Board Game Arena

Ceaseless wrote: ↑11 March 2024, 20:51

ChiefPointThief wrote: ↑11 March 2024, 19:41

Ceaseless wrote: ↑11 March 2024, 18:13 That depends on the game's outcome, which is unknown. The high rated player could be getting a few points less, or may be protected from losing several points for a loss, which is presumably less frequent, but per game is more impactful. No reason for the rating gains to be numerically similar though. For one thing, that means the higher rated player gained more, as each point is worth more the higher rated you are. For another, the lower rated player has the greater achievement for winning vs a quitting player higher rated relative to them than the relative gap between the higher rated player and the quitter.

Deductions are exactly that. Them leading to a net loss of the total elo is fine. The quitter earned the deduction, the other players did not earn anything related to that. As for whether Razor earned their 0 points I'd have to see which games are referenced. I have no objections to a 0 point win if it's something absurd like a player beating someone 400+ points below them. That said, I do think the player shouldn't have any deductions for drawing the other players, even if the net change can never go below 0. The outcome between everyone else should be unknown, no change, everyone gets a win vs the quitter. A +1.14 gain is also a plausible solution depending on the game too.

The whole purpose to op proposal is that good players shouldn't be penalized for others quitting which would still happen in the discussed everyone 1v1 with the quitter "fix". The problem is that these aren't 1v1 and so they shouldn't be treated as such. In my proposal the "unknowns" are replaced with a known. All should've could've would've are thrown out and there is just 1 loser and several winners in a multiplayer game. Because the quitter created this scenario they are being held liable to compensate all the other players and there isn't even a need for an extra deduction because we can use the deduction already in place to help with this solution.

If you want to use reasons like elo means more the higher you get so it shouldn't be exactly even I wouldn't be opposed to that. But the high elo players deserve something reasonable for what they would've got for winning a 4p game. Whatever that reasonable number is depending on the game it can come from the quitter whose elo is already going nowhere anyway.

Yes, all should'ves, could'ves, etc. are thrown out. The quitter lost, and everyone else beat them. There was no first place in the free for all, just a last place.

Ceaseless wrote: ↑11 March 2024, 20:51 Everyone else beat 1 player, and they didn't beat or lose to anyone else. The gains will be less than what a player would have gotten for beating everyone, since that didn't happen, and more than they would have gotten if they has lost to everyone besides said quitter, since that didn't happen either.

Ceaseless wrote: ↑11 March 2024, 20:51 I'm not sure what you mean by "there is no need for an extra deduction" when said extra deduction is there to penalize the quitter and that's it. It's not to "compensate other players", it's just a penalty on that player.

ChiefPointThief wrote: ↑12 March 2024, 01:38

Ceaseless wrote: ↑11 March 2024, 20:51

ChiefPointThief wrote: ↑11 March 2024, 19:41

The whole purpose to op proposal is that good players shouldn't be penalized for others quitting which would still happen in the discussed everyone 1v1 with the quitter "fix". The problem is that these aren't 1v1 and so they shouldn't be treated as such. In my proposal the "unknowns" are replaced with a known. All should've could've would've are thrown out and there is just 1 loser and several winners in a multiplayer game. Because the quitter created this scenario they are being held liable to compensate all the other players and there isn't even a need for an extra deduction because we can use the deduction already in place to help with this solution.

If you want to use reasons like elo means more the higher you get so it shouldn't be exactly even I wouldn't be opposed to that. But the high elo players deserve something reasonable for what they would've got for winning a 4p game. Whatever that reasonable number is depending on the game it can come from the quitter whose elo is already going nowhere anyway.

Yes, all should'ves, could'ves, etc. are thrown out. The quitter lost, and everyone else beat them. There was no first place in the free for all, just a last place.

Not according to bga. They label all other players as first place and put a message at the top saying the skipped player lost and all others won the game.

Ceaseless wrote: ↑11 March 2024, 20:51 Everyone else beat 1 player, and they didn't beat or lose to anyone else. The gains will be less than what a player would have gotten for beating everyone, since that didn't happen, and more than they would have gotten if they has lost to everyone besides said quitter, since that didn't happen either.

Yes. The gains would be less than if they won a completed game just like in my scenario, and more than what they would've gotten for losing to everyone just like in my scenario. The only 1 big difference which would be the entire point of the "fix" is that the high elo player would get some points in my scenario. I don't understand the objection to the high elo player getting some points when in a scenario like razor's he has a probability of 80-90% over all players.

