BGA has a serious bully problem

[Ceaseless]

Your opinion (key word, opinion) is that Elo is broken because ratings fluctuate a lot in high-variance games. My opinion is that there is no issue here, because ratings fluctuations are natural in high-variance games.

I think this is the heart of the matter. You can look at Elo definitionally (it tells you how a player has done recently for some definition of “recent”), or predictively (it should tell you how a player is going to do). For pure strategy games like chess, these are much the same thing. For luck-based games, not so much.

If you want Elo to function more predictively, lowering the k-factor would give you a longer-term average. But bear in mind if we were to cut the k-factor in half, you’d get half as many points from each win (and lose half as many to each loss). Would you ever have hit 550 in such a case? I’d say not. Extremely generously, you’d have been lucky to hit 300. And guess what? Instead of bobbing up and down by 20 points out of 550 with each win and loss, you’d be bobbing and down by 10 points out of 300. And I think you’d still be unhappy with that, since everybody else would be bobbing as well and sometimes overtaking you and sometimes not, and you’d still have the same issues with that.

I just don’t think Elo is best-suited for what you are hoping for from such a system when it comes to high-variance games.

You quoted the post you wanted to highlight, but it only tags the user you quoted. You should probably quote the person you're replying to as well, both for clarity to who you are speaking to, and to ensure they actually notice your post and don't end up skimming past it.

[Ceaseless]

Once again, my settings are set for good players and above.

Winning 68% of the time against players in the 200s Elo does not make someone a great player at that game.

How about the fact that I was the 1st of like 7 people to reach expert at the game, Im the only player to reach 550, or that I won the 1st arena and only didn't play in following seasons because I understand how flawed arena is, or that I've been the world elo leader for the majority of the time the game has existed on the site and theres close to 100,000 players? Do any of these things qualify me?

"good players" on this site are 200 Elo and above. Despite the title of that Elo range, there are a lot of low-skill players rated in the 200s.

This analysis is game dependent and simply false which is part of the point of this discussion. Applying the same rating system/titles across all games is flawed.

A mediocre player could play entirely against beginners (in the 0-100 Elo range) and have a high win percentage that would not correlate well with that player's actual skill.

A player could play entirely against experts and masters and have a 50% win rate which again would not correlate with that player's actual skill.

Yeah. This COULD happen, but didnt, so what relevance does this have pertaining to what I said?

Nor does it mean that player should have a higher Elo than someone who wins 50% of the time against Elite players.

I set the filter for most of my games to "Strong players" and above. In some cases, "Expert" and above. Thus, I'll have a lower win% in the long run, but a high Elo.

To put it another way: If I play a lot of games against the highest rated "Earth" player, we will, on average, perhaps have a 50% win rate each, while still maintaining Elos in the high 600s.

I gave you a very specific example and you decided to add a bunch of irrelevant and inapplicable things. There is currently only one expert player in the game I referred to and like 0.6% of the entire community is strong or above. Why are you comparing it to earth which has a luck rating of 1?

You talk about your settings but did not mention your opponent's settings. For all we know they're playing 300+ players for their games, and that's why they can maintain a lower win percentage than you while having the higher rating.

Well we do know because, If you only tried to play against strong players and above in this game you wouldnt find many tables as only 0.6% of players even fall into that category, they play in arena and it is impossible to filter in this way, & because their play history is public information

For the record I will eventually climb back above this player but a proper rating system shouldnt seesaw your strength level. I shouldve never fallen below them in the 1st place and it shouldnt be so easy to do so. I played captain flip arena season 1 and competed w/ another great player for the top spot. We both played at an elite level throughout the season but took drastic falls at some point playing game by game trying to get a few more points than the other. Having a system in which you can play at an elite level over hundreds and hundreds of games over months and to lose all of that within 20 minutes on a bad streak in a 4 luck rating game is insane & an improper rating system. I've already had this debate with you on several threads and shown you several examples and data. These convos usually end with "well dont play luck games" lol. Bga COULD fix this issue but I've accepted that it will never happen.

You're the one who decided to make your skill about a 68% win rate. Since win rate is a joke, don't be surprised when everyone else laughs. You saying "how can you ignore all of these other stats I didn't bring up" is rather silly. You didn't mention them, you shouldn't expect others to know about them. That goes for several of the things you refer to. If you leave out important information, you have no one to blame but yourself if people aren't using it.

You say "yes, this player could have played against top raters, but didn't". How are we to know? They apparently have a similar rating to yours, according to you. You talk about the user's profile being public, but given you're presently ranked 43rd it's not like I can say with certainty who to even check. We're not going to know things like if they only play Arena, etc.

