Arena scoring question

[euklid314]

Is a persons league considered in arena calculations? Ex. If I am elite and beat another elite player w/ a 1560 score do I get the same pts as beating a bronze player w/ a hidden 1560 score?

The Arena score is independent of the leagues or the Elo values of the players. Thus beating a 1560 player will always give you the same AES change.

[shinuito]

I see a lot of problems in your suggestions.

One problem I see is that you’re interpreting what I said as suggestions as opposed to merely criticism of the current system in comparison to other ideas.

I don’t believe I’m suggesting any of those would be a better replacement? Rather they’re contrasting other systems in other places.

One can equally debate whether combining live games and turn based correspondence games into one league is really such a good idea anyway, other than to just simplify the system. It’s not necessarily in the systems best interest.

[shinuito]

I suppose if I'm thinking a bit more, and trying to collect thoughts on it, it's not that the current system doesn't work; it's fine, it works for correspondence and live games, it works for 2 player and multiplayer games, it's a good way to get competitive games particularly at the start of a new season, as mentioned there's no penalty to stopping playing, so it's not soley about who played last or who can play the most games.

I think my issues are with the metagame around the arena and the pairing, for me personally in smaller two player strategy games. It is advantageous to hit elite league with a decent number of points and then stop playing. I think it's a reasonably good meta strategy, and then play only if you need to retake the lead.

In the smaller sub communities of two player abstract games, it can be very difficult or impossible to find live games in a certain game, sometimes difficult to even find turn based ones, and having a meta strategy which pushes people to stop playing to hold their place (first second third - whatever has a badge) isn't that great.

Then with the pairing, it might be unavoidable to play the same players over and over sometimes with small pools. Other times though constantly pairing an elite league player with bronze or silver players, particularly if this player is quite strong, probably doesn't give other players a chance to move up the leagues at all. It might not even matter how many points the elite player has if they can just blockade the elite league and keep people at silver/bronze 0.

I think it's just some things to consider. Maybe the perspective is a bit narrow, since I primarily like two player abstracts and strategy games, with or without some randomness. Anyway hopefully that might compact things a little bit from the other posts.

[ChiefPointThief]

I suppose if I'm thinking a bit more, and trying to collect thoughts on it, it's not that the current system doesn't work; it's fine, it works for correspondence and live games, it works for 2 player and multiplayer games, it's a good way to get competitive games particularly at the start of a new season, as mentioned there's no penalty to stopping playing, so it's not soley about who played last or who can play the most games.

I think my issues are with the metagame around the arena and the pairing, for me personally in smaller two player strategy games. It is advantageous to hit elite league with a decent number of points and then stop playing. I think it's a reasonably good meta strategy, and then play only if you need to retake the lead.

In the smaller sub communities of two player abstract games, it can be very difficult or impossible to find live games in a certain game, sometimes difficult to even find turn based ones, and having a meta strategy which pushes people to stop playing to hold their place (first second third - whatever has a badge) isn't that great.

Then with the pairing, it might be unavoidable to play the same players over and over sometimes with small pools. Other times though constantly pairing an elite league player with bronze or silver players, particularly if this player is quite strong, probably doesn't give other players a chance to move up the leagues at all. It might not even matter how many points the elite player has if they can just blockade the elite league and keep people at silver/bronze 0.

I think it's just some things to consider. Maybe the perspective is a bit narrow, since I primarily like two player abstracts and strategy games, with or without some randomness. Anyway hopefully that might compact things a little bit from the other posts.

[ChiefPointThief]

In the smaller sub communities of two player abstract games, it can be very difficult or impossible to find live games in a certain game, sometimes difficult to even find turn based ones, and having a meta strategy which pushes people to stop playing to hold their place (first second third - whatever has a badge) isn't that great.

Then with the pairing, it might be unavoidable to play the same players over and over sometimes with small pools. Other times though constantly pairing an elite league player with bronze or silver players, particularly if this player is quite strong, probably doesn't give other players a chance to move up the leagues at all. It might not even matter how many points the elite player has if they can just blockade the elite league and keep people at silver/bronze 0.

I was in 1st place in tic tac match by about 60pts. I had a great run to get there. I beat the world elo player 3x straight (they were #2 in arena) as well as other wins. The smart thing to do if I want to win the competition would’ve been to stop. Like you said games are rare to come by and I enjoy playing so I didn’t. I went on a losing streak and my wins got wiped away very quickly. Yesterday I played someone who was playing solid but had a lower rank. After 4 games I didn’t renter. Not because I was tired or wasn’t having fun. We played 4 games in which I went .500. They were getting 3x as much pts in a heavy luck base game so playing more would just dig a deeper whole. Imo this is problematic.

In response to your 2nd paragraph I realized I was keeping several ppl in the field down last season. I recently took two losses in a different game that I was leading arena. But before that I was ahead of the field by 100pts playing in 9x fewer games and only 1 match against top 10 opponents compared to over 80 last season.

[ExaltedAngel]

We played 4 games in which I went .500. They were getting 3x as much pts in a heavy luck base game so playing more would just dig a deeper whole. Imo this is problematic.

If you win and lose the same amount of games against someone who is lower rated than you, either you are too high, or they are too low. So you lose more points than what you win and vice versa and that makes sense and is perfectly reasonable. This is how Elo works and shouldn't be changed.

