question?

From what I understand: WHR purpose is to compare you to other players (color is added after depending on your position inside the finished ranking). So in a hypothetical case where you get beaten 20 times by player A and you beat 1 times player B and A beat B once the ranking will be A (1st position), you (2nd position), B (3rd position).

You will "go down in ranking" much faster if you loose from a player with a (previous) rank much lower than you (because the algorithm will be like "hey I thought this guy is good but he got beaten by the other one").

As there are no fights between every other player, the algorithm does things transitively (if you beat X and X beat Y probably you are better than Y).

If the ratings are correct and there is no transitivity problem, the expectation value of elo change or WHR change is zero, no matter who your opponent is. This means your small chance to gain a lot and your big chance to loose a little add up to zero.

If you loose always, your rating will converge to minus infinity. But the convergence will get slower and slower. Also, the probability of always loosing goes towards zero if you still have some skill. If you loose for an infinite amount of times, you deserve the rating of minus infinity.