the *decreasing ELO gain if you win against the same players rule*

[joyoy]

1. I know that Elo is based on who wins/loses only. I dont know how many more times I should try to explain the same...
the game result determines the Elo - but who wins or loses is determined or influenced by personal style or approach in a lot of / most games!! the more so the more interaction there is.

if we conisder 4 players, a b c d, with the same Elo. if their relative styles or strenghts work in a way that:
a > b > c > d > a
(a wins most against b, who wins most against c, who wins most against d, who wins most against a)

do you still maintain that Elo does not depend on playing style?
- if a never plays against b, and most against d..
- compared to a plays most against b and never against d..
his Elo will be much much higher in the second case. even though initially the Elo of his opponents were the same, but a's compatibiliy was much worse or better.
here the rule is to limit that effect.

ELO doesn't care how winner and loser are determined. If you're playing a game with high or low interaction, just flip a coin or let players discuss who is the winner doesn't matter.

Your example with a having a higher ELO because he only plays against b is not a problem with ELO calculation. It's a problem with pairing. If you really want comparable ELO scores, you have to make sure that a plays a comparable number of games against d and not change the ELO system.

I dont think anything you wrote actually refers to me.
in fact you confirm that it is a problem with pairing, the same I showed. and exactly what this rule can influence - in games with high interaction. but not in games with no interaction where there is no preferred pairing for players.

from now, I dont wish to repeat the same or similar content every time. especially after getting skewed repeated times.
if any reader wants to retrace the issue and request, you can read again from the first page. I dont want to write anything else here.
I tried my best according to my abilities to phrase & explain it in many different ways. more deatiled than most replies in this thread. I am not a native English speaker and if this is not enough, so be it.

[BarnardsStar]

Have you considered trying to explain what you mean in German? Automatic translators exist, we can all use them. If you think your English is not up to the task of explaining what you mean, I don’t see any reason you have to continue in English. My German certainly would not be up to the task!

All this stuff about cyclical winners, with a > b > c > d > a etc etc, has been a part of interesting discussions here in the past, about whether Elo is even sufficient sometimes. Can players’ skill at a game truly be reduced to a linear scale? Here, indeed, is where “interactivity” can play a big role. In a game like Gamme Logic or Reflection, where it’s just “how fast can you do a thing”, it’s already a linear scale and so Elo can map to it pretty straightforwardly. In other games, maybe you can’t actually do a real linear mapping, but you can still do an approximation because what else do you have?

However, none of this is relevant to why the “decreasing Elo gain in a winning streak against the same player” rule exists. Here’s where we come back to the possible language barrier. I, and others, have told you this repeatedly but you keep bringing up “preferred pairings” (and I’m not really sure what this even means) and such, still trying to argue for some special exception for Gamme Logic, and it seems to me like you maybe didn’t understand the part where I quoted lordalx’s announcement of this rule and his explanation of why it exists.

It is purely anti-cheating. Elo assumes that all players are trying to win. If one of them isn’t then Elo is irrelevant, and this rule is a (very crude) attempt to address that. Did I mention before I think it’s stupid? It is, but that doesn’t change the intention behind it or the fact that there is nothing about Gamme Logic that suggests any exception or modification of the rule is warranted.

[Jellby]

I think I can guess what they mean.

In a game with non-transitive win probabilities (A > B > C > A), elo cannot be completely accurate. But if A is matched regularly with B as often as with C, it may be as good an approximation as it can be. On the other hand, if A is only matched with B and never with C, their elo will be overestimated (within the A,B,C population). One might think that the elo reduction for consecutive wins was aimed at correcting for this case. (Whether or not it would be an effective measure, that's another matter.)

However, the elo reduction was never meant to be a "correction" mechanism for biased pairings. It was introduced as an anti-cheating mechanism, that's all, even if the wording could make it sound as if some kind of "fairness" was intended:

Power levelling
We encountered many situations of power levelling, linked to multi accounts, even if we are still fighting against such practices, it became obvious that the current system might encourage it.
To prevent this, any consecutive wins against the same opponent will reduce the amount of ELO points awarded from this victory. This reduction will be gradually higher as more consecutive wins are accumulated. This way, we encourage you to challenge other players to keep the original spirit of the ELO: show your skill at a game compared to the whole community, not just the same ones.

