Multi-hit WS accuracy testing

[CDF]

Actually, there is enough to show that the number of observed secondary hits is significantly less than expected, statistically speaking.

Obviously, the expected number of secondary hits over 100 4-hit weapon skills (no double attack) is 150, assuming 50% hit rate. The corresponding variance is 75 (the variance of the sum of the number of 2° hits of Guillotine is the sum of the variances).

The upper bound for the (approximate) 95% one-sided confidence interval for the actual number of 2° hits is 143. The (approximate) p-value is .00656. So, you could conclude that Guillotine has an inherent accuracy penalty or the actual hit rate was lower than 50%.

[Nameless]

Very interesting. Now if we do a test under the exact same setting but for Insurgency, I would like to see the secondary hits being close to 50%. If that is the case, then it is clear that guillotine has accuracy penalty and should be taken into account when gearing for it (go go 25 ws acc helm).

[Siniroth]

I still think we need a larger sample size than 100 Guillotines

[Starr]

We need over 9000 for every test we do

[CDF]

100 Guillotines is equivalent to 300 secondary attempted attacks. Is the bigger number more persuasive to you?

It's possible that 129 observed hits was just an "unlucky" result, but based on this sample alone it's reasonable to think that something might be going on other than random variability.

(Note: the actual p-value is .00889, not .00656 as incorrectly stated earlier.)

[Yugl]

Does anyone know what the ACC Penalty on Penta was before SE fixed it? Is it possible that some WSs have a base ACC?

[Kirschy]

Not only that, do we even know that the 2nd to 4th hit of drakesbane or raging rush get the critical hit bonus from tp or not. I am pretty sure they can critical, just don't know if they get the bonus critical. Also, are we sure for those WS that has ftp based on tp like KJ that the 2nd and 3rd hit are just 1.0 ftp instead of being the same as the first hit?

I think the test we have had used to determine that all following hits are 1.0 ftp is on single hit WS, and it could just be that any addition hits from DA are 1.0 ftp, but not necessarily the additional hits from multi hit ws. It would be interesting to see and can probably explain why KJ is so powerful, since at 200% it is 1.25 ftp, and if all hits follow that ftp, it could do surprising damage, which has been reported.

And if the theory is wrong, that only the first hit's ftp is based on tp, then it would be interesting to see if the critical hit ws work the same way. I am betting that the 10% critical hit % only apply to the first hit while the rest of the hits critical rate is only depended on your natural critical hit rate.

How we gonna test this I don't know....

When the 2H patch was first introduced, normal hits capped at 3.0 pdif every time. It made testing very easy. I went to qufim and began doing tests on guillotine against the crabs there. They had ~500 HP or something, so only the first two hits of Guillotine would land, and since each hit was a 3.0, I could ascertian the base damage easily. fTP was only applied to first hit and not subsequent hits. (This is true for a 4-hit WS like Insurgency where it says damage affected by TP. The extra damage from TP only applies to the first hit, and not the remaining ones.) But the secondary mods work on each hit.

Criticals can be tested in a similar manner even now by SAing a multihit critical on a mob with low HP. As long as only the first two hits land, and we know the first is a critical, it will be easy to tell if the second was a critical.

[Vodou]

Actually, there is enough to show that the number of observed secondary hits is significantly less than expected, statistically speaking.

Obviously, the expected number of secondary hits over 100 4-hit weapon skills (no double attack) is 150, assuming 50% hit rate. The corresponding variance is 75 (the variance of the sum of the number of 2° hits of Guillotine is the sum of the variances).

The upper bound for the (approximate) 95% one-sided confidence interval for the actual number of 2° hits is 143. The (approximate) p-value is .00656. So, you could conclude that Guillotine has an inherent accuracy penalty or the actual hit rate was lower than 50%.

This is true, but if and only if the measurements have no errors.

On the surface, I think its a pretty solid effort.


1) Wiki says these mobs are thf and have evasion boost. How was that handled or is wiki wrong?

2) How exactly was the count determined for the swings after the first one? Is there any room for error in the methodology?


Before I accept that there is an acc penalty on multi hit WS, I would want to know at least this much. If there was no evasion boost and no errors, then yes, statistically, it looks like there is a penalty.

I certainly applaud the attempt. We've discussed this issue for a while, and I am glad to see someone step up to prove it.

[Nameless]

Yea, I was thinking about Insurgency too, if all hits follow first hit on ftp when damage is based on tp, then it would have .5 for all hits, which would make it parse even lower than it has been. However, if only the first hit benefit from tp bonus, then I would assume that applies to critical ws too. Would be something to test.

[Fwahm]

This is true, but if and only if the measurements have no errors.

On the surface, I think its a pretty solid effort.


1) Wiki says these mobs are thf and have evasion boost. How was that handled or is wiki wrong?

2) How exactly was the count determined for the swings after the first one? Is there any room for error in the methodology?


Before I accept that there is an acc penalty on multi hit WS, I would want to know at least this much. If there was no evasion boost and no errors, then yes, statistically, it looks like there is a penalty.

I certainly applaud the attempt. We've discussed this issue for a while, and I am glad to see someone step up to prove it.

1) Evasion boost directly impacts their evasion stat, so it directly impacts the /check method for determining an opponent's evasion, and thus is accounted for.

2) I checked the TP for my WS after each hit.

