u dumb, why they should handle like you are not. omg.

ps.: i cant wait to see the nonsense after this like, why so mad and any other useless bs from this ppl.

 

hate or not. but as for "noob" it's hard to understand what he write there. and as rnika said - too much fancy math skills and too less plain info. just calculations and forulas what is not interesting for majority of peoples. good that someone doing those tests, but bad that he doing in geeky way what very few people can understand. as for me it would be nice to see simple test.
attacker 99. defender 103
hit - 9000
hit - 9001
hit - 8999
attacker 100. defender 103
hit 9100
hit 9099
hit 9101

or something like this. and now with this "test".. calculations calculations.. well.. i was good in maths in school, but that was 15y ago. and everything what i need from that is basic stuff to get %, sum or length or area in cm/m2

 

Yes, next time make test results with colorful pictures and table, so people, that finished school 15y ago would understand, coz (lvl1-lvl2)*difference in dmg+base dmg is just way too hard :/
Anw Thanks Palka, test more useful than 90% of previous ones we got last year

 

1. The way you described the experiment clearly says it was done incorrectly. If you test for a single factor, in this case, level difference influence on damage, you have to make sure every other factor is constant in the experiment. That means you need to have same P.Atk, P.Def, attributes, so that the only change in damage comes from the level difference, and not bigger P.Atk/lower P.Def due to level change.
2. I always do all the experiments that require it without passives or buffs or extra unrequired items.
3. The numbers in L2 are whatever devs choose them to be, sometimes it's linear, sometimes it's a function that's combined from 5 other functions.
4. I did a lot of function approximation during my uni times, so I learnt how to collect data properly to have a chance at guessing the function.

Thats why I always try to give general formula -> for people who care and want to calculate and speculate, and I also give an example to show how it works. But all the "it's not exactly clear" issues will be fixed with my 3rd experiment, expect a surprise

I will do that also in a much better way, soonTM

 

problem is not that i don't understand or something. i just want to see results. not calculations. and somewhere in middle of that, under spoiler is actual number what is most important for peoples who requested this test. now i see you edited post. but before iirc it was under spoiler button.
so in short: when i come to see test, i want easy and simple dmg logs and some easy calculation, not some fancy formulas, and as i heard you can call a devil if you try to read them loud 3 times in middle of the night in full moon.
for example how you get that 0.0087? saw that number on wall?

 

"Linear approximation using least-square criteria." It's a way of finding a line that's closest to a spread set of points.

My algorithm found that to be y=0.0087x+1
Now in linear function, the number in front of X, 0.0087 means "if I increase x by 1, result will increase by 0.0087".
0.0087 is 0.87%

To confirm it you can look at a point where x=0 and y=1. If you add 1 to x (level difference), the next point is at ~1.0087.

Nop, it was there, I made it bigger

 

well.. that part with your "algorithm" confusing me. what if it's wrong? and it should be 0.0088? or 0.009 ?

 

1. The algorithm is designed to find the "most-fitting" candidate. I didn't design it, I just use works of famous mathematicians.
2. It's used by engineers worldwide to build bridges and electric power plants, and they work kinda ok I'd say.

 

so from where and how you figured out that 0.0087?