The problem, however, kind of exists, and it's because Elo isn't very good for heavy luck-based games. Winrate of players in such games is a lot more compressed around 50% than in heavy skill-based games in which top players could have 90+ or even close to 100% winrate if paired randomly with players from the whole spectrum of skill. This usually doesn't happen because generally people are somehow matched based on their skill level, since heavy skill-based games played between players of a very different level are just not interesting for both parts (hello BGA Arena matchmaking?), so top players' winrates are actually much more reasonable than that, just because they usually play their own kind.

i.e. Magnus Carlsen has 63.82% winrate, but if you would randomly match him with users of * of any level, he'd probably have 99.99%.
Nothing of this can apply to heavy luck-based games, where the vantage of top players is minimal and having anything higher than a few points of winrate more than 50% (depends on how luck-based is the game of course) is already sign of very solid predominance.

What's the solution? Idk, maybe just don't take Elo so seriously for that kind of games. Especially because it still is relevant, but the more you thin the edge of good players on bad or mediocre ones, the bigger the sample needs to be for Elo to accurately reflect skill levels. And I didn't actually study the math behind that, but I'm pretty sure it's not a linear growth, but more of an exponential one.

*moderator edit: please do not name other game websites

[ExaltedAngel]

What I said is pretty obvious when looking at the ratings of top players of different games. For example Ticket to Ride and Lucky Numbers.
I hope we all agree that, despite still not being a 50% coinflip game, Lucky Numbers has an higher luck component than Ticket to Ride. This makes the players to be less spread on the Elo range, thus ratings for top players are lower, both in arena and permanent Elo.

This btw exposes how bad are the arbitrary titles that come from Elo rating (good player, expert, etc) since they are the same for every game but they are not at all accurate. A 500 Elo Lucky numbers player is definitely a master, while a 500 Elo TTR one is still nothing more than a good player.

I'm not sure I have good examples of even more luck-based or more skill-based games here on BGA, but if you do and go check these values (beware of games that not played enough to make the sample relevant), you'll very likely see the same trend.

[ChiefPointThief]

We played 4 games in which I went .500. They were getting 3x as much pts in a heavy luck base game so playing more would just dig a deeper whole. Imo this is problematic.

If you win and lose the same amount of games against someone who is lower rated than you, either you are too high, or they are too low. So you lose more points than what you win and vice versa and that makes sense and is perfectly reasonable. This is how Elo works and shouldn't be changed.

The problem, however, kind of exists, and it's because Elo isn't very good for heavy luck-based games. Winrate of players in such games is a lot more compressed around 50% than in heavy skill-based games in which top players could have 90+ or even close to 100% winrate if paired randomly with players from the whole spectrum of skill. This usually doesn't happen because generally people are somehow matched based on their skill level, since heavy skill-based games played between players of a very different level are just not interesting for both parts (hello BGA Arena matchmaking?), so top players' winrates are actually much more reasonable than that, just because they usually play their own kind.

i.e. Magnus Carlsen has 63.82% winrate, but if you would randomly match him with users of * of any level, he'd probably have 99.99%.
Nothing of this can apply to heavy luck-based games, where the vantage of top players is minimal and having anything higher than a few points of winrate more than 50% (depends on how luck-based is the game of course) is already sign of very solid predominance.

What's the solution? Idk, maybe just don't take Elo so seriously for that kind of games. Especially because it still is relevant, but the more you thin the edge of good players on bad or mediocre ones, the bigger the sample needs to be for Elo to accurately reflect skill levels. And I didn't actually study the math behind that, but I'm pretty sure it's not a linear growth, but more of an exponential one.

*moderator edit: please do not name other game websites

The purpose of me sharing that story was more about shinuito's point that players at the top are incentivized to stop playing and I feel like players should always want to play. Rather than just playing and winning I feel like playing in arena can come down to manipulating the system. Do I want to play early and get a quick lead? Play late as a strong player so I can get more pts for my victories on the way up. Stop when I get to the top even though I would love to keep playing because there are no more matches beneficial to me. Some players will even block ppl so the aren't paired w/ them.

You say the system shouldn't change but yet acknowledge it is very flawed for luck based game. The don't take it serious counter argument is bad considering that applies to all board game related topics here. Heavy luck based games range closer to 50% like you said but I'm expected to win 100% of my matches in the example above to see a gain. If you increase the sample size I still need somewhere around 77% before I could get a small gain. That isn't ideal for a game that ranges close to 50%.

[buu441]

Who would have thunk you need more that 50% winrate to gain elo? *insert surprised pikachu face*

[ChiefPointThief]

This btw exposes how bad are the arbitrary titles that come from Elo rating (good player, expert, etc) since they are the same for every game but they are not at all accurate. A 500 Elo Lucky numbers player is definitely a master, while a 500 Elo TTR one is still nothing more than a good player.

I completely agree on this point. For example I'm a mid 500s player in stone age which is nothing spectacular. I'm ranked in the 300s. But there are other games w/ less players that I've reached world elo and 600 elo. It isn't really possible to reach 700 because of the available elo in the field so the two aren't comparable. It definitely means nothing for co op games. For team games it should be much lower.