The error (standard deviation) in such counting method is the square root of the number. If you provide me with the raw data I can calculate how much better that nova sorc if you want.
2024 Spring RFL Sign-Up and Running Thread: Round 2 (Moo Moo Farm)
- Thread starter PhineasB
- Start date
The error (standard deviation) in such counting method is the square root of the number. If you provide me with the raw data I can calculate how much better that nova sorc if you want.
- May 11, 2022
- 482
- 812
- 93
I think I might be topped out at 12 hours of Cows with two qualifiers. Vang started the Random Tournament and I'm having more fun with that than Cows 
Maybe I'll fit one or two more hours in before Sunday...
I still consider 12 hours and any qualifiers to be a small victory for me. I normally can't get into farming Cows that much. It turns out that I can tolerate Cows with the right character.
Maybe I'll fit one or two more hours in before Sunday...I still consider 12 hours and any qualifiers to be a small victory for me. I normally can't get into farming Cows that much. It turns out that I can tolerate Cows with the right character.
- Oct 2, 2010
- 895
- 370
- 63
I'd be interested in hearing your thoughts about the data and what would be a sufficiently large number of drops. I'd hope 20 hour sets produce enough drops to extrapolate the number of cows killed somewhat reliably.
I used these Hell Bovine drop odds for each Jewel, SC, or GC:
P7 - 1:506
P5 - 1:571
Some runs are done in P5, so I included P5 odds here. I use the P5/P7 ratio to convert to the equivalent P7 run efficiency.
SC, GC, Jewel quantities, listed in the same order:
P7 Nova Sorc (2024 RFL) - 424, 430, 437
P7 Nova Sorc (2023 RFL) - 415, 436, 425
P7 Mosaic Sin (2023 RFL) - 334, 371, 325
P7 1.14d Necro (2022 RFL) - 164, 181, 176 <- 10 hour set
P5 1.14d Necro (2022 RFL) - 365, 320, 329
P5 1.14d Necro (2021 RFL) - 314, 331, 299
P5 1.14d Nova Sorc (2020 RFL) - 337, 317, 336
All 20 hour sets except for that one 10 hour set.
- Jan 1, 2011
- 703
- 1,240
- 93
Well no way I'll finish the second set, but I'm hopeful for getting to 5 qualifiers again. Just shy of 5 hours and I've got 2 qualifiers.
I've got to say that I'm really enjoying running Cows. Way more fun than running LK.
I've got to say that I'm really enjoying running Cows. Way more fun than running LK.
- Feb 24, 2004
- 6,390
- 1,341
- 113
Found it!
All set items acquired!
Edit: At least all set items for the cow level...looks like I'm still looking for a green caddy.
All set items acquired!
Edit: At least all set items for the cow level...looks like I'm still looking for a green caddy.
Last edited:
Simplest way, I think, to deal with "inverse" problems, is to simulate them. Here I simulated 1k 20h sets for the 2024 Nova sorc.
As we can see, rhoughly 60% chance that 178-187 cows were killed each minute, and ~85% chance that the number of killed cows/min was 175-190. Mean =181.6 , StandardDevitaion = 5.1
Here is 2023 Mosaic Sin
Mean = 144.9 , StandardDeviation = 4.7
Very clear separation.
What is interesting is that there is a factor 1.25 in the mean values between these two builds, but the standard deviations is only a factor 1.09 ...
Perhaps I should give this task to one of my math students next year for their diploma work...
Last edited:
I hate them, also hate travincal. I can not cope with item cluttering. It is 2024 and we still don't have loot filters
Close to 13 hours. One more qualifier, still nothing impressive so far.
Died once too... I know it's trivial but it pisses me off losing the gold.
Died once too... I know it's trivial but it pisses me off losing the gold.
- Jan 1, 2011
- 703
- 1,240
- 93
@Excalibur Well at least someone thinks I'm reasonable. My wife also questions my sanity! The difference between now and years past is that my best characters did no area damage, so it was one cow at a time.
