I enjoyed playing Anarchy Online, mostly due to the learning process. I love to demystify and understand how, why, and what things are done ingame. In my opinion, the evade formula is the last, and only true mystery left in the game, and so I have taken some steps toward finally revealing it.
Lots of theories have been posted, all explaining bits and pieces of what factor effects what other factor, why it's one way instead of another, etc etc., but although many of the people presenting these theories may swear by them, they lack the data to substantiate any theory so far.
Therefore the purpose of this thread will be to post actual, hardcore data from ingame, for analysis - following a strict set of rules.
I will share the data I have collected thus far at the end of this post, but I would first like to explain my approach to this task:
I decided to analyse the percent chance to hit your opponent with regular hits [no specials, no nanos, etc] in order to determine the
relative effect of each variable on the percent chance to hit equation.
Variables of Interest
Defender:
- Evade of interest (dodge rngd or evade clsc depending on attacker weapon type)
- Add All Defensive modifier, including research, perks, gear, towers...
- Position of Defender Agg/Def Bar
Attacker:
- Weapon Attack Skill (as shown in IP arranging window - base attack skill)
- Attack Rating (as shown in Stats window, base attack skill + add all offense modifier)
- Position of Attacker Agg/Def Bar
and, of course:
- The number of hits Attacker Lands on Defender
- The number of misses Attacker makes on Defender
[which combined, will give you total number of attack attempts made]
Percent Chance to Hit = (Hits) / (Hits + Misses) or (Hits) / (Total Attacks)
While some variables above may seem odd to record, such as attacker weapon attack skill, and attacker agg/def bar position, it is important to make note of them. It is unknown whether a point of Add All Offense has the same effect as a point of weapon skill on chance to hit. It seems likely by intuition that it is equal, but we can't rule out the possibility that it is not. Similarly it is difficult to say for sure whether attacker agg/def bar position also has an effect on the frequency of hits landing vs missing on your opponent.
Method
After doing some inital tests and fooling around with statstical error calculations, I realized that 100, 200, even 500 total attack attempts was not going to be enough to eliminate the wide range of error possible. I decided to perform tests with at least 1000 attack attempts, in order to get a more correct percentage, not due to random chance. Ideally 10,000, or even greater, would give more precise results but it would take roughly 5 hours to perform one test like that, so this is simply constrained by time.
I believe the best way to find out the effect of one variable on the hit/miss ratio is simply to keep all other variables constant, while you slightly alter just the one variable and take multiple test readings at different values. This way, you acquire a range of data that shows the change in hit/miss frequency based solely on variable X.
How I performed these tests was by creating a log window on the defender, which logged only other's hits and misses - allowing attacker to hit defender for about half an hour - then copy/pasting the log into word and doing a replacement search for "Bla hit " and "Bla tried to hit " - word is smart and tells you how many replacements it made in the document (im sure you could do this with some damagedumper like aoparser even, but i was just too lazy to dig my old dumper up). Log was obviously wiped after every test run, in preparation for the next.
Current Data
Since I can't seem to think of anything else to talk about, I'll go ahead and post my initial results, where I tried to alter different sets of variables to see their effect on hit/miss ratio.
Feel free to analyse those results in any way you'd like - I excel graphed the hell out of them, calculated the error values, etc etc and didn't find any easy-to-identify patterns. Then again I'm not too great with stats, so I'm hoping to catch the eye of someone curious, like me, with more knowhow, hehe.Code:Agg/Def Bar Position Weap Skill AR Evade AAD Defender Attacker Hits Misses Total Shots % Shots Landed 1763 1900 2209 440 50 100 631 545 1176 53.66 1763 1900 2209 670 50 100 611 582 1193 51.22 1763 1900 2209 1060 50 100 1230 1512 2742 44.86 1763 1900 2209 1280 50 100 619 963 1582 39.13 1763 1900 2209 1540 50 100 396 770 1166 33.96 1763 1900 2209 1825 50 100 378 808 1186 31.87 1763 1900 2209 2540 50 100 312 944 1256 24.84 1763 1900 1729 1540 50 100 657 1090 1747 37.61 1763 1900 1751 1060 50 100 854 940 1794 47.60 1763 1900 2532 1060 50 100 412 785 1197 34.42 1763 1900 2532 1360 50 100 384 842 1226 31.32 1763 1900 2532 1540 50 100 335 879 1214 27.59 1763 1900 2209 1060 88 100 533 685 1218 43.76 1763 1900 1729 1540 88 100 525 690 1215 43.21 1763 1900 2209 1060 0 100 345 888 1233 27.98 1763 1900 1729 1540 0 100 411 925 1336 30.76 Where Full Agg = 100, and Full Def = 0
Please try to keep random theories and speculation out of this thread unless it is based on data somehow, and if so, please provide data to back up your argument.
Let's do this scientifically, once and for all