SFC174, on 11 April 2022 - 01:12 PM, said:
I used to think the same, but the math says otherwise, and it took a bit to convince me. Over a long enough time frame the only thing that determines your win loss ratio is you.
You can put together anecdotal situations about how you played with a high level of skill and still lost, but play enough games and you're going to get the crap teams and the great teams enough times that those factors are neutralized as a determinant of WLR. The drawback is that it takes a couple hundred games to get a WLR that is an accurate representation of your skill level. That's why I've always favored some sort of additional metric to help seed players, but hey, we all went through the matchmaker changes and tier reset so we can tolerate a couple hundred games of WLR settling to get better matchmaking.
The great thing about this is that it wouldn't be particularly hard to do. WLR is already present in the statistics and Nightbird put together a proposal for how to do it within the structure of PGI's existing architecture. He's put out a lot of data on this (probably have to go back to 2020 to find some of the early posts) that's worth reading if you're interested.
The issue is that when PGI asked for modifications to the present PSR system MM (after Solo/Group queue merge), PGI was not willing to change the MM up completely to a W/L setting. Nightbird had even twittered PGI and was shot down, essentially it was not within the requirement range that they were asking for. Coding is and has been difficult with PGI. Even with the current MM, there is no secondary consideration, it is just plain Tier+tonnage, and they really cannot get that close either...
And using a W/L MM, if a 4-man group has a high W/L, be aware that (chuckles) Sorry.. I do not believe PGI could get that coded correctly, but what would happen is that most of remaining 8 players would have the worse W/L record, while the other players on the other side would have the better W/L of the that selection, which would mean that 4-man would need to actually carry, or at least try to get the remaining 8 to actually work with them.
Elo selected players/groups within a set range of the seeded player. Per Paul, the MM alternated between teams. Still matching weight class and the groups were limited to 1 mech/weight class. Then once the teams were filled out, one team with the higher Elo average was the team expected to win. If they won their Elo went up very little, the losing team Elo numbers also dropped very little. It was when the expected winners lost that their Elo drop quite a bit, the expected losers went up quite a bit.
Then original PSR went live. It was actually okay at first, but PGI never revealed the actual PSR numbers. The players though had calculated when PSR > MS thresholds, and it ensured all but the lowest MS scoring players would eventually reach Tier 1 with enough time and games played.. Nothing like it is today. Some players, one with an avg 171 MS, 0.83 W/L ratio were able to reach Tier 2 and Tier 1 by blunt force games played. In late 2019 said player was in the top 10 most games, with at 25k games having been played. He had setup his web profile to show tiers, and he had gone from Tier 2 to Tier 1, then bounced back and forth between those tiers.
Could the MM be better? It could be different, even if PGI kept the current tier system for visible queues but utilized a hidden stat, ie W/L for the actual MM. But could PGI actually code that change effectively? Math is hard.. especially for PGI...
2
