Latest Diablo 3 News
DiabloWiki Updates
Support the site! Become a Diablo: IncGamers PAL - Remove ads and more!

counting your mf runs

Discussion in 'Single Player Forum' started by Tolano, Apr 11, 2005.

  1. Tolano

    Tolano IncGamers Member

    Joined:
    Aug 10, 2004
    Messages:
    51
    Likes Received:
    0
    Trophy Points:
    26
    counting your mf runs

    i noticed that alot of people like to do experiments and stuff with drop calculations that involve counting the amount of runs they do, i have seen the numbers be as high as 1500 runs. i was just curious as to how people count them? is there a program? or possibly they do a set number of runs per day and they write them down? if there is a program that does this could anybody point me in its direction? if not i guess ill go and buy a notebook, lol.
     
  2. nepeta

    nepeta IncGamers Member

    Joined:
    Sep 19, 2004
    Messages:
    3,829
    Likes Received:
    46
    Trophy Points:
    258
    ... don't forget the pencil ;)

    (good luck)
     
  3. Chimaira

    Chimaira IncGamers Member

    Joined:
    Feb 28, 2005
    Messages:
    1,709
    Likes Received:
    0
    Trophy Points:
    120
    You just have to do it manually...I did 100 Mephisto runs once and noted everything down, got quite boring in the end, but the result corresponded almost 100% with the amount of MF I had on...unsurprisingly. It would be cool if there was some kind of tracker in Diablo II where you could go and check out your statistics!
     
  4. Artagas

    Artagas IncGamers Member

    Joined:
    Apr 3, 2005
    Messages:
    472
    Likes Received:
    0
    Trophy Points:
    143
    yeah, sometimes i wish for a "number of xyz killed" list as well (a simple asset of several crpgs), but than on second thought at times it could be very frustrating. (make you say "geez 278 mephistos killed and not a proper item yet!". nah, i don't think i want that. better not count and keep thinking it was only 78 or so)
     
  5. Cormallon

    Cormallon IncGamers Member

    Joined:
    Oct 11, 2004
    Messages:
    1,704
    Likes Received:
    0
    Trophy Points:
    466
    It's pencil and paper. One time I used a precounted number of small items (coins for example) I put from one box into another, and when the first box was empty I had that number of runs done.

    I really much would like some kill statistics in a potential D3 though. I've seen that in other games (Wizardry?) and such a feature would come very handy doing these 30 sec fast runs.
     
  6. Nightfish

    Nightfish IncGamers Member

    Joined:
    Aug 14, 2003
    Messages:
    3,909
    Likes Received:
    25
    Trophy Points:
    347
    I remember back when 1.10s was fresh and everybody believed MF was no longer working on act bosses I did a lot of runs to test whether this was true or not. (it wasn't) I used to do it both with pen & paper and later on with excel. Both good but if you do it with excel you can do statistics a lot easier since you've already got the data typed in.

    One thing you need to watch out for is what you write down. You need to consider failed items and the dilema of blue and yellow bows / Xbows. If you're not careful there you can mess up your stats before you even start counting. (This is only important if you want to compare things, of course. Like, meph vs baal, ws k vs pit, etc).

    If you want to do some statistics you don't have to have 1500 runs, though. 30 is generally the lower limit for sample size.
     
  7. tenaka

    tenaka IncGamers Member

    Joined:
    Jun 22, 2003
    Messages:
    506
    Likes Received:
    0
    Trophy Points:
    120
    Well there is a counter-argument to this. Lets say I did account for failed items and didn't have any bows to worry about. If my results were 100 rolled uniques and 90 actual uniques after X runs, and this data was consistant with other people doing X runs, then one could conclude that when running HelloIAmABossPeopleLikeToMFOn one in ten uniques rolled will fail. What really matters is what you receive not what is rolled. Whether you get 90 true and 10 failed or 90 true and 1000 failed at the end of the day you get 90. When people look at what is the best place they look for that 90 and compare it elsewhere so tracking failures give you interesting data but it won't make a place look any better or worse for MFing.
     