[Ehronde]

I dont think anything you wrote actually refers to me.
in fact you confirm that it is a problem with pairing, the same I showed. and exactly what this rule can influence - in games with high interaction. but not in games with no interaction where there is no preferred pairing for players.

It absolutely has. People here explained a couple of times why it doesn't matter if games have interaction or not, when it comes to ELO calculation. And therefore it makes no sense to change ELO reduction for games with low interaction.

[nik592]

I think I can guess what they mean.

In a game with non-transitive win probabilities (A > B > C > A), elo cannot be completely accurate.

But hang on, in a game like Gamme Logic, where time is the only measure of success, there's not going to be non-transitive win probabilities. If A > B > C (ie. A's time is better than B's time is better than C's time), there is no way C > A. It's purely linear, so Elo should likely work better in this case?

I admit to not really understanding all the complexities of Elo (nor wanting to), but this seems pretty straightforward to me.

[BarnardsStar]

I think I can guess what they mean.

In a game with non-transitive win probabilities (A > B > C > A), elo cannot be completely accurate.

But hang on, in a game like Gamme Logic, where time is the only measure of success, there's not going to be non-transitive win probabilities. If A > B > C (ie. A's time is better than B's time is better than C's time), there is no way C > A. It's purely linear, so Elo should likely work better in this case?

Sure, I’d say so, and I’d say it’s an interesting discussion to have … But “how well Elo works in this game” isn’t a factor in the power leveling rule that started this thread. It doesn’t have anything to do with why they added the decreased gain for consecutive same-player wins.

[FrankJones]

We're on page 5 of this thread, and I still am completely baffled as to what the discussion has to do with the title of the thread.

[RazorOz]

However, the elo reduction was never meant to be a "correction" mechanism for biased pairings. It was introduced as an anti-cheating mechanism, that's all, even if the wording could make it sound as if some kind of "fairness" was intended:

Power levelling
We encountered many situations of power levelling, linked to multi accounts, even if we are still fighting against such practices, it became obvious that the current system might encourage it.
To prevent this, any consecutive wins against the same opponent will reduce the amount of ELO points awarded from this victory. This reduction will be gradually higher as more consecutive wins are accumulated. This way, we encourage you to challenge other players to keep the original spirit of the ELO: show your skill at a game compared to the whole community, not just the same ones.

The thing is I've never really bought that it does anyway. If you're that determined to cheat, you're going to find ways around it. I suspect most of these can be caught with pretty simple methods tbh, because really what lengths are people going to cheat ELO, when at the end of the day it doesn't matter. Like in terms of a sense of accomplishment, how much accomplishment are you feeling if you know you cheated, and then what ELO tends to be used for, matchmaking, if you have inflated your ELO you will soon be brought back to earth.

All the while it makes it significantly worse for the vast majority of the player base. Eg. take the line there about encouraging you to challenge other players, but often you don't have a choice in this. In Arena matchmaking is private, you don't know your opponents before the game begins, and then depending on what game you're playing, even outside Arena turning down opponents because of the ELO reducer can lead to significantly higher wait times.

There's just so many different things they could have done, but the blanket implementation is pretty much the worst implementation. Eg. this could easily have been made to not to apply to Arena or tournaments where you can keep getting the same opponent(s) out of your control. Linking to time is stupid, because that is literally encouraging players not to play, you could easily link it instead to games played, which would actually be more annoying for a cheater to get around too.

[BarnardsStar]

The thing is I've never really bought that it does anyway.

Of course not, that’s why I’ve said it was stupid. And I don’t think you’ll find anybody not employed by BGA who thinks it is a good idea. (Cue Upton Sinclair: “It is difficult to get a man to understand something, when his salary depends upon his not understanding it!”)

The best thing to do would have been to just ignore the whole “power leveling” thing. Why do I care if somebody is artificially boosting their Elo when the whole point of this the cheating method is they will never play me, and if they ever did I’d either gain far more Elo than I should against a player of their actual skill, or I’d lose far less than I should? If one of these cheaters ever got high enough in the rankings to show up on any list, then I say hide them.

[menettm]

The FAQ saids "Consecutive wins: beating the same player repeatedly makes the outcome less outstanding.".
Can we have a more detailled definition in the FAQ for this case ?

I ask to play Arena for a TTR Maps game. It is often that the same player starts the game with me.
In the last tables, I win 0 ELO and he loses 0 ELO.
But I win Arena points.
What will be the effect on the current table with the same player if I lose ?