Targeting high ELO players

[buv3x]

Hi, here is a point for a discussion, which I post because I am myself on fence here and can see good arguments for both point of views and would be
genuinely interested in what the community thinks.
In short the question is: "Is fair/ethical to base you game decisions on your opponents ELO (e.g. targeting higher ELO players) in multi-player games, even if game state itself does not necessarily call for those (e.g. in the very beginning of the game)?"

Some arguments from my side, but first I think it's important to define fairness itself regarding multi-player games. In my mind, the best definition of fair play is: "Every player is making decisions that in his understanding lead to maximizing his win probability (or maximizing his expected position, the difference between those might be significant, but that's the point of another discussion). So, based on that, specifically targeting some other player is not good in itself, but it obviously can be good if the game situation dictates it (e.g. targeting the leader).

Now, back to the initial point. One way of thinking about is, say, you have a 3-player game, therefore at the beginning all have 33% win chances. If you pick one player and make a move that targets that player and shifts the probabilities to 20% for that player 30% for you and 50% for the third, that might not be fair if you also have a move that just make it, for example, 40% win chance for you. You are ruining other player's game for the sake of ruining it.
But the thing is, ELO rating is an objective way of predicting win probabilities for players, it's actually its main function. So maybe a more practical way of looking on the same situation would be to take ELO into account. Let's say, when ELO is taken into account, the player you target has 70% win chance from the start and you and the third player have 15% (percentages are exaggerated to illustrate the point). Then it might turn out, that you "targeting move" might become a better and more fair way of playing.

For me, the second way of thinking feels a bit wrong and it also kind of negates the ELO idea itself. On the other hands, when thinking rationally and not emotionally, I lean to the second point of view more. Maybe it's that ELO in multi-player games does not fit as well as it does for two, maybe there are other arguments I'm missing. What do you think?

[voriki]

See it differently. If you are a hiring manager, and you want to select the strongest candidate(in this case opponent), and you have an interview lined up with 2 people. Normally they have a 50% chance each of getting hired(attacked).

Now Player 1 has a better resumé, a work history that shows he is a stronger candidate, and ofcourse it impresses you more. But you still have the interview with both candidates.

Now if neither candidate has said anything during the interview to make them stand out more, obviously you are still leaning toward hiring(attacking) player 1, since he has a stronger work history and is deemed a good candidate(strong opponent).
But if on the other hand Player 2 started taking the lead in the game, gaining more points, your attention clearly shifts towards him. If he impresses you in the interview, the work history becomes less relevant.


Bottom line is, I would too target highest ELO until the current game tells me to target differently.

[pjt33]

It is most definitely that case that Elo is designed for two-player games and doesn't work in multi-player games. The ways in which it breaks down vary between games. E.g. Just One Express isn't too far removed from the model of everyone plays everyone else individually, because there is essentially no player interaction. Cubirds, on the other hand, is set up to have one winner and consider everyone else as tying, and there the pairwise comparison breaks down almost completely; if Elo were updated solely between the winner and each other player then it would more-or-less work. There are some games where you play in teams, and there the game treats team-members as playing against each other and tying, which is completely insane.

On the main question, if the aim is to maximise your own Elo improvement then low-Elo players targetting high-Elo players is certainly the rational play. But flipping that round, it's also the rational play for high-Elo players to target low-Elo players: losing to someone 300 Elo below you is a lot more painful than losing to someone within 10 Elo of you. It's a matter of opinion whether either or both of those are perverse incentives, and the weight of opinion might vary between games. There are also some games where it's much easier to target a single person. I can think of at least one where I try to avoid playing against players with much less Elo than me because I've played a few times where I felt ganged up on and it isn't fun.

[Shivaware]

But flipping that round, it's also the rational play for high-Elo players to target low-Elo players: losing to someone 300 Elo below you is a lot more painful than losing to someone within 10 Elo of you.

I hadn't thought of that before. That makes me think of that scene in Watchmen: "None of you seem to understand. I'm not locked in here with you. YOU'RE LOCKED IN HERE WITH ME!"

[Silene]

When you play with one player of 100 ELO and one player of 700 ELO and you become 2nd - it doesn't matter who is first and who is last. Net ELO gain/loss will be the same. So you should play for position anyway. But yeah - maybe targeting higher ELO players first might increase your chance of becoming 1st instead of 2nd. However going easy on the lower ELO player increases your chance of becoming 3rd instead of 2nd. So I don't think there's a really good answer of who to target would benifit you the most.

It is most definitely that case that Elo is designed for two-player games and doesn't work in multi-player games.

And also this!
So when applying this system to 3+, which it is not designed for, it leads to chaotic effects no matter what. I don't think there's a fix to this unless you're masking who your opponents are / disguising who is who (maybe this should be done in tournaments / arena). But even then the competition of course has more outer influences that are beyond your planning like an additional luck-component.