Critical chance

[Joe[IS]]

A few days ago, I was testing something, firing single shots of impact ammo at a target.  I happened to notice that my crit chance was nowhere near what I expected it to be.  I'm a PW with max Critical Targeting and a 200% Crimson Force Barrier, which adds ~26%. 

 

The way I always thought it worked was that the bonus from your Critical Targeting skill plus the highest critical targeting bonus from gear was the chance you would deal a critical hit.  If this were the case, I should see very few non-crits.  However, what actually happened was that I saw a non-crit almost every other shot.  I didn't fire a huge number of shots, but I did see quite a few non-crit shots - far more than I'd expect to see if I had a 96% chance to crit.

 

I don't think that it's just a multiplier of my base crit chance either.  Base crit seems very low - I'd bet that it was less than 10%.  If I roughly doubled this estimate, I'd be looking at about 20% crit chance, which also doesn't seem even close.

 

Has anyone actually tested how crit chance works?  How does Rally affect it?

[Zackman [LDEV]]

Hi !

 

What was the mob level you were fighting ?

[Joe[IS]]

I want to say it was a L20 Claimjumper in Carpenter.  I know it was in Carpenter and it was a ship-type mob that wasn't Red Dragon faction.

[Zackman [LDEV]]

Heya!

 

I was equipping my test toon with the very same setup (tho i used Plasma ammo) you are using on your Progen Warrior and had a crit almost every shot.

Make sure that the Combat trance is up and the bug didnt hit you (tho it should be fixed) - so the CT circle is spinning around you.

[Joe[IS]]

Interesting... I didn't think to mention it before, but I was using Impact ammo at the time.  Also, I didn't think Combat Trance affected my critical rate?

[Zackman [LDEV]]

Combat Trance will raise projectile accuracy. With misses you cant hit critical - but this is not relevant in this case since you fought a L20 mob being OL150 yourself.

Also is the ammo type not relevant for crits - i just meantioned it because i saw CHEM on your weapons when i was copying your setup for testing.

 

Like said, the crits were flying constantly on the mobs head when i tested and it shouldnt be diff. for you.

[Joe[IS]]

I'd like to test this better on a mob with much more HP but low damage output.  Alternatively, should using a lower level weapon affect my chance to crit at all?  If not, I'll just go get some low level gun and record a larger sample.

 

I went over to the Claimjumpers earlier with a single Archos with impact (makes it easier to tell if the shot that just landed was a crit or not), but they don't really last long enough to get a reasonably large sample.  Out of the 30ish shots I counted (not including the kill shot, which doesn't display damage), 7 or 8 were non-crits.

[Zackman [LDEV]]

Dont need to use lower guns, your setup is fine. One single shot archo is good for testing.

Remember that your angle of facing the mob and movement also affects the crit chance!

 

A crit happens really "between the eyes".

 

Also remember that a crit chance is based on a random formula like

Random from 0- 100

Crit chance 96

If randomizer pulls number from 0-96 you have a crit.

 

If you are unlucky, randomizer pulls the 97 5 times in a row (just an example), so you need way more shots to have a reliable average.

it can also happen the randomizer pulls from 0-96 for 200 times....so the average is building up over the amount of shots.

 

30 shots is not reliable.

 

 

 

I was testing it and checking the code and its working bug free, so all a question about a reliable average.

[Joe[IS]]

I didn't know it was based on angle or movement.  Do they heavily influence the formula or is it only a slight difference?

 

Also the reason I wanted to try out low level guns is because they do less damage, which means I get to sample many more shots than I do with an Archos on the same mob -> more reliable sample.

[Joe[IS]]

I just tested a low level gun against Trever Ardun, CL25 named Claimjumper.  I used autofollow to close the distance and did not touch movement controls at all during the "fight".  I fired exactly 600 shots, and 112 of them came up as non-critical hits.

 

I will test again when I come home from work tonight.

 

In a binomial distribution using p = .9591 (Critical targeting skill contributes .7, CFB contributes .2591 for this total), the probability of getting less than or equal to 488 successes out of 600 is 2.22 * 10^(-16).  In conclusion, either I'm missing some very important variables of critical rate, or the PRG is badly broken.

 

Can you detail all of the critical rate calculations?

 

edit: Fixed statement.

[Rossdie [BT]]

Actually zach just said its a randomizer from 0-100 so the only way to guarentee crit hits is to get your percentage of crit to 100% or higher

[Zackman [LDEV]]

Ihasabucket, it is called "critical chance", not "critical guarantee".

