I know these topics come up from time to time and are never productive but here is another.
So I had noticed or at least it felt as if I was consistently getting bad luck. I would play with my friend and it seemed like he was consistently getting good luck. At first there was really no way to tell but then bobs buddy was released. I started recording the result predictions and actual results in a spreadsheet just to see if I was actually experiencing bad luck or it was some sort of confirmation bias.
I have recorded 2496 matches as of now. I separated the data into 3 groups. The first set of data is the fights that have expected outcomes that match the actual outcome. This data is basically not interesting to me as it is just confirmation that what was expected to happen did happen.
The next group is the fights that the most likely outcome did not happen. To make it simple I just look at whether I am supposed to take damage or not. So for example 40%W 20%F 40%L should favor me but the outcome is I lose and take damage. Or 20%W 20%D 60%L should not favor me but I win or draw and donât take damage. or even 33%W 33%D 33%L but I lose and take damage. When I look at this group of data I find that in 82% of the fights the outcome was bad for me as opposed to good. This should approach a 50/50 good/bad luck for me but it isnât even close itâs instead 82/18 against me.
The 3rd group was the coin tosses. So for example 30%W 20%D 50%L or 50%W 0%D 50%L. In these fights I was also under-performing. 73% of these fights I lost the coin toss instead of it being well 50/50.
I would be curious if anyone has a better way look at the data or if anyone else has been collecting it.
Conclusion⌠Either Bobs buddy is very wrong. I am analyzing the data wrong, my sample size is just too small, or blizzards game has some fun RNG biases.
Why do you think theyâd care which of the eight of you wins?
Why do you think theyâd care enough to design a complex, multi-faceted system affecting many different variables to nudge the odds just enough to actually accomplish meaningful changes in winrates but also not enough to be readily detectable, instead of spending their limited development time doing things like fixing actual bugs and designing new features?
2496 is alot of games to play when you think it is rigged. I would really love to play you at poker, rig the games, and have you play 2496 hands with me anyways.
I think you should upload some replays so we can see how you are going wrong because game is not rigged. Admit that it could be your play and you can improve or continue to insinuate that it is rigged and never improve.
This is your answer right here, the program doesnât work properly. It has trouble evaluating many situations, especially those involving random deathrattles and divine shield resets. That program is good for nothing but stressing people out.
I would agree that we are not ready for a conclusion on this because I donât think you have looked at your data properly.
The first thing I would point out is that by throwing out the whole first group you skew all the rest of the data. How many cases are in the group where the expected outcome occured?
You have also combined your data in a way that maybe misrepresents it badly if you take cases like this
As a toss up. That has a clear most likely outcome - that you are going to take damage in half the cases. If that happens, bobâs buddy was correct rather than you losing a coin toss.
It really matters which answer is most likely and how many times it is correct.
Beyond all of that, I am not sure how we can figure out if the buddy is even good enough at guessing to separate the noise of the measure from the signal of any potential âriggingâ
I am happy to hear more about your trials and how you looked at them, but we canât begin to assert that something is amiss with the BGs from the data you are looking at.
This is possible but you canât claim this to be true without some data to prove it. I donât even know how someone like me would go about proving that bobs buddy is wrong. This is what the developers have to say about its accuracy.
It doesnât let me include the link so google âBobâs Buddy: Week 2 Status Updateâ. It talks about the accuracy a bit.
No, because it informs whether weâre looking for intentional manipulation of game results or an error in one or more of their random results, bobâs helper, or OPâs analysis.
By calling it âriggedâ you assume the conclusion then look for evidence to support it. Thatâs backwards.
If one of us has a reading comprehension problem then I am going to say it is not me. In any event I said what I wanted to say. I am out of this thread. Go ahead and have the last word.
I would compare Bobâs buddy to zephrys. I would not say he is âwrongâ but he is limited, and there are certain combinations, minions, interactions etc that can really mess with these kinds of predictive systems.
Certain comps have a knack for having huge disparities in their outcomes â amazing early pirate boards that self buff have a great way of snowballing off of each other so results compound of each other.
Other comps like deathrattle with baron have a good way of attaching complete functionality to own minon.
There are also some very high variance minions like sneed and pirate ship, or even the recursive deathrattle interactions that I highly doubt Bobâs buddy can figure out when zephrys so utterly fails at it.
Something like divine shield dragons vs divine shield murlocs is probably one of the easier matches to predict for example, but one with sneeds, foe reaper, kangors and baron will be such a clown fiesta that you will get more variance than Bobâs buddy accountS for accurately.
I understand you get triggered by word âriggedâ though I guess in this case the âriggedâ in the topic name is kinda misleading.
OP is giving us some data he has collected and hes asking if there are any other players with data. Hell he is even proposing he may be wrong with his analysis.
The poster you quoted didnât call it rigged and the OP offered a list of possible conclusions, most of which were not âriggedâ as an answer. As in, the OP made a genuine effort to propose more benign reasons for his result, going so far as to acknowledge his own limits in data analysis.
Rejecting his inquiry out of hand is small minded and not scientific. You could at least, as MegaCharzard suggested, consider what the OP asked before rejecting everything including the data.
I am at least happy someone tried to record a couple thousand matches to help understand what is happening. It is progress.
While I do not support the conclusion that the game is rigged based on this or anything else I have seen, I applaud the effort and will continue to give each case a serious look.