For Arkonen: here the first part of the analysis for the Cornundacauda dataset .
Due to the sample size I did set the focus on Mature and the combo of Prowler and Stalker, hence I’m going to show results only for those 2 groups.
Empty Rate:
For Prowler I did combine the results from Alpha, OA and Prowler. For Mature I did use also Old.
Code:
Maturity ER
Mature 0.57
Prowler 0.49
Stalker 0.43
Using this empty rates I did correct Loot according to Loot corrected = loot * (1-ER).
Edit: Please not, I have to redo the figures due to an error in empty correction. Data in tables should be correct.
Survival analysis according to maturity
[br]Click to enlarge[/br]
The survival function is clearly different between Mature and the combo of Prowler/Stalker (p<.001 Log-Rank). I used this combo because I’ve noticed that their survival functions do look more or less the same. Hence to have a better sample size I did combine those groups.
Model verification class C0 and C1.
I previously published loot classes for mobs with HP > 1000. One of the main conclusion was, that base loot is more or less HP/20 + HP/10 PEC. Now let’s verify if this is also valid for Cornundacauda.
Base Loot Mature, Stalker/Prowler
[br]Click to enlarge[/br]
The observed mean for Mature is 18.17 PEC and 55.82 for Prowler/Stalker. The exponential distribution does not fit perfectly but approximates base loot quite well in terms of means. I've already analyzed this in detail in my loot analysis thread. The main conclusion was, instead of using single classes with respective mean and weight the exp. dist. can be used to estimate the mean for base loot.
What I’ve further shown in my loot analysis thread was, that those means are related to hp dmg done.
Code:
mean mean/HP min min/HP
mature 18.17 0.12 7.70 0.051
st/prow 38.03 0.09 20.11 0.050
Results from Stalker/Prowler do correspond quite well to the assumption that min loot is HP/20 and mean for C1 is HP/10. The observed factors are .05 vs .05 and .09 vs .1.
Matures do behave similarly with a mean/hp of .12 instead of .1. It could be that we have to assume HP*3/25 instead of HP/10 on lower HP mobs but I have my doubts about that.
Loot above C1 has a freq. of about 5% for Matures and 5.1% for Stalker/Prowler. The error margin is high with this sample size on those frequencies but interestingly I do always get those 5% even when the sample size is small.
Here now the right tail of the loot distribution.
[br]Click to enlarge[/br]
I've used a GPD (generalized pareto distribution) to model it due to the small number of cases.
It tuns out that mean loot above C1 for mature is 645 PEC and 2495 PEC for Stalker/Prowler. Now apply the weight of 5% and you'll get the expected mean loot.
If I assume HP/20+.95*HP/10+.05*645 for Mature then I'll get an expected mean loot of 54 PEC.
For Stalker I'll get HP/20+.95*HP/10+.05*2495 = 188.5 PEC.
Please note, those are preliminary results, n=10 on mature and n=12 on stalker, and I'm sure the mean of about 2500 is to high for stalkers, but you can already apply you're expected cost to estimate your return rate.
If I apply my loot model validated for mobs > 1000 HP I'll get for prowler/stalker a mean loot of 178PEC (empty corrected, hence the observed mean is 310 PEC) and a mean of 62.6 (uncorrected 146) for Mature.
If I'll take your mean cost of 182 for Stalker/Prowler and 69 PEC for Mature per kill, I'll get a return rate of 98% and 91% respectively. That's quite a good result I would say.
Your guess about 65% of the loot comes form the higher loot class is correct as you can easily verify.
All in all it seems that Cornundacauda with a hp >=150 do behave as those above 1000 HP.