Players have bad streaks and lose a lot of rating. That's normal, it happens even in games like Chess. There's not some magical solution to get around what happens when you perform poorly.

[Ceaseless]

Far not well in co-op and high luck factor games. Too easy to level up as beginner and too easy to fall down from expert level.

That’s because in those games there isn’t as much difference between a beginner and a top player. Your problem is with the game, not the system.

You could ask that the k factor be different in high luck games, so that rankings don’t move as easily, but I don’t see the justification why. That would be like hiding the true nature of the game, making the better players appear more in control than they actually are.

Counterexaple: Texas Hold'em Poker. Luck factor is very high but there IS mutch difference between beginner and top player. In normal ranking system (money) top players looking for play with less experienced players (they called fish) but with present ELO system is better to play with similar or higher ranked player to get higher ranked.

I never did get a reply describing what was meant here either.

Define money being used as a ranking system, I have a few different possibilities in my head that would get different responses, so I'd like to pin this detail down first to set the context of the discussion.

[BarnardsStar]

You quoted the post you wanted to highlight, but it only tags the user you quoted. You should probably quote the person you're replying to as well, both for clarity to who you are speaking to, and to ensure they actually notice your post and don't end up skimming past it.

A fair point. Hope people don’t object to me repeating myself here but tagging the user I’m actually addressing:

No bga doesn’t need to create 1400 different rating systems but they could create a handful.

I think this is the heart of the matter: You can look at Elo definitionally (it tells you how a player has done recently for some definition of “recent”), or predictively (it should tell you how a player is going to do). For pure strategy games like chess, these are much the same thing. For luck-based games, not so much.

If you want Elo to function more predictively, lowering the k-factor would give you a longer-term average. But bear in mind if we were to cut the k-factor in half, you’d get half as many points from each win (and lose half as many to each loss). Would you ever have hit 550 in such a case? I’d say not. Extremely generously, you’d have been lucky to hit 300. And guess what? Instead of bobbing up and down by 20 points out of 550 with each win and loss, you’d be bobbing and down by 10 points out of 300. And I think you’d still be unhappy with that, since everybody else would be bobbing as well and sometimes overtaking you and sometimes not, and you’d still have the same issues with that.

I just don’t think Elo is best-suited for what you are hoping for from such a system when it comes to high-variance games.

[stillframe]

I think this is the heart of the matter: You can look at Elo definitionally (it tells you how a player has done recently for some definition of “recent”), or predictively (it should tell you how a player is going to do). For pure strategy games like chess, these are much the same thing. For luck-based games, not so much.

It still has predictive value. The only difference is that for higher luck games, it takes more future plays for a prediction to be meaningful.

[tbhp]

Well we do know because, If you only tried to play against strong players and above in this game you wouldnt find many tables as only 0.6% of players even fall into that category, they play in arena and it is impossible to filter in this way, & because their play history is public information

Then their 50% winrate is probably due to the time when they weren't very good at the game.

If you play 1,000 games badly, but then all of a sudden you start playing seriously and rise to expert elo in the span of 100 games, then your overall winrate will mostly be reflective of the period when you weren't good at the game. That doesn't mean your current elo is incorrect.

I played captain flip arena season 1 and competed w/ another great player for the top spot. We both played at an elite level throughout the season but took drastic falls at some point playing game by game trying to get a few more points than the other. Having a system in which you can play at an elite level over hundreds and hundreds of games over months and to lose all of that within 20 minutes on a bad streak in a 4 luck rating game is insane & an improper rating system

If you are both elite players then what else do you expect to differentiate you, if not a few games here and there?

Your problem is not with the system, it's that you expect to be handed the title freely despite not achieving anything more than your opponent.

If the game was chess, by the same logic, you would still be complaining that your bad streak made you lose your position.

[tbhp]

Your opinion (key word, opinion) is that Elo is broken because ratings fluctuate a lot in high-variance games. My opinion is that there is no issue here, because ratings fluctuations are natural in high-variance games.

If you want Elo to function more predictively, lowering the k-factor would give you a longer-term average. But bear in mind if we were to cut the k-factor in half, you’d get half as many points from each win (and lose half as many to each loss). Would you ever have hit 550 in such a case? I’d say not. Extremely generously, you’d have been lucky to hit 300. And guess what? Instead of bobbing up and down by 20 points out of 550 with each win and loss, you’d be bobbing and down by 10 points out of 300. And I think you’d still be unhappy with that, since everybody else would be bobbing as well and sometimes overtaking you and sometimes not, and you’d still have the same issues with that.

I felt something was off when reading the linked thread, and you put it into words well.

