Tarteso, on 20 March 2023 - 02:02 PM, said:
I´m interpreting those numbers as unadjusted % of use.
The numbers you're seeing on Jarl's are "adjusted" to fit the intent of Jarl's: To make performance metrics between players and their "preferred" and most often used mech chassis comparable between players.
Tarteso, on 20 March 2023 - 02:02 PM, said:
I don´t know for what are these weight class coefficients being used for
So you don't actually have understood what the statistics you tried to throw at me actually mean.
Tarteso, on 20 March 2023 - 02:02 PM, said:
and can´t guess why mech usage should be adjusted, at all.
Again a slight reading comprehension problem then. The text that tells you about those coefficients also should tell you why it's done. Hint: mech usage is indeed not involved but I do seem to recall that "we" talked about more than just "mech usage", didn't "we"?
Tarteso, on 20 March 2023 - 02:02 PM, said:
If anyone can confirm that absolute % usage are being adjusted that way, I´ll redo my analysis. But I don´t think it could make any difference, given the huge deviations in every ton class.
Oh, more misunderstanding there. The problem concerning % usage of mechs is an entirely separate problem with your methodology and has nothing to do with the coefficents (which deal with key performance indicators like average damage, average match score in order to make them comparable between classes and from there to the players themselves).
The problem as far as % usage is concerned was your heavy selection bias when doing a "quick analysis" based on the top 100 players. In order to get to proper results you'd have to compare about 380 to 650
randomly chosen players from the roughly 30000 recorded "active" players on Jarl's to have a decent enough chance of getting usage percentages that have confidence levels of 95 to 99% and within 5% margin of error.
Tarteso, on 20 March 2023 - 02:02 PM, said:
The funny thing is that, if these coefficients are related to the "recent global performance", here you have the proof, apparently, that lights are (moderatelly) overperforming.
The "proof" is apparently that Lights require an 1.11
multiplier on their performance values to make their results actually comparable to that of an Heavy. If you have to inflate the actual number by multiplying it with a value that increases the original value in order to make something comparable against something else, there's little chance that the something that you are inflating actually has "overperformed".
Tarteso, on 20 March 2023 - 02:02 PM, said:
And thus, I think that you are wrong in the meaning of these coefficients (your last paragraph).
Thus I'm very confident that you still have no clue what the statics you're trying to throw at me actually mean.
Tarteso, on 20 March 2023 - 02:02 PM, said:
Provide a handy table with such data.
Fun fact: The Jarl's list is one such table of such data ... currently spread over about 998 pages.
Tarteso, on 20 March 2023 - 02:02 PM, said:
Heck, if you have such thing, perform stat analysis yourself and share your meaningful and unbiased results with everybody.
You and me both have the same access to said data. Unlike you I didn't make a faulty "quick analysis" and nothing I wrote so far put me in a spot where I'd have to actually go and do this for you or anybody else
You on the other hand did put yourself in such a spot and now you're trying to pass the onus onto me but I'm not taking it.
Tarteso, on 20 March 2023 - 02:02 PM, said:
I´m not going to hesitate about your results, I promise.
Nor would I hesitate yours if they reflected you actually understanding what you're talking about and indicated proper procedure.
The "funny thing" is: I specifically asked the OP (and since you tried to pick up the mantle you as well) to provide such numbers because I'm familiar with others having done such analysis (and even posting the results on this forum in older threads) and those prior results always put heavies and assaults at the top both in usage and key performance indicators like average damage and average match result while Lights firmly were the bottom-feeders.
So: Convince me that this has actually changed by properly doing the analysis of available data to show that what you claim to be true is actually true.
Tarteso, on 20 March 2023 - 02:02 PM, said:
Then please, enlighten me with your exact knowledge about the real mech class performance/usage, based on real data, and settle the question once and for all. I´ll be very grateful.
Not my (actual) claims,so still not my job
Edited by Der Geisterbaer, 21 March 2023 - 01:02 PM.