🌿ELO System Updates

[6element]

A few months ago I said that when Elo is stolen it feels bad, both because these Elo points have been earned, and because the act of "punishing for consecutive victories" somehow implies these victories weren't completely fair, that there was some kind of cheating involved, which is an insult. Well, it still feels bad.

But now I can add that being rewarded for consecutive losses also feels bad. The feeling I've got was some sort of "yeah, we know you're tired and possibly drunk and not playing well ATM, but it's Ok, good job, take a candy, keep your high Elo". That's an unexpected feeling, to be honest.

PS: A "drunk master" trophy "lost 500 points comparing to the highest Elo achieved" would probably feel like a funny joke. The current implementation ("Elo consecutive loss protection") doesn't.

[LesserJester]

I was getting so fed up with all this nonsense from BGA I actually took a few weeks off the site to get away from it and think about happier things. I hoped there might be some happy news on my return. I see there isn't, and now they're apparently doubling down making extra unwanted changes instead of just fixing their mistake. Wow. This is something special. I think I'll take a few more weeks out.

[h_illes]

Dear Players,

We appreciate you taking the time to share your feedback regarding the ELO system. We value your input and are committed to ensuring a fair and balanced ranking system for all players.

...

We're always listening to your feedback and working to make Board Game Arena the best place to play. Keep those messages coming!

But seriously. People have been explaining what's wrong and ways to set it right in simple terms. So why are you not listening? Do your own promises carry so little weight?

[RazorOz]

Here is the perfect example that showcases why the current system is terrible:

https://boardgamearena.com/table?table=508410637

So in this game I beat the #3 player at the game with virtually the same ELO as me, yet they gain more ELO for 2nd place, than I do for 1st place. How can anybody look at this and think this is a fair system?

A number of things are being shown here, firstly the system is ridiculously harsh, my overall H2H with this player is 36-29, in other words it was 50:50 7 games ago, but because I've won the last 7 I now gain 0 ELO for beating them, and this is the #3 player who has spent much time recently as the #1, it's a total nonsense.

I really don't think BGA are aware just how bad this system is and the sort of behaviour they're now encouraging. It's massively in my interests to red thumb Howie Varez and say I don't want to play with him, purely because playing with him right now is so EV- for me to do, is that the sort of behaviour BGA are now encouraging, red thumbing players because it's no longer in my interests to play them? This part is actually even worse with lower ranked players, because Howie will beat me sooner or later to break this streak, against lower ranked players you can win 15+ in a row and not win ELO off beating them ever again pretty much, as you will have always beaten them 6 times in the last 30 days.

The gradient of the system is just ludicrous. I don't think it's been understood whatsoever just how common a 7 game win streak is, if you are 50:50 vs somebody and play them a lot, you are going to get 7 game win streaks either way all the time. Which leads into the the bigger picture, that the system encourages players to decrease their activity significantly, you are encouraging inactivity, is that something BGA now actively want?

[GeoPaladin]

I really don't think BGA are aware just how bad this system is and the sort of behaviour they're now encouraging. It's massively in my interests to red thumb Howie Varez and say I don't want to play with him, purely because playing with him right now is so EV- for me to do, is that the sort of behaviour BGA are now encouraging, red thumbing players because it's no longer in my interests to play them? This part is actually even worse with lower ranked players, because Howie will beat me sooner or later to break this streak, against lower ranked players you can win 15+ in a row and not win ELO off beating them ever again pretty much, as you will have always beaten them 6 times in the last 30 days.

I'm sure you'll be glad to know that according to LordAlx's last message, the staff, in their wisdom, are removing red thumb protection from Arena so that you can't even avoid playing the same player repeatedly. We won't even have that much protection soon - we'll be forced to play the same players.

Your example is extremely helpful for showing what an asinine system this is. You did not choose to play this player in Arena. You have a fairly even record against him. Your last game against him prior to this was 4 days ago and you played ~22 games before facing him again - yet you're still being penalized as if you were exploiting him for elo.

It astounds me that the staff have failed to address this despite it being the most common complaint and despite having so many clear examples as to why it is a flawed system.

I wish they would display a more professional approach.

[nik592]

I'm sure you'll be glad to know that according to LordAlx's last message, the staff, in their wisdom, are removing red thumb protection from Arena so that you can't even avoid playing the same player repeatedly. We won't even have that much protection soon - we'll be forced to play the same players.

If you're referring to the last update in this thread, I believe that said they were actually still considering the best way to proceed with red thumb adjustments to prevent abuse in Arena (notably, where players use red thumbs to avoid the better players so they can climb the rankings by only playing weaker players - something that is also acknowledged by the community, but is widely agreed that preventing red thumbs in the Arena is NOT the way to combat this). So there may be a slight hope for better judgement there, and for now, that protection still exists - along with the more important protection of being able to avoid people who behave poorly (eg. disappearing from a game of reasonable length shortly before it is actually ended).

