I’m certainly no genius, but there’s truth into your words as smurfing may be more frequent at certain MMR ranges than in the whole thing. I’ll take in consideration both point of view though. Thanks for the input (often rich in crispy anecdotes btw 
), Eternity.
This has been the most proper conversation ive seen in a while on these forums.
1 Like
** Black swan event** check it. You can understand. When it happned, we find the best method to solve it as soon as possbile. Try to fix the worst and then you will gain additional market share. Don’t try to quantify the entire market and control it. If the smurfing of certain MMR ranges lead to the whole thing goes worse, fix it as soon as possbile. if not, you needn’t try to fix it, you just waste your time because you can’t control the entire market. When you try to control the entire market, you are creating the new problems to ask for troubles.
Well…I just feel boring, hope can help you.
You are biased because you are only taking your own matches, and not a randomized sample from whatever range you are talking about.
This wasn’t an issue of trying to quantify the entire market, this is an issue of trying to use a model out of data, in other words if you have a model then you only assume it works within the range of whatever data you have, you can’t extrapolate.
Within my MMR range, are my opponents selected in a predefined, unchanging, unvarying manner ?
for a sample to be unbiased it means that every player in your “range” has to have an equal chance of being selected. Do you happen to sleep around the same time every day? If so then it is not unbiased.
It’s the same mean. When you want to extrapolate it out of those bounds, you are trying to control the entire market. Big data, according to the history but no one knows what will happen in the future. That’s why we still have many non-performing loans. Who knows pneumonia will happen? We only can find the best way to press for some customers to repay their debts as soon as possbile. According to the history data, they can repay it but they can’t now.
Yes, but the statistical method itself is sound, its just that people are executing it incorrectly, if its done correctly , i.e. random sample and no extrapolation then it will be good.
Well… My universtiy tutor really hate me becuase I use the divine financial mathmatics knowledge to play a video game ---- World of Warcraft.
Why not? Make more damage is fun!!! 
And it’s more fun play with Tyrael than daily work.
The times I do play are different on week-end and during the week. That still doesn’t makes my data collection independent from me, but at one hand besides the MMR range, the other factors (several players queueing, matchs duration, etc. etc.) are random and out of my control. So much that so far there has been only two smurfs I could identify has having played twice in the last six months. And, at another hand is having a players problem seen from a player’s perspective that completely irrelevant ?
I understand that the limited range sample I do have isn’t bias free, but there’s still a part of randomness in the collection process. And what you don’t want to admit is that limited process still provide limited results. To take an image your might understand better, it’s as if regarding the recent coronavirus outbreak, you completely discarded the local hospitals data we first had on the motive that they we’re not done on a worldwilde multicentric double blind pre-randomized sample. In other words, you’re saying that the sole and only data ever usable are from perfect studies (if there is such a thing than a perfect study (even randomness, sometimes, does not do things well)).
What I’m telling you is that there are degrees in evidence, different proof levels, and that from the moment you do know them, you can weight/use even imperfect data while awaiting to get better one. The world isn’t full white or full black, there are shades of grey.
No you cannot. This is not debatable. I’m trying to help you actually get to the truth of the matter. If you listen to what im telling you, you will actually have legitimate statistically sound evidence.
Ah…random sample and no extrapolation can do well? OK…Ummmmm…Would you like know the truth? Really?
The truth is a little harsh. You only can get information from the sample…and…extrapolate it out of the bounds of sample…
So sad!!! 
Ah…Just focus on the event that leads to the whole thing goes bad. Or…Try to control the whole thing and create new troubles…find the new event (the new worst event) that leads to the whole goes bad…solve it…
yes you can do that, but then you can’t be surprised if your model is wrong. We still do it all the time and its definitely worth it to attempt this even though this is not statistically sound. As long as your well aware of the risks and the limitations.
Finance is weird though, its quite a lot more unpredictable than anything else i can think of, I don’t think its fair to assume its the same in other fields.
If you can make troubles to make me feel fun in Starcraft 2. Why not make it? Balance Team always do it. They really make the game a lot more fun. At least, pro players will give me new strategy about late-game after they destoryed the starcraft 2 world every time.
A friendly reminder: If you feel starcraft 2 is a bull shxt game. You just need to try other Blizzard games and wait the new patch. When you come back, it will be more fun. Learn new strategy is fun, of course.
pero yo nunca dije eso
我没有生气，我只是觉得这很正常，我试过去平衡这个游戏，那是不可能的。
我有和你一样的感受。
To stay in the medical example, there actually are proof levels rated from IV to I with subdivisions (IA, IB etc.) depending on data types and their level of bias. So there are at least some people who thing bias levels and uncertainties can be weighted (fraudulent bias put aside ofc). And if I may, it’s actually more cautious to consider there can always be some kind of bias, and weight it, than to consider grade I studies as absolute truths, and discover some years later that there were untold oversights.
So I still think than although biased, my current methodology is better than nothing.
However I do take in account your remarks, and will try to see if the data I’m interested in can be retrieved via SC2API.
Unranked should be completely removed --> you can play custom game
or completely seperate
it means:
ranked vs ranked only
unranked vs unrankved only
and the old matchmaking system should be taken back from WOL and HOTS…
Because matching Bronze and Gold vs Master in 1v1 or teams its complete ridiculous.
Do you care Mr. Blizzard? Oh I forgot there is no more Blizzard… but only $$$…
Its really bad…2v2 and 3v3 are affected the most. By bad matchmaking. Blizzard won’t fix it. Hasn’t touched matchmaking in years.
When you are trying to estimate the proportion of smurfs on ladder what you are doing in a different term is to estimate a population paramater. This falls squarely into the field of statistical inference, not whatever else you think you are doing. In statistical inference you cannot take such an incredibly biased sample that you are trying to do and come up with any sort of “evidence”.
If you took the time to literally google the term “how to estimate a population proportion” you would’ve figured this out in a day. Stop grasping at straws and just do things properly and accept their results.