  8. Nightfish

    Nightfish IncGamers Member

    Joined:
    Aug 14, 2003
    Messages:
    3,909
    Likes Received:
    25
    Trophy Points:
    347
    Yes, that is true, but *not* if you want to compare things. If you don't pay attention to failed uniques and sets you might arrive at wrong conclusions. If you think about it a little I'm sure you'll see why that is so.

    If you only compare real uniques and sets you'll be comparing numbers small numbers which are easier influenced by chance. Kinda like if you flip a coin 5 times you might well arrive at the conclusion that the chance for tails is actually 0%. It's not quite the same but it's similar. Fact is that your numbers will be wrong. Doesn't really matter much but if you want to really compare things you have to do it correctly. (trust me, I R TEH SCIENTIST!!!1111one)
     
  9. Chimaira

    Chimaira IncGamers Member

    Joined:
    Feb 28, 2005
    Messages:
    1,709
    Likes Received:
    0
    Trophy Points:
    120
    I agree with Tenaka...It won't do you much good to know how many failed uniques you get, cause statistically you'll still get the same amount of failed uniques the next time you do 1000 runs..so other than interesting data I can't see the appeal of noting them down. I'm not completely sure I understand your example...perhaps you can enlighten me, since your the scientist:)
     
  10. Nightfish

    Nightfish IncGamers Member

    Joined:
    Aug 14, 2003
    Messages:
    3,909
    Likes Received:
    25
    Trophy Points:
    347
    Well, if you go into depths about statisitics and stuff I could sit here all day. THe short version is that if you do not count failed uniques you need a much larger sample size to get good results when comparing one MF target to another. That's because you need to cancel out the mistake you're making in gathering your data with an increased sample size.

    I'll try to give a simplefied example: Let's say you do 10 on two targets which gives you one item each.

    Target A)
    1) Unique
    2) Set
    3) Nothing
    4) Nothing
    5) Failed Unique
    6) Failed Set
    7) Failed Set
    8) Failed Set
    9) Failed Unique
    10) Failed Set

    Target B)
    1) Nothing
    2) Set
    3) Nothing
    4) Nothing
    5) Unique
    6) Set
    7) Set
    8) Failed Set
    9) Nothing
    10) Failed Set


    So, if I count this I have
    Target A) 3 Uniques, 5 Sets
    Target B) 1 Unique, 5 Sets

    If Tenaka counts:
    Target A) 1 Uniques, 1 Set
    Target B) 1 Unique, 3 Sets

    So, in this case we both arrive at different results. I'd favor target A while tenaka would go with target B.

    Now please don't go and nitpick about the small sample size in this example and stuff. It's just to make a point. If you don't count failed items you are biasing your results which isn't really good. Or rather, you're adding another factor to the equation. (item selection within treasure classes. That's what determines failed items.)

    The bottom line is that you get much better results if you count everything. You don't feel cheated by bosses as easily, too. And it takes you a lot less runs to determine what is the best target to run. Yes, in the long run it'll even out, but if you are taking the trouble of gathering statistics you might as well do it correctly. And if you deliberately ignore information that's definetly not accurate... Maybe I'm a bit more anal than I need to be on this but this is part of what I study after all.
     
  11. tenaka

    tenaka IncGamers Member

    Joined:
    Jun 22, 2003
    Messages:
    506
    Likes Received:
    0
    Trophy Points:
    120
    I completely agree with your argument Nightfish, you're just approaching it from a different angle as me. I qualified mine with the statement
    which should weed out that erronious data by increasing the sample size with a combination of information from many people. In order to get a good comparasin on your own in a relativly short ammount of time your way is best. My contention is just that a lot of people lately have been posting their findings on runs lately so one can incorporate that data into their own and come up with a significant result without the extra tracking.
     
  12. Nightfish

    Nightfish IncGamers Member

    Joined:
    Aug 14, 2003
    Messages:
    3,909
    Likes Received:
    25
    Trophy Points:
    347
    I don't know enough about statistics to tell you how bad your results are if you leave failed items out. I can tell you that if I did something like that in my diploma thesis I'd get flayed alive, though.
     
  13. killian27

    killian27 IncGamers Member

    Joined:
    Jan 27, 2005
    Messages:
    520
    Likes Received:
    0
    Trophy Points:
    165
    "Do, or do not. There is no try." -Yoda


    A failed unique tried and failed, therefore it is a do not.
     