[Rossdie [BT]]

All I know is I see plenty of crit hits from my missiles on my TE

[Joe[IS]]

I know it's a chance and not a guarantee.  Hence why I'm using probability theory to back up my findings.  The probability of getting such a discrepancy between the advertized critical chance and my sampled critical rate is extraordinarily low.  If the PRG was indistinguishable from random, as is the goal of a PRG, then the probability that my sample contains as few or fewer criticals as mine did is astronomically low.

[Joe[IS]]

I just finished test #2.  Target was CL31 Manes.  Number of shots was again 600, with 153 non-critical hits.  Again, auto-follow with no movement. No shots missed. Something I noticed in both trials is that the client is combining numbers of two shots together.

 

The numbers I saw in this test were 8 (-2), 16 (-4), 16 (-4) (CRIT) and 33(-8) (CRIT).  This means that a single non-crit must be the smallest number, 8, and that a single crit must be 16.  I never saw a damage number of 24 in the entire test, which implies that the client does not combine crit and non-crit shots together.

 

Using a binomial distribution again with probability of success, p = .9591 and n = 600 trials, the probability that we pick a sample containing 447 critical hits or fewer is 1.15 x 10^(-74).

 

I'm not sure what's going wrong but it's very clear that my actual critical hit chance is far below 95.91%.

 

I would be glad to help you look over the code, if you'd like.  It's quite helpful to have a second set of eyes for debugging.  I'm proficient with C++ - would just need Kyp to set up the access.

[Stanig[IS]]

I think its likely that equipped crit isnt working again.  70 percent crits on a PW?  Thats the CT rank 7 skill, but try it with a JE with energy beams with a same CL mob to really test equipped crit.  In my tests last year, equipped crit was not working at all.  (use energy because you can get lost real quick counting plasma ticks, they tend to overlap)

 

I was barely managing a 3 percent crit rate with an expected 20 percent on 1000 test shots.   Seriously, out of 1000 shots I had about 34 criticals iirc, id have to look it up.  This was last year though.

 

It is very likely, zack, that the equipped critical value is not actually being calculated in.

[Stanig[IS]]

Currently running test on DEV server.

 

150 JE with 3x Level 9 Purple Dragon Lasers M9 Equipped (18.17 crit chance buff) and all prototype gear so the toon has no other buffs on it.  Testing vs a CL50 Bio Extractor.

 

 

Results for test heat 1:

 

300 Beam shots fired

300 Beam shots hit (100 percent hit rate)

14 hits flytext as critical with double damage (4.667% Critical chance observed, well below expected 18.17)

 

Will run additional trials if necessary, but my conclusion is equipped crit is NOT functioning, and you are only seeing skill based crit bonuses + whatever base chance to crit exists.

[Zackman [LDEV]]

I think its likely that equipped crit isnt working again.

 

It is very likely, zack, that the equipped critical value is not actually being calculated in.

 

I see where you come from, checking this...

[Zackman [LDEV]]

We need to put this on hold.

 

I will bring this up internally as i can see something, but i dont know if this was intended.

After this has been clarified internally, we will come back here.

[Stanig[IS]]

:panic:

[Zackman [LDEV]]

Hi Players!

 

The critical targeting had a mechnism in that was handling certain conditions diff. from what it's supposed to do.

This has been fixed now, so critical targeting should show the results you expect after the next tuesday maint.

[Stanig[IS]]

[Joe[IS]]

I... I think I mirror this sentiment?

[Willbonney]

As reported in a posting I put in a couple years back here: https://forum.enb-emulator.com/index.php?/topic/2274-cl-66-was-never-a-mobs-true-level/

 

I believe this fact to be the Combat Level issue of the mobs. It seems, to me at least, that the True Combat level of the mob is the mob level shown in the indicator. I do not remember that being similar to this is in the live standing. From what I understood in live, everyone had an innate 50% chance to crit on a mob as the same level as you. As you went up and down a level, it changed by 10% (therefore 60% for you being 1 higher, 40% if you were one lower). This calculation was added before buffs, equipped, or activated modifiers, so that it stacked, but the max level was 99% and not 100%.

 

This was my understanding, and was why we saw an "almost" perfect crit rate against mobs 10 levels or up, less than us even without equipment, and did so as well for mobs 5-7 levels higher.

 

This ties in with the mob level deal, because even with it being a OL level (which is why we used the acronym for Overall Level and not Combat Level even on mobs and not just players), we never saw more than a 30-40% decrease on our crit rates even against CL 66 bosses.