The higher the k-factor, the faster the gaps in elo will close, but what the post seems to forget is that in arena people don't start with an elo representative of their skill, they all start with the same elo.

From that point onwards, the uncertainty works in both directions. You are just as likely to benefit from luck when you go up as when you go down. It's only logical that a gap of 100 elo points can be closed more easily with a higher k-factor, since it can be created more easily in the first place.

[ChiefPointThief]

May I ask, why do you play players with much lower Elo than you, thereby putting your Elo at risk in a high-variance game?

There are only 16 players at that game rated 400 or higher. With 2 million total games played. Only one player is higher than 472, and nobody is higher than 500.

You answered your own question already.

Never said I went from 550 to 380 in 20 minutes . You keep making up a lot of things.

Ummm.

posted by you literally on this page:
"We both played at an elite level throughout the season but took drastic falls at some point playing game by game trying to get a few more points than the other. Having a system in which you can play at an elite level over hundreds and hundreds of games over months and to lose all of that within 20 minutes on a bad streak in a 4 luck rating game is insane & an improper rating system."

You proceeded to quote me not saying what you said I said I'm confused

[ChiefPointThief]

You've misquoted me numerous times as well as stated things that have nothing to do with my position so I will attempt again to make my position to you clear. I believe the probabilities currently used shouldn't be applied universally amongst all games because that is faulty. I believe that just because the ceiling is lower for maximum player strength in a game due to its luck factor that that doesn't mean having a rating system that will gravitate players towards the same rank rather easily is acceptable. There may be a smaller gap in between what is considered a great player in a luck base game (10% better win rate over another player vs same field) but what they are accomplishing is great for what they do. Just like a basketball player shooting 90% free throws doesn't mean they are a superior athlete to a baseball player that bats .350. I think that the rating system should still properly reflect the difference amongst players in luck games. If one player is able to use methods to win 5% 0r 10% more games within the same parameters than another player in a game that is inherently closer to 50/50 that is a huge margin.

Im taking the time because I think you may actually want to have a conversation. Lets start here. Currently the way probabilities are calculated for team games is faulty. Can we agree on that? It uses head to head matchups and doesn't consider your partner. Meaning if I have a higher elo than both my opponents I may be ranked 70% over one and 80% over another. It doesn't count them as a team and doesn't consider who my partner is (who could be by far the worst player at the table). I play a few team games. I will use spades in this example. I had a spades community at one time and have partnered or played against a lot of players. I know who is great, good, and average skill wise. But because they are playing in a faulty system with false probabilities their ratings seesaw all the time. Their skills never change. It is the system being used to rank them that is inaccurate. It constantly takes elo from stronger players and gives it to weaker players keeping the field closer than it really is. Even more so for a game like bang but I will fast forward.

Based on what I've witnessed in team games these false probabilities lead to inaccurate rankings. Elo was designed for chess. A non luck game. How can the same probabilities be applied to a game with no luck and a game with 4 luck rating? If a player is projected to win 60% of the time if they have +60 elo over player b with no luck factor than how can it also be 60% in a game that has 4 luck factor and so many variables? So as an example with 20 k factor a player would lose 12 elo if win probability is 60% but considering the luck factor of the game should reduce it to different rates like 11, 10.5, 10.25.

[ChiefPointThief]

I played captain flip arena season 1 and competed w/ another great player for the top spot. We both played at an elite level throughout the season but took drastic falls at some point playing game by game trying to get a few more points than the other. Having a system in which you can play at an elite level over hundreds and hundreds of games over months and to lose all of that within 20 minutes on a bad streak in a 4 luck rating game is insane & an improper rating system

If you are both elite players then what else do you expect to differentiate you, if not a few games here and there?

Your problem is not with the system, it's that you expect to be handed the title freely despite not achieving anything more than your opponent.

If the game was chess, by the same logic, you would still be complaining that your bad streak made you lose your position.

Maybe you misunderstood me? Im not sure. But my issue wasn't either of us falling to the other it is going from 2nd place to 400th in the blink of an eye after hundreds of games being played at a high level. There is usually only a few players per season to reach 1800s out of 4000. Because you are above the majority of the field your rates to break even become very high 75%+. You can maintain this over hundreds of games which is elite. And that can be wiped away with one bad 10 game streak. If you lose 7of10 games at around 35aes that is 230aes. Considering captain flip is 2 min game you can do that in 20 min. Falling to hundreds of players that you outperformed by a large margin throughout the entire season.

I don't want to be handed anything. In fact, I want the opposite. I believe bga made the "competitive" mode less competitive than normal games to appease to a larger audience. If you approve of a competitive format that allows players who performed way worse than other players but finish higher then we can agree to disagree.