Luck vs Skill in Arena

[Vandaley]

Every Arena season change I see many comments that either defend or deride a change with some version of this argument:

"This game has too much luck when (in this version, or with this amount of players, or something) and for arena the best version is (insert something that is usually a low amount of players) as it is true measure of skill and not luck."

I am interested in if this is a common take. I really dislike this take as it acts like navigating luck is not itself a skill, and seems to approach all board games as needing to be something more akin to Chess or something that has no luck.

For me the element of luck in many games is important and fun, and as long as everyone is playing by the same rules then arena is still going to be a competition that measures who is the best at the game. But it is not clear if I am the person in the minority opinion or if those who constantly voice this view in arena announcements are.

[Lumin_S]

Navigating luck is of course a skill.

But I think you might be missing the difference between navigable luck, where the game has enough skill for a stronger player to overcome a luckier, weaker player -- and non-navigable luck, where a stronger player can't overcome the weaker player being the beneficiary of luck.

The game where both you and I have a solid Elo is 4p Saint Petersburg, and I think I can make a good example there. Let's say there's a "Good" (200-300 Elo) player and 3 "Strong" or "Expert" players. If the "Good" player gets the 1st round Aristocrat flip as 18-12-7-4, they have an inappropriately high chance at winning just due to that flip. I consider that non-navigable luck and will probably be playing for 2nd place unless I spot them make a significant blunder.

I think reducing non-navigable luck in Arena should be done whenever possible, and it'll be very hard to convince me otherwise.

I also think reducing luck is generally a good thing for competitive formats where we want to find the best player(s). But less necessary if the luck is navigable as opposed to non-navigable.

[Vandaley]

Seeing as how I hate that St Pete went from 4 to 2 player in Arena it is not the point you think it is. I am perfectly happy having that happen occasionally, and it is only something that happens occasionally.

[frogstar_A]

I prefer arena formats that are more fun rather than have minimum luck, so for example sushi Go with 4 players not 2

[RicardoRix]

You make a very good point.

You can't please all the people all the time, so you're always going to have complainers.

The (probably) bigger and over-riding reason for lower player count settings is because it's easier and quicker to fill tables. You maybe beating a broken drum.

What is nice, is when games have a good rotation of settings like 7-wonders, but even then it could be better with wider player counts. Rather than a simple vote for next season, it would be good to get a rotation for the next 4 seasons say to catch a wider audience. Players will know to wait a season if they don't like it, or try a different game. The 'vote for next season' always feels like the players weren't listened to if the settings aren't their own personal choice, and so there will always be a chorus of complainers come a new season.

Lumen's argument doesn't really count. Arena is played over many games, so 1 player might get lucky 1 game, but over the course of a season the luck should even out.

The only problematic luck in Arena could be, that it might be considered you need a lucky run of consecutive games before you park your high position. The higher the player base count, the more likely this is to come into effect.

[dschingis27]

Arena is played over many games, so 1 player might get lucky 1 game, but over the course of a season the luck should even out.

"Over the course of a season" sounds very euphemistic here. With the current setting, in 2p games, at most only the last 50 games count, the rest gets forgotten. The most recent games you play have a tremendous influence on your score. For games like Can't Stop, your score is more or less based on the results of your last one or two hours, not the season at all.

The only problematic luck in Arena could be, that it might be considered you need a lucky run of consecutive games before you park your high position. The higher the player base count, the more likely this is to come into effect.

Exactly. But of course the best players are the most likely to get such a streak.

[Lumin_S]

Lumen's argument doesn't really count. Arena is played over many games, so 1 player might get lucky 1 game, but over the course of a season the luck should even out.

My argument counts, even if you think it doesn't. I disagree with your implication that, regardless of luck, the length of an Arena season is sufficient to reach expected long-term outcomes*. And for reasons related to variance and the Elo/AES systems, higher luck leads to less favorable Arena results. Elo-style systems (like both BGA Elo and AES) are known to be a poor rating systems for multiplayer games, and many arena formats are >2 players.

One reason Elo and AES are bad is that they are dependent on order of results, and therefore manipulatable: a player will end up with a higher rating by losing 2 games then winning 2 games, as opposed to the opposite order. This can be pretty easily done by turn-based Arena players who have multiple games finishing at similar times.