This had another effect, with skill types such as "Rage" that even though you were fighting an OL 66 Boss, you were seeing as though you were fighting a Combat Level 58 type character (as though an in game player was combat level 58 or OL174)

Special skills then have a better chance of landing on a mob, instead of seeing constant failures. Hell, I remember what it was like fighting a PS with Menace in the ring that had gear to increase his skills to the rate of a Level 58 mob without gear. I remember even with 50 Psiongic resist, it would hit often, but was definitely not every time.

 

Another thought that just came to mind and I haven't checked it out yet. Don't you all remember that when you got a certain resist amount over a mob, you took 0 damages. Or that if you weren't doing more damage than its resists, you again saw 0 damage hits (even crits that showed 0). This was due to the fact that the resistance value was based off a real number, and not some Percent deal. If you have 50 resist, you took 50 less damage. If you had 48 resist, you took 2 damage more than someone that had max.

Again, I haven't really looked at it lately, so dunno if it's different now, however it seems to be a percent based formula. Now however, I just tested on a Level 4 Mob, with my cl 17 OL 62 TE and looked at the damages and resistance changes.

 

Original shots they hit me for were 7-10's for the highs, and 2-6 on lows. I then hit my Rally skill, and saw very similar numbers, except a (-1) or (-2) that my resistances protected me from.  I have level 4 rally, and thus have a value of 20 to all my resistances.

 

I also did not fire a shot, and they agrod me. They saw me, targeted, and let loose their arsenal. I remember that after a certain level over the top of a mob, even "Agro" type mobs would not agro until you did damage or used a skill against them. Tie into this my old theme, it also effected the range in which "Agro" type mobs would agro you even with the same signature on your ship. Each mob had a certain site range, and it's ability to see you would get affected by your level (don't remember if it was combat level or overall level). So say you were CL 40, an OL 50 mob would "see" you from about 12k away. But when you were cl 50, that range would reduce to 4k on some, 6k on others, up to a TE's max range of 8.5k for others (never seemed to be more than that, on any mob in the game including the fishbowl).

 

Well, that was more of a reply than I had originally intended, I'm a bit famous for my "walls of text" in the past, so have the habbit of hitting return more.

 

Have fun and fly safe. o7

Edited February 1, 2013 by Werewolf

[Mattsacre]

I don't 'member as a player taking 0 becuase of resists, I do remember mobs that took 0 however for certain dmg types, like the mobs on arduinne planet that actually healed from plasma and a named chavez that took 0 from plasma. Onlt 0's as a player we took was becuase of actual miss. The resists had a absolute complete 1 for 1 resist up to 50 and then a deminished return over 50. I.E. at 50 you resisted 50% of a certain dmg type, any resist you got over 50 was halved so you had to stack 60 to get 55%. If I remember right, that made you only be with all the right buffs and gear top out at true 72% resist.

 

I do remember the agro envelope (the range at which a mob would autoattack) did decrease the more you were over thier OL compared to yours. Mobs would be less and less likely to entiate combat with you, without hostile action on your part the more over thier level you gained, this was great for miners etc. to avoid the "annoyance" factor of low lvl mobs when mining for build ores etc. This did preclude factioned mobs however, their scan range was their agro range when they hated you as a faction. They would persist in annoying if there was a faction hate unless they were a mob with a escape AI like chavez that ran when you killed them to a certain level or some of the player class npc.

 

I too, remember the crits had a higher occurance when you were a sugnificant level over the mobs, as you explain it, it sounds logical but never did parse the numbers to see. Seems also that if that was that way the miss skill/resist would modify as it should to mob OL, I remember with my PS if I hit menace on a mob -10 level to me I almost never missed, and mobs at my level there was about a 60% hit rate with a miss rate dropping 10% per level approx. with a no less than 10% hit rate no matter how more it was over me. Example: my PS without his menace buffer seemed to have a 20% hit rate on 55-56 volts, and never had a better than 10% on the L60 volt, when I plugged in the menace buff the 58s seemed to have a 20% hit rate.

 

When I had my JE guilder along, he would fit a PSI debuffer and a chimera and I'd forgo the basalisk/chem ammo and make it plasma, with his psi debuff, my manace chances went to about 50% as well as the grav link, we combined, plowed through those buggers much faster than many full groups could, and we would do that, especially if fellow guilders were having problems with KSers, some of them would be stunned that a PS/JE combo could KS them back.