Way to make ELO matchups more similar?

[FrankJones]

Now, with regard to chess. Well, that's an interesting question. Does the Elo model work in chess? The math would say I have a non-zero chance of beating Magnus Carlsen.

But I think I have zero chance. That said, maybe there is some luck in chess. Maybe he goes to make a move and some misfortune, bad luck, spasm, whatever, causes him to touch a pawn that is between my bishop and his queen, and now he has to move the pawn, and he loses the queen without gaining anything.

I'm not sure what skill level someone would need to be to win against Carlsen if he is playing minus a queen, but the win probability would certainly go up a lot. I would try aggressively force trades of equal material before I become ensnared in some sort of clever trap that costs me my queen.

[Meeplelowda]

Maybe I'm changing my mind somewhat on some firm stances I expressed in that other thread, though I'm not sure.

This is intellectual maturity. Taking in opposing arguments and considering whether you maybe should reassess your position.

I had a similar evolution on conceding where I used to argue that everything should be viewed the same as abstract combinatorial games where conceding is routine and expected, and I no longer hold that view.

[Meeplelowda]

I'm not sure what skill level someone would need to be to win against Carlsen if he is playing minus a queen, but the win probability would certainly go up a lot.

Even a large multiple of an infinitesimal amount is still negligible.

[tbhp]

The luck factor comes in what 200 and 700 mean. In a high-luck game, 700 is probably the most you can ever hope to reach, and even then it would be overestimated. In a low-luck game, 700 is probably just high-average.

It's more accurate to define it as the amount of control a player has on the outcome, rather than luck per see.

For example, Tic Tac Toe has zero luck, but its Elo would reach much lower than a game like Poker, which has a lot of luck.

What really matters is how much variability there is for skill. Poker allows a ton of possibilities for better players to dominate a game, whereas Tic Tac Toe is a very simple game where everyone is pretty much on the same footing.

I'm not sure what skill level someone would need to be to win against Carlsen if he is playing minus a queen, but the win probability would certainly go up a lot. I would try aggressively force trades of equal material before I become ensnared in some sort of clever trap that costs me my queen.

Magnus Carlsen won the EWC 2026 final despite losing his queen in the opening. Granted, it wasn't a random occurrence, he did it on purpose because his opponent tried something that he saw he could exploit (by sacrificing his queen no less), but it's still pretty impressive.

[tbhp]

So, perhaps this goes back to what ChiefPointThief (and others) were saying in a 20-page thread awhile back: The Elo model is imperfect when games have varying amount of luck.

In another game, I believe myself to be 700 level, though I have not reached it yet, haven't played enough games), but in that game, I believe three are enough variance factors in the game mechanics that I could lose to a 200 Elo player.
(It's also possible that game has too much potential variance and luck within a single game for anyone to maintain a 700 Elo, which also ties in to that previous long thread.

Maybe I'm changing my mind somewhat on some firm stances I expressed in that other thread, though I'm not sure.

That thread was talking about another imperfection, being that the more luck based a game is, the harder it is for Elo to consistently settle around your "true level".

What I'm saying is that it may also be imperfect in that even at similar Elo ceilings and similar luck factors, the win probabilities across different games may not be similar, and so the win probabilities it states should be taken with a grain of salt.

The fact that I can see myself losing to a 200 Elo player in Harmonies is not related to the luck factor. It's related to what they have at their disposal. They possess all the same information I do, so I don't see why if they put the necessary amount of thought into it, they wouldn't be able to beat me.

In some other games things may be different, because experience plays a bigger role, so a beginner can't just logic their way into a win. They need some experience of what works and what doesn't work.

Basically, when we apply the Elo system to a game, it's like we are trying to make it fit neatly into a box, when in reality every game has its own quirks. In some games beginners will have better chances than in some other games, and that is not properly reflected in Elo when it says that the chance is always 5% or whatever.

[RazorOz]

I wouldn't describe anything you can reach 700+ in as high luck tbh, at true high luck games like Can't Stop and the like you will struggle to break 500 ELO, let and out 700 ELO. 700 is probs more middle of the pile, like most mid-weight Euros you probably couldn't reach much higher than the 700s. Anything that's like 900+ tends to be very low luck with a high skill ceiling.

As for the chances of beating someone, you also have to bear in mind how much ELO fluctuates. ELO is probably most accurate on very low luck, high skill ceiling games. Otherwise what will happen is I could be ranked 800 ELO, go on a bad run, and now be ranked 600 ELO, all of a sudden it will say my chances of beating the 200 ELO player are much lower, but in reality they've probably not changed, just when I was ranked 800 ELO my ELO was inflated and now that I am ranked 600 ELO I am ranked lower than my true ability. My true ability likely always lied somewhere in the middle.

The variance of a game probably is also relevant, in the sense of I imagine there's some games that cap out in the 700s which have some variance, but not tons, and the ability to get completely screwed is low, just the skill ceiling isn't high enough for players to get their ELO much higher. There are other games though where relatively the skill ceiling can be high, but the chance to get completely screwed is higher than the game in the previous bracket. This is most obvious in something like Magic: The Gathering, I think they have moved away from ELO or glicko as they used these days, but the equivalent of like 700+ ELO player playing a 200 ELO player at Magic would feel like you are 0% to lose, but Magic is also a game where in some games you can get so screwed that you effectively can't play the game and lose to anyone. Though Magic is a weird one on this scale regardless, in the sense of in a lot of formats match up matters, you could be a better player than me, but I could be highly favoured to win, because my deck has an 80% matchup vs your deck.