No, I think you misunderstand what he is saying. e.g.
Being handicapped is exactly what we are talking about here. Having a difference of 2000 SR between teams IS A HANDICAP.
And this is what MOST online games do, except in the case where the community controls the servers, and players “clan up”. But in the past, nearly all games allowed the reshuffling of teams to rebalance them, according to the estimated ELO. (With the exception of periodic tournaments where fixed teams would compete.)
But the key difference here is that in most games, estimating an ELO is relatively easy. They usually have a small roster of player class, with a limited number of weapons, without outsize synergies between “ultimates” or CC combos. In OW, it is nearly impossible to estimate ELO. Just look at the performance difference of certain DPS between pre-match FFA and in-game. It’s like two opposite ends of the spectrum. Doomfists that pwn in FFA are regularly and continuously shutdown in team play. Widow is the opposite, often failing to get a single kill in FFA (where the map design and the lack of objective allows almost anyone to kill her), but can pwn in-game.
The games that invented the concept of ELO did not exhibit this discrepancy. Players who were good in duel/FFA were also good at CTF or TDM. And ELO was a fairly accurate approximation of outcome.
Given the inability to measure individual ELO, it makes no sense to rank players based on whether they win or lose in a pool of randomly selected players, or even if one team is self-selected and the other is randomly selected.
The only way to really test skill in Overwatch is to have tournaments with fixed teams. The competitive ladder controlled by Blizzard is just a popularity contest, biased toward sociability.
As for wanting a stronger matchmaker, I don’t believe that it’s possible, given the inherent problems with estimating ELO. I think if a community were allowed to develop, you could organise tournaments, or even see the formation of clans who would compete with each other regularly.
But trying to estimate outcome when there are 4.5 billion permutations of 3 roles, 31 heroes with 4-5 abilities each, with varying degrees of co-ordination and synergy between abilities, would take billions upon billions of simulation trials to build a predictive model. It’s just easier just to throw in 4-5 players 500 SR lower or higher into one team or the other to skew the outcome towards some kind of volatile homeostasis, and rely on sociability for grouping up to push the team players to the top.