@drmalawi There should be all sorts of math projects available using Diablo. The clutter does get horrible sometimes. In D2R I usually have show items on, but have to turn it sometimes so I can see the cows!
@drmalawi There should be all sorts of math projects available using Diablo. The clutter does get horrible sometimes. In D2R I usually have show items on, but have to turn it sometimes so I can see the cows!
- Oct 2, 2010
- 895
- 370
- 63
@drmalawi I suspected simulations were the way to go, and thank you for running the simulations! For the purposes of comparing build efficiency, around how many standard deviations apart would you consider sufficient separation? For example, the mean between my 2024 Sorc and 2023 Sorc is so close together (difference of only 2), I wouldn't be able to confidently say the 2024 Sorc performed better.
It depends on what confidence you wanna say that build X is more efficient than build Y.
In my working field, elementary particle physics, you want 5 standard deviations apart in order to claim a "discovery" - which is like 1 in 3.5 millions. In many other fields of science, one or two standard deviations apart is enough to make such claim.
Yeah that is impossible to judge by the charm+jool counting method.
I know I made a program that simulates RFL scores in 2018. Perhaps I could merge that into these simulations and make estimates for "if you found X number of sc+gc+jools, what is the distribution of RFL scores?".
Woke up early, got a nice qualifier. Currently at 14 hours with 11 qualifiers but far away from a winning score.
Heyho, despite my best intentions I ended up not running after all this year, with RL being busy and PoE taking up what's left of my spare time. 
But wanted to quickly drop in to say it's great to see everybody run and still more theorycrafting being done! I look forward to seeing the results.
But wanted to quickly drop in to say it's great to see everybody run and still more theorycrafting being done! I look forward to seeing the results.
- Jan 1, 2011
- 703
- 1,240
- 93
8 1/2 hours into set 2 and I've got another qualifier (now 3) for this set. Should get over 12 hours by Sunday night and hopefully will get to 5 qualifiers. Also found a Taebek's Glory and a Wraith Flight tonight. The Cow King even gave up an Oculus .
No idea how I'll rank, but very happy with my finds so far.
No idea how I'll rank, but very happy with my finds so far.
- Oct 2, 2010
- 895
- 370
- 63
Thanks for the advice, I'll keep that 1-2 SD in mind. The 10k simulation you posted looks good. SD around 5.0 cows/min isn't too bad, but then again it involves a data set that took 20 hours to collect.
I think I'm the only one diligently collecting and bookkeeping those charms and jewels though. I forgot to mention the other "error" that occurs is missing charm/jewel drops, so the real mean should be slightly higher.
In the 2022 RFL, I used the kill count feature of the MF Run Counter to get the real kill count (D2 LoD only feature). I think the P5 Necro kill count was about 5% higher than the drops method count, but of course there is deviation, so it's hard to say what the true under count is. Necro tends to kill in larger piles, so I expect my other builds to not be as bad in this regard.
Couldn't one just do exp counting? But that leaves out doing any TZ while running.
Collecting and book keeping charms and jools takes some time as well. If it is just a pure rune hunter that is.
Collecting and book keeping charms and jools takes some time as well. If it is just a pure rune hunter that is.
- Jan 1, 2011
- 703
- 1,240
- 93
- Oct 2, 2010
- 895
- 370
- 63
@drmalawi Other runners have used the exp. method before, and I think I tried it long ago for some quick data. I use the drop method now because I collect charms and jewels anyways, regardless of whether I need the data or not. The hardest part of exp counting was accurately estimating the number of champions/minions/unique monsters killed, as those monsters offer 3x or 5x the experience. Some builds may tend to clear boss packs more than others. Runners will also need a level 94+ character so that the exp. penalty is the same for both normal cows and boss packs. There is already a SPF acceptable kill counter for D2LoD, but I would love one for D2R.
Google