I disagree with both, but since I’m not the one making the claim, it’s on the claimant to provide the mathematical basis for his claim.
2 Likes
Depends what you are playing on, console it is Diamond, on PC it is Plat. You are welcome.
1 Like
Suspend your arbitrary disbelief about about the mathematical result (C). i.e. the claim of asymetrical or multi-modal pdf disparity.
On what grounds do you disagree with the assumptions (A) and (B)?
If A and B were true, why would C be false?
You need to show something to disprove it.
Show something that explains why you claim it’s false.
1 Like
I disagree with A because we have no data showing how many people own alt accounts and with current systems, we cannot possibly know.
With B, I disagree because we have no proof which alt accounts are higher than main accounts and which are lower, and anecdotes are not data. It’s unknowable with the data collection systems we have.
Now, where’s your proof?
EDIT - changed the autocorrection “unprovoked” to “unknowable” for clarity.
2 Likes
Of course there is no data. There was declared when I said “I don’t”.
This is why everyone here has to use assumptions. Mine are extremely basic and conservative.
I repeat: if a player has an alt, that player will contribute >=1 alts:main ratio. That’s (A). For (B): Alts are by definition more disposable and less serious, so they will naturally drift to lower SR (in general) compared to the main. Higher skill is less frequent and maintainable than lower skill (effort based skill sieving). You will own or occupy more lower SR than higher SR for you as an account-holder than you as a player (if you have more than 1 account). The edge case: since we know at least one person has more than 1 account, we know this must be true.
A booster with 18 accounts all in GM we count the highest placed account as the “main” and the others as lower alts which again sour the distribution (case for multi-modal). A plat with 2-3 accounts less than plat again affects the skew.
These “assumptions” are more like primitive facts which are data-free.
And their direct result (C) is the normal pdf for player % by SR goes away and instead you have a degenerate looking pdf for number of accounts by “SR”.
So again, explain why you refute A and/or B. Now that I think of it, you might only need A to hold for C to be technically true. The existence of a single alt somewhere already disrupts the “players by % skill” statement.
Finally, assuming A,B, explain using math why C is not reachable. Thanks.
1 Like
Nah, you admitted you have no data, which is all I wanted. You’ve now shown the thread that you’re offering nothing useful to the argument and I’ll take my leave, since I don’t feel like trying to mathematically disprove an argument you’ve made on feelings.
Have a good night!
Right since you are both idiots as I said before in my other post ask me anything regarding ranking…I have been through the ringer on this…just stop fighting please…this is enough on this.
But you don’t have data either? Your “data” is for player% not accounts% and it is years out-of-date (and you didn’t even post the reference/source).
I’m using existence proofs. we don’t need data just the fact that alts are a thing.
that is a MUCH stronger and data-heavy statement to make, yet you haven’t provided anything recent or w.r.t. accounts not players.
my claim is based on distribution structure, your claim is based on raw data about a distribution that isn’t valid.
1 Like
Usually above average is diamond and good at the game is masters
1 Like
We don’t have newer data, and I don’t feel like pulling guesses from my fundament, so I’ll simply state that this is the last official data we have:
…and unlike other people, I don’t feel like making uneducated guesses based on feelings. And since again, you keep providing nothing relevant in the way of data, I’ll again wish you a good night and end this lack of discussion.
1 Like
The data you provided is years out-of-date and based on player% not accounts%. It’s not valid in fact it’s misleading.
What isn’t misleading: If a single alt exists the way I have defined it (A and B) then C is true. It’s a data-free argument based on existence proof.
You can’t talk about “playerbase” distribution when speaking about SR, and you can’t really talk about SR when speaking about skill (mmr).
Alts exist, corruption exists, and no-reset deteriorations exist. In the base case for n=1 we already know the noise effects on the pdf. Scale that out to whatever N you want, you already lose your claim about the pdf.
Data-free will always beat data-driven nonsense using the wrong metrics and out-of-date sources. We’re literally inducting for the set of any/all data.
1 Like
Average is 2350. Anything above that is technically above average.
3 Likes
Wait, you forgot to consider split correlation of multi-modal probability distribution functions!
Considering conditional probability of A in respect to B you can obtain a Bayesian estimator for ladder’s intrinsic error, that would allow us to compute the BBATG coefficient for each player and determine if ranks are statistically accurate.
Alternatively we could use the method from Bauszat et al. to construct dense error
prediction from a small set of sparse estimates, that would allow for a better data noise rejection and lead to somewhat more accurate results. It’s much more complicated and computationally intensive tho.
2 Likes
Rigged is pure comedy gold. For a mad scientist, he has an amazing amount of time to troll the forums.
5 Likes
Man invented data science and he sits here arguing with rando’s on a video game forum?
3 Likes
It’s for the greater good. Overwatch has the potential to cure cancer and feed the world.
3 Likes
I’d almost believe the user if everything they said actually made any sense.
1 Like
A serious “mathematician” or computer scientist would at least, out of respect, take the effort to explain complex processes in a generally understandable way. At least that can be attributed to Cuthbert.
Instead, Rigged/Receips/Outside talks completely crazy stuff, because he is seriously delusional or because he wants to distinguish himself as particularly smart.
Since nobody understands his nonsense, of course, one cannot contradict specifically, except that his statements simply do not stand any examination.
4 Likes
Rigged, I can’t remember if it was you talking about statistics a lot that got me toying with the idea of taking the plunge in that discipline, but if it was, let me offer you a HUGE bit of thanks. I’m only a chapter into Bayesian Statistics and it’s one of the coolest things I’ve ever learned about. The way it’s already changed my thinking about my daily experiences is as quick and profound as it was unexpected.
I see that it could have a lot of applications surrounding the topics of debate on these forums.
Oh but I did! Please don’t jump to conclusions!
Without data we can only assume bounds on the mixture densities.
The only assumption I require is that the lower modalities are of higher variance than the higher modalities. It stands to reason that lower SR gameplay is more volatile and variable whilst higher SR gameplay is more consistent. Is that a fair assumption for the first few moments, and if not, why?
ewww. we’re trying to be param-free dist-free and data-free so that it holds robustly.
and we’re trying to make this computationally efficient.
exactly. we don’t need mcmc denoising lol
No. The reconstruction used by Bauszat et al. (basically a plagiarism from Liu & Zheng) is operationally akin to Sobel filtering and not really applicable here. It requires gradient information and underperforms on fine grained details…like sandbagging your performance during critical moments whilst padding your overall match stats.