[h_illes]

Let me point out again that the win probability displayed at the end of the games is wrong ever since the ELO changes were introduced. And under the current system, there is no fix for the formula; for old (regular) ELO, there is a simple formula that is based on the difference between the ELO ratings for the players before the game. However, for the new version of ELO, this formula is no longer valid: the rating values do not carry sufficient information to calculate win probabilities, as the players' ratings also depend on their gaming patterns (e.,g. strong players who play less often are hit less by the win streak penalty, so they are inherently rated higher than players of the same skill who play more often, also on opponent selection - none of these would matter for regular ELO!).

In other words, by making unsolicited changes to a properly working systems, you managed to make even the win probability calculations incorrect (in addition to all the other problems). Incorrect information is displayed for ALL games, for months now - that means millions of games I think.

[KenFin1]

Let me point out again that the win probability displayed at the end of the games is wrong ever since the ELO changes were introduced. And under the current system, there is no fix for the formula; for old (regular) ELO, there is a simple formula that is based on the difference between the ELO ratings for the players before the game. However, for the new version of ELO, this formula is no longer valid: the rating values do not carry sufficient information to calculate win probabilities, as the players' ratings also depend on their gaming patterns (e.,g. strong players who play less often are hit less by the win streak penalty, so they are inherently rated higher than players of the same skill who play more often, also on opponent selection - none of these would matter for regular ELO!).

In other words, by making unsolicited changes to a properly working systems, you managed to make even the win probability calculations incorrect (in addition to all the other problems). Incorrect information is displayed for ALL games, for months now - that means millions of games I think.

Probably not millions, but a high number. They're clearly not going to do anything to change it though. Doesn't make it right or sensible, but the lack of response suggests either a complete indifference or ignorance of the criticisms and their validity. Either way they are going to keep ignoring everything that doesn't endorse their view that the changes were good ones. I still fail to see how.

[yzemaze]

@KenFin1 I wouldn’t doubt h_illes. He’s right about so many things (in this thread). The number of games is no exception:
448842913 (ID of a game played December 12th, one day before this thread was started.)
509754500 (ID of a game played yesterday.)
A large number (majority?) of IDs in between will lead to completed games. So millions it is.

I hope BGA gets their act together soon, but I’m definitely not convinced …
While they are at it they should fix the one thing that’s wrong about their Elo-implementation all along: The spelling! It’s no TLA, it’s just named for its creator Arpad Elo (né Élő Árpád Imre).

[GeoPaladin]

Dear Players,

...[snip]

It has been two months since you last addressed us. When you did so, you completely ignored all of the most common concerns and instead doubled down on the system without addressing anything meaningful.

If nothing else, the following changes need to be made:

1) Playing games with other players should reduce (and eventually remove) the penalty. It is absolutely ridiculous that I can get a 100% penalty for a win against a player I haven't played in a week, after completing 20+ games against other players in the meantime. Completing games against other players is a pretty solid sign that you're not gaming the system.

2) The person with the *win-streak* should not lose so much elo when their winstreak is finally broken. It's ridiculous that I can gain a winstreak and have no chance of winning elo against another player, but will lose full elo if they win. Reduce the K-factor instead of penalizing one player & rewarding another player who's done nothing to deserve it.

3) A winstreak of 6 in a period of a month is not that hard to achieve if you're highly ranked in a game - in fact, the penalty is exceptionally punishing if you get a mere 3 wins in a row, which even an average player should hit more than a few times. The numbers should be substantially higher before penalties come into effect.

4) Arena involves blind-matching making and should not have the same penalties attached.

Finally, remember that the elo system is already meant to account for repeated wins. If I'm expected to win against a player 6 times for every 1 win they get against me, then my elo will stay approximately the same as long as I meet that standard. It is not only unnecessary but counterproductive to invent an additional handicap.

Finally, remember to focus on having fun, challenging yourself, and enjoying the competition!

ELO isn't a prize or a badge of honor. ELO ratings serve as a tool to gauge relative skill levels and find well-matched opponents, not a measure of personal worth. Focus on enjoying the games, learning from your opponents, and refining your strategies to become a true board game master!

You say that elo is a measure of skill and not a badge, but this is wrong on both counts. First, a measure of skill *is* something to be proud of and strive for - I would argue it's arrogant to tell others not to do so nor to enjoy it. Setting a goal to improve your elo (and necessarily yourself as a player) is a worthy goal.

Second, this system has warped elo with an arbitrary handicap, making it a less effective measure of skill in the process. The system was designed to work as-is. Your arbitrary handicap has unbalanced it, as many other commenters have explained to you over 85 pages of content.

Please fix this.