A side note, ELO systems. ELO or MMR systems use some variation of th ELO algorithm created for ranking chess players. It works by assigning players a number which denotes their skill.

When two players play eachother, the probability of either person winning is calculated based on the difference in their scores. The general formula, is that two people with equal scores have a 50/50 shot at winning. For every point by which one player iss below the other, that players assumed odds of winning g are reduced by some percentage for every point. So 1 point below you might reduce his odds by 1% so he has a 49.5% chance of winning. Then another point so a other 1% now they have a 49.05% chance of winning and so on and so on.

When the game is over, the players scores are adjusted according to how much of an upset the results of the game were. So, if a player won a game they were likely to lose, they'd get more points than if they were likely to win it. Eventually, you're so likely to win that you would gain less than some threshold of points. At this threshold you usually just gain 0 rather than a fraction of a point. Both to keep the math easy in the future, as we as to represent that you played someone so much worse than you that it basically doesn't count.

The amount of points you can win at once are also bounded. So the harder the odds are against you, the less pay off you get the more unlikely it is. In other words, you have to win increasingly more unlikely games to get the next additional point.

Because the way the odds work, the game is considered to be too wide in skill gap before the odds get so extreme that you get 80 points in one game. This solves the problem of outlier games where sometimes an unskilled player will get lucky against a far more skilled opponent, and has to play many more games to even things out.

Finally, all the points one player gains, the other loses. Meaning that if you lose a bunch, your rati g goes downdown, then you win all your rematches, you should end up in a similar place, though technically be a use of rounding error you might not.