Additionally, Elo and AES fail for multiplayer games because they assume the game is multiple independent 1v1 matchups, and assign a rating based on that assumption. Most multiplayer games on BGA just don't work that way! Most winning strategies for a N-player game don't rely on N-1 independent victories, and many times a worse final position can be attributed to causes outside the player's control - but the rating system will just assume you won or lost more 1v1 match(es), which often tells a more extreme story than what actually happened**.

Here's what I believe:

• If there is greater luck in a game, then a consequence of that is higher variance within the win-loss results. That is, a player who wins X% of their games in the long term is less likely to win (X+/-Y)% of their games in the short- and medium-term if there is higher variance in win-loss results.

• With higher variance in win-loss results, there is a wider range of possible outcomes for every player's rating by the end of the season, and a lower probability that their rating will be close to representing their true skill level. This is exacerbated by the problems with Elo and AES, which cause ratings to have even more variance.

• With a wider range of outcomes for each player, it is less likely that the Arena standings reflect a representative order of players' skills. And I'm pretty confident that the purpose of Arena standings' existence is to reflect the order of players' skills.

To summarize, even if luck "evens out in the long run", the rating systems here are affected differently and will be less accurate with higher luck. I also have doubts that an Arena season is truly a "long run" for most Arena games.

----------------------------------
*I agree with your point specifically for 2-player games where everyone is playing very large numbers of games every Arena season.
**BGA does scale down the sum of Elo adjustments in multiplayer games, but I claim it's not enough to address this issue.

[RicardoRix]

[My argument counts, even if you think it doesn't. I disagree with your implication that, regardless of luck, the length of an Arena season is sufficient to reach expected long-term outcomes*. And for reasons related to variance and the Elo/AES systems, higher luck leads to less favorable Arena results. Elo-style systems (like both BGA Elo and AES) are known to be a poor rating systems for multiplayer games, and many arena formats are >2 players.

I argue against the specific scenario which involves just 1 game. I highlight that multiple games are going to diminish that effect.

One reason Elo and AES are bad is that they are dependent on order of results, and therefore manipulatable: a player will end up with a higher rating by losing 2 games then winning 2 games, as opposed to the opposite order. This can be pretty easily done by turn-based Arena players who have multiple games finishing at similar times.

Manipulation is not possible. In turn-based games you have to play each game in turn.

Additionally, Elo and AES fail for multiplayer games because they assume the game is multiple independent 1v1 matchups, and assign a rating based on that assumption. Most multiplayer games on BGA just don't work that way! Most winning strategies for a N-player game don't rely on N-1 independent victories, and many times a worse final position can be attributed to causes outside the player's control - but the rating system will just assume you won or lost more 1v1 match(es), which often tells a more extreme story than what actually happened**.

Most probably, but it's the same scenario again. You're high-lighting a single bad game, and it's the same for everyone. If you ultimately have a better strategy, then over multi games it will prevail.
If you want to come up with a better ELO calculation, or a better system, then let's hear it.
It sounds like it would be best to push everyone into the Elite league sooner, so there can be more games played.
I've said it before that the league system doesn't work at all like personally expected (sports leagues). It suggests that you will only play people in your own league, which in reality barely happens. And all the games climbing out of the lower leagues don't really help anything.

It follows from your logic (always reduce the luck if available) then that you are suggesting to play Agricola 2-player?

The over-riding luck factor is going to be the order in which you get the win results.
If you play 10 games and are going to win 7. Then it's best to win 7 in a row and then park that score. This is independent upon the luck you may get during any individual game (I had to think pretty hard about that).

The more I think about it, the more I agree with the sentiment of the original post. Using 'luck' and 'serious' in the same sentence sounds bad, and I think puts people off immediately. People really hate bad luck, and confuse their good luck for skill. It needs better marketing.

[dschingis27]

For anyone with more in-depth interest in the topic, I recommend this very good talk.
https://www.gdcvault.com/play/1024920/B ... -Day-White
The speaker distinguishes input randomness and output randomness which is similar (but not quite the same) to what Lumin was talking about.

[Vandaley]

I fully support rotations in Arena games. I am one of the top players in Thurn and Taxis and based on 7 Wonders got those who only wanted to do 2 player (for both the reason of this post as well as the real time playing aspect) to agree. So now we switch and I think the community as a whole is generally pleased.