Fundamental issue with the elo system and repeated wins

[MasterMystery]

(for simplicity 2 player)
Elo gives a win % for each player, A and B. So if the elo difference is such that A is supposed to win 90% of the games and B 10% if 10 games are played and A indeed wins 9 the elo of the two players should stay roughly the same. The only uncertainty/deviation from this state occurs due to the timeing of the win of B within the 10 game series. The closer the elo and equal the expected win % the less this recency bias.

But now we come to the point this does not work on BGA, as you get fewer elo points the more wins you have against a player (repeated wins). This fundamentally is meant to avoid elo farming (and noob stomping). But it also means that you can never get the elo back you lost against a player which is much lower in elo, if you loose once.

Its not that i personally care in that particular case (as its even more annoying its not really wins as the other player just passed or early gave up and i don't want those pts anyway), but it is a fundamental issue if games are played out normally. Here is my record as an example: https://boardgamearena.com/gamestats?pl ... ateStats=1
At game https://boardgamearena.com/table?table=715893229 the expected wins are 85%:15%, which is 3 in 20 rounding easisest to 3 in 21 so 1:6 and the higher ranked looses (which results in -17 elo). So given the higher ranked should be expected to win 6 games the elo should equal, its 3+3+2+2+2+1=13 not 17 due to the repeated penalty. After 6 games the penalty is 100% so no elo is gained at all.

This fundamentally jeopadizes the elo system:
- it looses part of its self correcting character
- punishes higher elo players ratings and therefore reducing their elo, effecting overall the main info what elo should provide , an indication of balance and win%
-

Nevertheless, I think that a general repeated win elo win reduction has a merrit for the overall "ladder" and noob stomping, but not in such a harsh way. It just should not go already to zero at 6 wins if 6 wins is the expected average win rate before a loss. I am not sure if the number 6 is based on elo difference or just always.

This mainly concerns games with low playerbase , but in the end all the above mentioned holds true in any case, just the likelyhood of facing similar players is lower.

[6element]

There is 100+ page discussion on the topic https://boardgamearena.com/forum/viewtopic.php?t=33785 where BGA team representative asked for feedback and the feedback (mostly negative) has been ignored. So, you're not alone but likely nothing will be fixed with the current BGA team / attitude towards customers.

[Cos-]

I didn't know BGA makes any adjustments to Elo based on factors other than the Elo calculation itself. While that's actually significantly played for >2 player games and games with luck (Elo was designed for a 2-player no luck game and that's what it works best for), it's still a neutral and well understood system. If BGA is playing weird games with it for any purpose _other_ than record a person's relative skill at a game, that undermines the system completely. What's the point of it at all?

I know that some people view Elo here as a score they should strive to maximize, but this just plays into that twisted view by undermining Elo for everyone else and making it less meaningful in general without enhancing its value as a score for the people who care about it. It seems really really misguided to me.

From the link you gave, I also learned that they have "Elo decay" for people who haven't played a ranked game after an arena season? *? This is really nuts. It means we don't actually have Elo on BGA after all, and they're deceiving us by calling it that.

*Moderator edit: please refrain from swearing on BGA, this is a family-friendly platform.

[Meeplelowda]

From the link you gave, I also learned that they have "Elo decay" for people who haven't played a ranked game after an arena season? *? This is really nuts. It means we don't actually have Elo on BGA after all, and they're deceiving us by calling it that.

I don't think they ever actually implemented that. Just like they never implemented the change announced in that thread that red thumbs would not affect Arena.

[danielvictor]

I have been playing two different players many times in two-player games [at Forest shuffle] - because in both cases we both wanted to play.I did notice that I wasn't getting any ELO points when I won,but it doesn't really bother me - I'll just get them back in games against other players.

[Cos-]

now we come to the point this does not work on BGA, as you get fewer elo points the more wins you have against a player (repeated wins). This fundamentally is meant to avoid elo farming (and noob stomping). But it also means that you can never get the elo back you lost against a player which is much lower in elo, if you loose once.

In some games, I sometimes get repeatedly matched with the same player over and over, specifically when I try to _avoid_ "noob-stomping" by setting my tables to not allow newbies and beginners. I just recently posted about this because it's been frustrating:

https://boardgamearena.com/forum/viewtopic.php?t=44541

[FrankJones]

I hate this.

3 wins in a row against the same player is NOT "Elo Farming". I usually offer a rematch, especially if the game favors player 1, that way we each get a chance as player 1.

If I win both, and then I host an open game, and that player joins, I will get an Elo penalty if I win.

This is true even if I have played against other player in between these 3 wins against that same player!

So, I could win twice against one opponent. Play 20 other games against other opponents. Then, play a 3rd game against the original opponent... and get an Elo penalty, supposedly to avoid ELo farming?

We could red-thumb someone after 2 wins, but that's also a sub-optimal solution.

[6element]

I didn't know BGA makes any adjustments to Elo based on factors other than the Elo calculation itself.

Well, 100 point "Elo floor" is well known hack and it artificially increases Elo points in the system.

On the other hand, "Elo punishment for quitting" is actually artificially removing Elo points from the system.

From the link you gave, I also learned that they have "Elo decay" for people who haven't played a ranked game after an arena season? WTF? This is really nuts. It means we don't actually have Elo on BGA after all, and they're deceiving us by calling it that.

There is no Elo decay, they realized it'd be causing way too many issues while not solving anything, so it has been abandoned / never activated. This is good.

[Cos-]

I didn't know BGA makes any adjustments to Elo based on factors other than the Elo calculation itself.

Well, 100 point "Elo floor" is well known hack and it artificially increases Elo points in the system.

On the other hand, "Elo punishment for quitting" is actually artificially removing Elo points from the system.

From the link you gave, I also learned that they have "Elo decay" for people who haven't played a ranked game after an arena season? WTF? This is really nuts. It means we don't actually have Elo on BGA after all, and they're deceiving us by calling it that.

There is no Elo decay, they realized it'd be causing way too many issues while not solving anything, so it has been abandoned / never activated. This is good.

I'm glad they didn't implement that Elo decay, it's unfortunate they even considered it to begin with. But an Elo penalty for quitting? That also undermines the point of Elo. Yuck.

[RazorOz]

It's arguably even worse for games primarily not played at 2 player, because you are even more likely to have repeat opponents.

I'd echo what the majority have said, it's nonsense to be applied virtually immediately, and the reducer to be so exponentially harsh, you could be 50:50 to win and win 7 consecutive games vs players regularly.

If it has to exist it should be much less harsh, and probably shouldn't even kick in at all until about 7 wins, let out max out at 100% then. But really I'd prefer it just didn't exist at all. One of the problems with this system is it rewards not playing, and punishes frequent players, which should kind of go against what the site wants, I've noticed in some games the real best players aren't #1 anymore, because they play regularly and get bodied by the reducer, whereas the players who maintain #1 play a few games every 60 days so it doesn't affect them. That's not really encouraging a healthy system, because if everyone did that there would be no site, as games wouldn't start.

It's also kind of ironic, because in the long thread on this, it's stated one of their goals was to stop players seeing ELO as a trophy and sitting on #1, but instead they've created a system that does exactly that.

For the record I also think ELO decay is a bad idea too, problem is it will be very hard to get right. Say someone is the best player at a game, but then goes travelling for 6 months and doesn't play, if you decay their ELO it will be silly, and defy the point of ELO in the first place, to help people find games of players of roughly equal level, if that player then ends up having to work their way from the bottom again.