  14. Chimaira

    Chimaira IncGamers Member

    Joined:
    Feb 28, 2005
    Messages:
    1,709
    Likes Received:
    0
    Trophy Points:
    120
    Thanks for the explanation Nightfish:) I still think that if you do enough runs the discrepancies will be so minor that it won't be worth the extra time needed to inspect your finds more carefully. I'm still not quite sure what the difference is between counting failed uniques or not...Say you did an infinite amounts of runs...then it wouldn't matter if you counted failed uniques...or am I totally mistaken. Since an infinite amount of runs would be the "definitíve" result. Also if you did a smaller amount of runs...lets say 500, Baal runs...Then you'd find maybe 50 failed uniques, then if you did another wouldn't you find approximately 50 failed uniques again...if that is the case I don't understand why it's so important to count failed uniques. Sorry for being dense :(
     
  15. tenaka

    tenaka IncGamers Member

    Joined:
    Jun 22, 2003
    Messages:
    506
    Likes Received:
    0
    Trophy Points:
    120
    Ahh but if we wanted to be truly accurate then do we not need to investigate why Target A has more fails than target B? Perhaps A has a higher likelyhood of base items who's unique/set is of a TC it cannot drop. If you are going to account for everything then you must account for everything. It may be the case that one area/monster will always have more failures than another and not accounting for that reason could give you false results. I have no idea what people look for in a thesis but I know if I went to my boss and told him A had more faulires than B he would tell me I'm not going home until I found out why. If I told him B had more sets/uniques than A he would say, Ok lets use B. As you can see I'm not a perfectionist. I just need a ballpark to aim for and I'll take it from there.

    (edit) @Chimaira: Nightfish's point is that the reason you get a failure is that a base item drops whose unique/set is not of a TC that monster can drop. If different base items drop you will see more or less failures. Over a smaller sample the number of failures is inconsistant. His way is definitly not more innacurate but I am agreeing with you that it is not necessary to have a decent idea of what you will see. To use for a calculator or term paper sure, but to look at to know I want to run A over B, no.
     
  16. Nightfish

    Nightfish IncGamers Member

    Joined:
    Aug 14, 2003
    Messages:
    3,909
    Likes Received:
    25
    Trophy Points:
    347
    I'll withdraw from the discussion at this point. :) You do it your way, I do it my way. :D
     
  17. Sint Nikolaas

    Sint Nikolaas IncGamers Member

    Joined:
    Apr 9, 2004
    Messages:
    3,937
    Likes Received:
    8
    Trophy Points:
    256
    I just run untill:
    1- I'm out of time (can be set or can be because else I'm late for something)
    2- My stash/cube are full.. when I mule my stuff off that's allmost allways the same time I stop :)

    I once wanted to count the # of pindleruns needed to level from 92-93.. needlessly to say I lost count at about 70 orso and I still had a long way to go :D
     
  18. Mage11

    Mage11 IncGamers Member

    Joined:
    Jun 22, 2003
    Messages:
    721
    Likes Received:
    0
    Trophy Points:
    105
    I was counting & recording a lot in 1.09. Then came 1.10, and I couldn't be bothered any more.

    However, I would say that both Nightfish and Chimaira have valid points. If you can do enough runs (and 30 certainly isn't enough for this sort of thing, particularly if the targets have rates that are fairly close), then it is reasonable to only count the actual uniques and sets if you are trying to determine who will provide more uniques and sets.

    If you want to determine who is more likely to get a unique or set hit, then you will want to include your + duration rares & magics, and this can provide you a quicker result.

    However, if you consider that one target may have a higher likelihood of a failed unique or set (due to TC - either there is no such unique/set item or the target cannot drop the unique/set item) than the other, and this is something you want to account for in your counting, then you would not want to include failed items in your counting.

    Two different methods, counting probabilities of two different things. If you understand the difference, then you can use them more precisely. Neither is wrong, but each is different in what it is actually doing. To be substantially accurate (and, in fact, probably to be able to differentiate between the methods if they are being used to compare targets), you would need to do quite a number of runs. And no, I'm not going to prefer a number. That would take more thinking than I'm prepared to do at the moment.
     

Share This Page