Ladder suicide
-
I consider it ok if it fits the game scenario, just a shame the game isn't recorded properly (I know its not fixable)
-
The bug didn't exist a few years ago, meaning there is an already known method for rating 1v1 matchmaking that works. It didn't work for non-1v1 games, but that doesn't matter if you just make a special case to handle 1v1 differently. I was told that this would be done if a solution was not found soon, but that was already like a year ago now. Devs are obviously just trolling me at this point.
-
There is already a special case for 1v1 you can go look at it with your own eyes. https://github.com/FAForever/server/blob/fbd13cac9ca4a5776da1a994ae184e6b4fe3cb82/server/games/ladder_game.py#L74
-
@melanol said in Ladder suicide:
In this game: https://replay.faforever.com/16883598 my opponent (blits) CTRL+K'ed to blow up my ACU; according to FAF Analitics (https://kazbek.github.io/FAF-Analytics/) he received points, I lost points. How does it work?
"Com-bombing", is what this is called. It's a very cheesy strategy, but it will work on ACUs that are below 2.5k HP at the time of their self-destruction. It's the ultimate way of saying "If I'm going down, I'm taking you with me!"
Some people take it a step further, putting their ACU in a transport so they can get shot down by anti-air weapons while they're flying above their target. Unlike a typical Ctrl+K, this is more akin to a traditional "kamikaze/suicide bombing", and is often done to damage enemy infrastructure on top of wound and/or destroy enemy ACUs - which is helpful in games where Full Share is enabled.
-
It is not a bug. It is a successful accomplishment of the task - the destruction of the enemy commander. Whatever cost, mission must be completed. It is normal logic
-
@mediamix1 Com-bombing to kill the enemy ACU, in a 1v1 (or in a multiplayer match, if you're the last 2 ACUs) logically should result in a draw.
The fact that in ladder 1v1s, it doesn't always result in a draw, is what makes it a bug.
This bug basically only impacts high-level ladder players, because at low- and mid- levels on the ladder, it's easy to find matches and it's easy to get more ladder points. For someone like Blodir or Tagada, it's not so easy. That's why higher-rated players want their opponents to not take advantage of this bug
But it would be silly to try to ban lower-rated people for this. Enforcing it would be a waste of time, educating players on this rule would be a waste of time, etc.
-
I thinks drawn games ELO rating should be calculated the same way as it does in tournament chess.
-
@jip said in Ladder suicide:
I'm afraid this is a difficult bug that we've tried to solve in the past. And by doing so, made it worse. It is not likely to get picked up any time soon. I'm not sure, but if I am correct it is bannable to consistently win games like this. @Deribus could you confirm this?
It's not something we're too concerned about because it requires pulling off a draw to even trigger, so someone attempting to do so consistently is not a worry.
However it's covered under "rating manipulation" so if it were a pattern of behavior yeah we could ban for it.
-
You can draw pretty consistently especially on the smaller maps, but at least for high level players that know each other, most of the time we'll pause the game and just offer a draw using the diplomacy menu.
-
if you draw through the diplomacy menu, there are still rating changes. which seems broken too but it doesn't effect the league leader boards.
-
Well yes, RLO there should be rating changes if you draw with the diplomacy menu, if lower rated player draws a higher rated one there should be some rating transfer, of course, it should be less than an actual win. The issue is when you ctrl-k with acu to get a draw, the game will count that as a win for the person who blew up, and I presume this is the draw-bug that the post is talking about. It's been an issue for a while now.
-
for instance if there is a disconnect at the start of the 4v4 tmm game and all remaining players select draw after 1 min, it affects rating. at the same time, if you are electing to draw, then you are agreeing that you are equal and there should be no rating change. 1v1, 2v2 or 4v4.
Draw = a game or contest that ends with the score even.
agreed maybe not the correct thread and i shall retreat
-
If you want no rating changes then there should be another button that invalidates the game. Archsimkat is right - if you draw versus an opponent while you are rated higher then there should be a rating difference. As you are apparently equal, so the rating is changed such that you're more equal rating wise before the game then you were after the game.
-
fair enough, but is the change proportionate? i.e. 10 point swing in rating for a win/loss and 5 point swing for a draw (as an example, it all depends on ratings of course)
Just curious at this stage.
-
If I am not mistaken it uses the true skill rating system in the backend, which is well studied. @BlackYps can you shed light on this?
-
-
The point change is not the same every time because it depends on the mean and deviation of both players' ratings. I'm not sure what you mean with proportionate. There is no guarantee that the sum of the ratings stays the same. For players with settled rating a win gives about 8 to 14 points. A draw maybe 2-3? I can't think of a game with a draw of the top of my head, but you can check for yourself in the client if you click on a replay in the client and then press the "show rating change" button.
-
If you continuously draw against the same player (and don't play against others) your mean rating will eventually become equal.
The speed at which is happens is in theory an adjustable Trueskill parameter, but only in general for all rating changes, not just draws. So if you reduce the speed, you will also reduce the points you gain/loose from winning/loosing. The parameter controls the "volatility" of the ratings. In theory you could measure which parameter value would make Trueskill predict the outcome of a match correctly most often, but that would be quite bit of work and i don't think the current value is too bad.
(I think there is a different parameter that is supposed to tell Trueskill system how likely draws are in general but i am not sure what changing that would do.)
