Is this Really Random?
Дата: 27.07.2021 18:45:16
umkhulu, on Jul 27 2021 - 18:28, said: To answer your question, I created a random team generator against which to compare WoT's matchmaker teams. The results are quit interesting.... The random team generator is an algorithm which creates any number of randomly skilled opposing teams and spits out the PR ratios for statistical evaluation. The program simulates the proportion of players falling within each skill bracket at Tier 8. (i.e. 5% Red, 37% Orange, 41% Yellow, 15% Green, 2% Blue and 0.2% Purple) I have generated 4000 matches for comparison purposes, but the stats remain pretty much the same no matter how many line-ups you simulate. Everything looks normal, but it would seem that some bell curves are more normal than others..... As you can see, the WoT bell curve is altogether more 'compact' with a standard deviation of only 3, a variance of 11 as opposed 135, and a positive kurtosis indicating less outliers. So, the question now becomes, why is MM's random so different from mine. I'm hoping that somebody has a reasoned explanation instead of the normal scathing and meaningless comments!....
DeviouslyCursed: A couple of things. 1. Are you sure your ratio of players is correct?2. Even if the ratios are correct, do they all play the same amount? For instance a blue who plays twice as much as the average# of games would should be represented twice (or close to it, since he can't be on his own team, it's not quite x2)3. I'm betting your random generator is assigning PR values within the brackets randomly, when that likely is not the case. For example, the reds are more likely to be to the higher end than the lower end, but I'm guessing your random generator will spread them equally across the whole range. For the actual PR numbers, all brackets above the average have more people closer to the lower end, all brackets under the average will have more people toward the higher end. This would likely have an effect of pushing things closer together for the actual WoT chart, because the skill spread isn't as large as your random generator is producing.4. If you are using an individual's (or a few individual's data) their own PR is likely influencing the actual WoT chart.5. Your mode in your random one says 37, and that's clearly not correct. I'm betting the biggest effect is #3 above. So, if I understand your chart correctly, you showing that the skill spread should be even higher than what it is in WoT matches?