Lets look at scenarios on how these proposed ideas would look. Here is a game of razor's that actually finished. https://boardgamearena.com/table?table=483398395

1st place won 5.12 - vs last 1.39
2nd Place won 10.28 - 4.78
3rd Place won 5.17 - 6.48
4th Place won 0 - 8.15
5th Place lost 20.85

3 out of the 4 players elo would be drastically different from what the actual results would be if we use the 1 vs quitter fix.

Everyone gets even elo would be 5.2 each. 1st and 3rd would get identical numbers while 2nd and 4th would be off but closer to what the actual game results would've been.

If everyone was to split the -10 karma hit from the quitter (2.5 each) and receive the 1 vs quitter elo that wouldn't be that bad either.

But any "fix" should not include the high elo players getting stiffed otherwise you can just leave the system as is.

Ceaseless wrote: ↑11 March 2024, 20:51 I'm not sure what you mean by "there is no need for an extra deduction" when said extra deduction is there to penalize the quitter and that's it. It's not to "compensate other players", it's just a penalty on that player.

Why have an elo penalty on a player for losing and simultaneously penalize their opponent by stiffing them elo ? That makes no sense.

Ceaseless wrote: ↑12 March 2024, 02:31 Having an 80-90% expected chance of victory going into the game is precisely a reason not to grant more elo. The greater your proposed chances of success, the fewer points you get for actually achieving it, that's the whole purpose. Also why they lose more for upsets, etc. I'm not sure what posting a game where Razor actually got first is supposed to do. There was a real first place player out of five. They beat the other four players and got ratings for those victories. The other 3 players also lost to players and got deductions, which don't exist in the scenario where there is only 1 loser and the other four players did not lose to anyone.

Ceaseless wrote: ↑12 March 2024, 02:31 And yes, splitting the unearned 10 elo would be terrible, especially if it went to the higher rated, where the points become more relevant. Free rating becomes more problematic the higher up the players get.

The whole concept of a player somehow being stiffed by not gaining rating from an opponent's penalties is absurd. The opponent lost the game, and so lost points. They also quit the game, so they lose additional points. If I won that game, I get points for winning. I had no part in making them quit, and even if I did, that would hardly be cause for a rating gain (more like moderation action.)

Ceaseless wrote: ↑12 March 2024, 02:31 Having an 80-90% expected chance of victory going into the game is precisely a reason not to grant more elo. The greater your proposed chances of success, the fewer points you get for actually achieving it, that's the whole purpose. Also why they lose more for upsets, etc. I'm not sure what posting a game where Razor actually got first is supposed to do. There was a real first place player out of five. They beat the other four players and got ratings for those victories. The other 3 players also lost to players and got deductions, which don't exist in the scenario where there is only 1 loser and the other four players did not lose to anyone.

Ceaseless wrote: ↑12 March 2024, 02:31 And yes, splitting the unearned 10 elo would be terrible, especially if it went to the higher rated, where the points become more relevant. Free rating becomes more problematic the higher up the players get.

The whole concept of a player somehow being stiffed by not gaining rating from an opponent's penalties is absurd. The opponent lost the game, and so lost points. They also quit the game, so they lose additional points. If I won that game, I get points for winning. I had no part in making them quit, and even if I did, that would hardly be cause for a rating gain (more like moderation action.)

RazorOz wrote: ↑12 March 2024, 09:39 At an absolute bare minimum #1 needs to change, so that all players actually gain ELO from the quitter. Also, this doesn't actually bother me, as I don't mind more ELO in the system at a time when that's become problematic because of other changes, but if we're really concerned about players unfairly gaining ELO, it should be noted that players gaining ELO from tying high ranked players, especially if they'd actually lost, is actually farcical if we're talking about fairness. As said before a tie because the game was voided is not the same thing as a tie because we ended on the same points at the end of a game, the system is currently treating them the same, but it shouldn't be.

ChiefPointThief wrote: ↑12 March 2024, 09:33 The opponent by quitting caused a situation in which they can't get their fair rating from anyone but the opponent.

h_illes wrote: ↑12 March 2024, 09:59

ChiefPointThief wrote: ↑12 March 2024, 09:33 The opponent by quitting caused a situation in which they can't get their fair rating from anyone but the opponent.