Primary hit only: 20 TP
Primary + 1 secondary: 21 TP
Primary + 2 secondary: 22 TP
Primary + 3 secondary: 23 TP
1 secondary: 1 TP
2 secondary: 2 TP
3 secondary: 3 TP

After every weapon skill i manually inputted the results of the Weaponskill in word, saving every few Weaponskills.

For example, if I did a 101th Guillotine, and got 22 TP, I would change the fractions to 92/101 for Primary, and 131/303 for Secondary.

[konoko]

me and a dnc friend did some testing (me proving to him that DE was stronger then Pyrrhic)

i didnt keep the actual data, but it was pretty obvious that for dnc, DE gets 1.1875 ftp on the first hit only, and Pyyrhic gets 1.5 on the first hit only, if every hit could get the 1.1875 ftp then DE would be the weaker of the 2 ws's

if all hits got the ftp boost then

PK:1.5 x 4 = 6.0

DE:1.1875 x 5 = 5.9375

DE comes out slightly behind as far as FTP is concerned

if the following hits only give 1.0 ftp however

PK: 1.5 + 1.0 x 3 = 4.5

DE: 1.1875 + 1.0 x 4 = 5.1875

DE comes out ahead by a much larger margin (which our testing in game seemed to agree with)

[Vodou]

1) Evasion boost directly impacts their evasion stat, so it directly impacts the /check method for determining an opponent's evasion, and thus is accounted for.

I'm sorry because I posted my comment that in a confused way.

I was meaning the evasion boost from warm-up, not the evasion bonus that mamools get as a trait. I apologize if not all mamools use it (maybe blus do and thfs don't for example), but I barely xp, so I am asking if that's an issue, not suggesting it is.

If they can't use warm-up, then that also answers that question!

As a follow-up, since I'm not clear. How did you handle the issue with tp being truncated in its presentation form in the game? E.g., if it says 53 on the screen, it actually could be anything from 53.0 to 53.9. Again, I'm sorry if it seems like I am challenging you. I admire your effort and am only asking because your results could be very accurate, but they are somewhat of a surprise.

[Fwahm]

I'm sorry because I posted my comment that in a confused way.

I was meaning the evasion boost from warm-up, not the evasion bonus that mamools get as a trait. I apologize if not all mamools use it (maybe blus do and thfs don't for example), but I barely xp, so I am asking if that's an issue, not suggesting it is.

If they can't use warm-up, then that also answers that question!

As a follow-up, since I'm not clear. How did you handle the issue with tp being truncated in its presentation form in the game? E.g., if it says 53 on the screen, it actually could be anything from 53.0 to 53.9. Again, I'm sorry if it seems like I am challenging you. I admire your effort and am only asking because your results could be very accurate, but they are somewhat of a surprise.

1) Due to my testing procedures, the Mamool Ja Bounder didn't use a single tp move in the entire test.

2) I'll copy the TP results, this time without truncation, so you can see they don't affect anything.

Primary hit only: 20.50 TP
Primary + 1 secondary: 21.65 TP
Primary + 2 secondary: 22.80 TP
Primary + 3 secondary: 23.95 TP
1 secondary: 1.15 TP
2 secondary: 2.30 TP
3 secondary: 3.45 TP

[Nameless]

I think the next thing is to test another multi hit ws. Cuz guillotine is actually the one SE said they gave acc penalty, I am not sure if it is completely undo by SE, but will be good to test if penalty applies to all multi hit ws.

[Starr]

Can't use Penta Thrust/Dancing Edge etc because of Acc bonus and I think it would be better to parse a WS with less hits like RR/KJ/Impulse Drive/Skewer.

[Ned]

Window forum link to TP returner

The post give you the basics and if you know what you are doing with spellcast you can use that to /echo whatever you want.

I use a variation of the code IneptVagrant posted, but I put a set of rules to check which weapon (so it doesn't give stupid numbers when I use a sword instead of a dagger).

OK, now I read it again it does make sense. I am not sure how I missed that before.

That TP autoexec code is gold - anyone know where i can find a base TP return list for all WS so can butcher it for other weaponskills - can't see anything on wiki?

[Ned]

That TP autoexec code is gold - anyone know where i can find a base TP return list for all WS so can butcher it for other weaponskills - can't see anything on wiki?

nvm - "TP" page on wiki shows normal TP return per hit.. i assume all WS gain TP depending on delya and then +1 for additional hits/ double attack? is this correct?

Was hoping for a list by weapon of normal WS tp return but as they vary so much i guess not possible.

Edit.. again.. does duel wield effect TP return on WS?

[Max™]

I did hours and hours of testing on Raging Rush basically naked on a static target (fort in garlaige citadel).

Over something like a thousand WSes, I got something like 80% accuracy on my hits other than the first with no neck piece, and 85% accuracy with gorget.

This lined up with a few things.

The TP bonus on critical hits seemed to sit at 0% at 100% and scale up slowly, maybe 1% per 10% TP, and the JP fast blade testing which showed the same amount of accuracy.

I threw out the first hit TP returns because I had already suspected the first hit got a huge accuracy boost after talking to Ruke about my freakishly accurate steel cyclones in my 360~ WS Acc gearset.


Summary: pchan is wrong.

[Starr]

Guess you don't have any parses of any of this data about do you?

[Max™]

Silence, dog, I have data which is easily reproduced by anyone who wishes to examine it, which fits the tests others have shown. It is simply a supporting example without better evidence.

Go play at being awesome elsewhere.