Posts from Bug reports.

I dare say that such a phenomenon can be explained by a peculiar protection against PTS overflow, just like the fact that in subsequent matches with the same result, the difference between one player's rating loss and the other's gain will increase (up to plus +0 for Lucky in this situation).

Well, if you look at what Lizard wrote above, one could assume that next time Lucky will get 5 PTS, meaning there will again be a division by 2 with no remainder)
And the identical deduction from the loser in this situation is due to the maximum rating gap, because more than 47 points will not be taken from the loser, no matter how huge the gap between the players.

I recall the rating mechanics used to be different, worse than the current one, for sure.

I won't list all the factors of rating change, but earlier there was such a buggy feature that a player desperately wanting to reach top 1 by means of, let's call it, "farming" Jebus, would find it much more profitable to surrender on day 2-3 after seeing a not-so-easy respawn, or carelessly stumble into max secrets and calmly surrender, losing only 3-5 points and go roll a new one.
As you might guess, the person's motivation was mutated, and the ladder gameplay could also soon mutate.

At some point, I was even slightly puzzled when a player with a solid rating, after surrendering on the second day, wrote to me about throwing the game in the gnome dwelling=) Moreover, okay, this happened once, but after that the same guy decided to surrender again on turn 111-112)

This is precisely the direct difference from HeroesWorld — if here the guys will squeeze everything out of the layout, some won't even try in the lobby)

At that moment, it bothered me, and I wrote, I think, to Baratorch about this situation.

To summarize, the idea of farming HeroesWorld/lobby rating is not great)

Don't you realize how absurd this system is if you correctly guessed how it works? The addition and subtraction of rating shouldn't differ, and especially not by such a huge margin – it's just terribly flawed. And the fact that the number of points awarded in the lobby depends on the submission deadline is just pure madness, sorry.

The rating in the lobby is now just an ELO rating with a coefficient that depends on the map size.

Are the different advantages and disadvantages balanced in one game?

Regarding the rating: the system has been changed periodically, but in the latest HD mod updates, obscure elements have been almost completely removed.

The rating in the lobby now is simply an ELO rating with a coefficient that depends on the map size.

The rule for changing the rating as a result of a game is now as follows:
R1 += K*(RES - 1/(1 + 10^((R2 - R1)/400)))
R2 -= K*(RES - 1/(1 + 10^((R2 - R1)/400)))
Here, R1 is the rating of the first player, and R2 is the rating of the second player. RES is 1 if the first player wins; 0.5 if it's a draw; 0 if the second player wins.

The coefficient K depends on the size of the map being played:
K = 60 * ([map width in cells] * (1.5 if with Dungeon) / (144 * 1.5)).
For example:
XL+U: K = 60
L+U: K = 45
XL-U: K = 40
L-U: K = 30
M-U: K = 20

Let R1 and R2 be the initial ratings of the players.

The player rating difference coefficients are calculated as follows:
f1=max(-1, min((R1-R2)/400, 1));
f2=max(-1, min((R2-R1)/400, 1));
where min and max are functions that return the minimum and maximum of two numbers.

The points awarded for a win (WP) and a loss (LP) are calculated as follows:
WP1=max(1, 16*(1+f1));
WP2=max(1, 16*(1+f2));
LP1=min(-1, 16*(-1+f1));
LP2=min(-1, 16*(-1+f2));

If the first player wins, then
R1->R1+(int)WP2;
R2->R2+(int)LP1;

if the second player wins, then
R1->R1+(int)LP2;
R2->R2+(int)WP1;

if the game ends in a draw, then
R1->R1+(int)((WP2-WP1)/2+0.5);
R2->R2+(int)((WP1-WP2)/2+0.5);

where (int) is a function that returns the integer part of a number.

There are no instances in the history where a unit skipped its turn without dealing any damage or having any defense.

This usually happens when under the Berserk effect.