The problem is that 'fair' cannot be defined in a universal way. What about games where there is a quitter 2-3 turns before the end, but the other players are very close, so the winner isn't even decided? Also, the player leading by any game metric at any point before the end might not be actually winning. Also, with 2-3 turns left, players might still make mistakes. (Just to be clear, I offer no hard conclusions, I just want to highlight that the concept of 'fair' is complicated.)

h_illes wrote: ↑12 March 2024, 09:59 While the current situation of getting rating denied due to circumstances outside of a player's control is definitely horrible, giving extra rating for high-rated players due to whatever reason (circumstances outside their control or not) is not a good idea either. I would hate to see rivals gain free rating like that, and to be honest, I don't want it even for myself; I'd rather earn rating the hard way. Unfortunately, I do not see any perfect solution for quitters in multiplayer games; the best attempt I could think of was what I described above, which essentially cancels the part of the game between the non-quitters, similar to when players agree to cancel a game.

RazorOz wrote: ↑12 March 2024, 09:39 To be clear I am not arguing players should gain ELO from penalties applied to quitters. There's two things though.

#1 I never gain ELO when someone quits. It's simply impossible for me to do so the way ELO neutralisation currently works. ELO neutralisation has clearly been designed around 2 player games where it might work as designed, but any other player count it's not really fit for purpose. Tied winners in a game someone quit should not be exchanging ELO, given it wasn't actually a tie, but under the current framework they do, which is why it's impossible for any high ranked player to gain ELO when someone quits, as we lose points to ties, and then the system neutralises the result to 0 rather than losing ELO (I have seen examples posted where this part doesn't happen, and someone does actually lose ELO, but have never had that happen to myself).

#2 A reasonable number of the quits the game result was already decided. I had already finished in 1st place. The vast majority of quits in Heat take place at the end of a race, a good number are after some cars have already finished. So it can be very misleading to say there was no winner, the system might be saying that, but often there very definitively was a winner. Others will be ahead of the last turn because someone knows they can't win as they haven't cleared the final corner, while 3 cars have. In that scenario the game result is already effectively decided too, as that player can no longer have any influence over the result.

At an absolute bare minimum #1 needs to change, so that all players actually gain ELO from the quitter. Also, this doesn't actually bother me, as I don't mind more ELO in the system at a time when that's become problematic because of other changes, but if we're really concerned about players unfairly gaining ELO, it should be noted that players gaining ELO from tying high ranked players, especially if they'd actually lost, is actually farcical if we're talking about fairness. As said before a tie because the game was voided is not the same thing as a tie because we ended on the same points at the end of a game, the system is currently treating them the same, but it shouldn't be.

ChiefPointThief wrote: ↑12 March 2024, 09:33

Ceaseless wrote: ↑12 March 2024, 02:31 Having an 80-90% expected chance of victory going into the game is precisely a reason not to grant more elo. The greater your proposed chances of success, the fewer points you get for actually achieving it, that's the whole purpose. Also why they lose more for upsets, etc. I'm not sure what posting a game where Razor actually got first is supposed to do. There was a real first place player out of five. They beat the other four players and got ratings for those victories. The other 3 players also lost to players and got deductions, which don't exist in the scenario where there is only 1 loser and the other four players did not lose to anyone.

So the fact that the higher elo players are projected to win this game the majority of the time when there isn't a quitter is a reason to give lower elo players more rating than they would normally get if there was no quitter? The purpose of me showing a game that actually finished is to show how the majority of the games ending with a quitter will result in the projected winners being stiffed. What I am proposing is the higher elo players get something relative to what they were projected to get without the quitters actions. Not more. It is only more compared to the system in place now because higher elo players are being stiffed.

Ceaseless wrote: ↑12 March 2024, 02:31 And yes, splitting the unearned 10 elo would be terrible, especially if it went to the higher rated, where the points become more relevant. Free rating becomes more problematic the higher up the players get.

The whole concept of a player somehow being stiffed by not gaining rating from an opponent's penalties is absurd. The opponent lost the game, and so lost points. They also quit the game, so they lose additional points. If I won that game, I get points for winning. I had no part in making them quit, and even if I did, that would hardly be cause for a rating gain (more like moderation action.)

It isn't free rating. It is rating they should earn for winning a 4p game. They are not being stiffed for not getting rating from the opponent's penalties. They are being stiffed because they are not getting the proper rating at all. The opponent by quitting caused a situation in which they can't get their fair rating from anyone but the opponent. So because of this the opponent should give them their fair rating from points they are ALREADY losing that goes to NOWHERE.