Accuracy/CE parse

Discussion in 'Melee' started by Sirene_Fippy, May 10, 2016.

  1. Sirene_Fippy Okayest Bard

    Parsed difference in +29 Accuracy on Test vs Live servers (1139 HDEX):

    [IMG]

    Here's the full gamparse thing:
    [IMG]

    [IMG]

    Overall it looks like a small DPS increase. Accuracy increased by 0.67% for +29 accuracy, so ~0.023% for 1 accuracy. CE is roughly 0.128% extra procs per 1 CE. Both very small. I wasn't expecting to get a patch with a big DPS increase; we will see these stats climb slowly as gear improves. Feels like a very secondary benefit of having HDEX though for sure.
    Maxxbuff1 likes this.
  2. Xianzu_Monk_Tunare Augur

    Umm isn't the accuracy cap 150?
  3. Sirene_Fippy Okayest Bard

    On Test the Accuracy cap is 300 :)
  4. Xianzu_Monk_Tunare Augur

    Ah, good to know.
  5. Brohg Augur

  6. Xianzu_Monk_Tunare Augur

  7. Brogett Augur

    Remember to consider the noise.

    I think if my stats recollection is correct, the standard deviation of a binary hit/miss chance is something like square_root((p*(1-p))/N) where p is the probability of hit or miss (23% here) and N is the total number of hit attempts. If that's correct, it looks like ~0.1% so a difference of > 0.3 is significant within a 99% confidence range.

    I think! It's been too long, but anyway do bear it in mind when parsing such minor things. This is why calculating crits was such a PITA! :)
    Lanthur and Sirene_Fippy like this.
  8. Lanthur New Member


    Best post regarding parses I've read on these forums. People post all these parses and say its a % gain, but if it isn't statistically significant, for all we know it looks like the developer just added fluff on the User Interface and there is no calculation being done server side.
  9. Sirene_Fippy Okayest Bard

    I've never taken statistics. I'm all ears if you (or anyone else) want to explain how to compute confidence intervals etc. for things in EQ.

    One general question I have for EQ statistics is how for an unknown value (i.e. miss rate) how many samples (attempts) do you need to determine the unknown value within a 1%, 0.5% or 0.1% margin of error.

    Also, for an example like the above, what is the margin of error. If what Brogett said is the correct method (same one for E in this post), it's 0.26%. So the result is the two parses have a differing miss rate of 0.667%, which is statistically significant.

    I do not feel comfortable enough with my statistics knowledge (google, wikipedia) to do stuff like that for my own parses. It frustrates me to hear you say that "People post all these parses and say its a % gain, but if it isn't statistically significant..." I've invested a not insignificant amount of time into putting information like this on the EQ forums for everyone's benefit. I try to provide good results by making extensive parses (this one was 10 hours, the Avoidance ones were 20 hours each, plus taking the time to type up, format, upload results and write a post). To me it is an assumption when seeing any parse that there is error, and it may not be a clear cut result. You (and anyone else) are invited to provide your own parse or analysis to add to the discussion.
    Brohg, Spellfire, Reht and 1 other person like this.
  10. iZealot Lorekeeper

    Small increase, almost negligible. If you are HSTR there doesn't seem to be a point in making the switch to HDEX. HDEX will raise your crit but lower your avg damage.
    Here are some parsing results I did. 420 point swings in HDEX and HSTR (+210 and -210) by switching out 10 Type 5 augs. These are both 8-9 hours parses.
    Though what's strange, I had more flurries on my HSTR parse.
    Had misses hidden /doh so I did not include accuracy %.
    Raid buffed.

    I'll do some more parsing on Night's Endless Terror and Roar of the Lion. if they significantly increase in proc rate/DPS due to increased HDEX, HDEX may be more worthwhile. But as it stands, since HDEX doesn't increase Thunderfoot or other procs, i'm still putting my eggs in the HSTR basket for now, DPS wise.

    [IMG]
    Sirene_Fippy likes this.
  11. iZealot Lorekeeper

    Dosen't seem like NET or Lion is increased by more HDEX, so nevermind on that.

    Though my parse listed above does have two errors.
    First, the red circle should say HSTR, not HDEX, but the data is the same, if that wasn't already obvious.

    Second, I had destructive force (DF) on BOTH parses, but for some reason, my HSTR parse increased flurries per minute more than HDEX. I wonder if I made a mistake or if for some reason DF increases more from HSTR???
    I also had Dicho 6, Zan Fi, & Infusion of Thunder on.
    None of these will be on next parse, just raid buffs.

    Anyways, will re-parse all this tonight and repost.