Can we revisit Red Thumb in Arena? Some ideas for improvement

[tbhp]

Back to what was discussed. Is it a common occurrence that someone can sustain a 50%-win rate against the 700 Elo players (or the top 10 Elo, if it's not a game with a 700 Elo top 10), but cannot sustain the necessary win rate against the players ranked 11th -25th, or the players 100 Elo lower than the top 10? That sounds odd.

It is not just common, it is expected. Everyone is different, has their own way to play, their own understanding of the game, and even things like their own motivation levels and confidence levels when playing against a variety of different players.

It would be extremely improbable for two top rated players to have exactly the same distribution of winrates across the whole range of possible opponents. Just because two players have an even 50% winrate when they play against each other, doesn't mean they are doing equally well when they play some other type of opponent.

It can be seen when observing players' success in tournaments. There are players whose Elo may be like in the 500s, where when you put them in a round robin tournament with only 600s Elo players, they will consistently do well in those tournaments. Yet their general Elo always seems to remain a bit lower than the Elos of the top players they regularly beat. Why is that? Maybe it's because they are less motivated when they play against lower rated players. Maybe it's because their brain is better adapted to take on a great player than to take on an average player. It's a myriad of things.

For Arena, the difference seems even more glaring, because like I said above, a top player who can win 95% of the time against random players is going to have a huge advantage over one who wins 90% of the time. Like, at least a 100 points difference in the Arena score. When those two players play each other, they may be around the same level and win equally as often, but in a context where all levels of players are present, one has a stronger ability of securing those wins against average players. And those make all the difference, as we all know how important it is not to lose those games if you want to hold your score in Arena.

[RazorOz]

Also because player pools tend to split via various modes to play a game, ELO isn't always accurate of reality.

For example in my experience the toughest way to play is real-time Arena, it's tougher than turn-based, and both are significantly tougher than non-Arena games. If you play say a 400 ELO player in Arena they tend to be stronger than a 400 ELO player outside Arena, even though logically that shouldn't be true. I suspect it's because in practice the crossover between groups is relatively low. It's also likely to do with you can't lose ELO below 100, and there are far more 100 ELO players in non-Arena games vs Arena.

There can be other principles too, so take Heat, you can play Championship mode which is effectively 4 races in 1, these will lead to way lower variance in your ELO, because you are far less likely to lose to someone ranked significantly lower than you over 4 races than 1. It's where ELO wouldn't really be telling, because say you are 80% to beat someone, the system will say you're still 80% to beat them in Championship mode, in reality you will be much, much higher. So in theory Championship mode would be far easier to gain ELO, in reality most the top players are not playing Championship mode so you will run into a lot of lower ELO opponents, and games obviously take 4x as long to play.

[Ceaseless]

Also because player pools tend to split via various modes to play a game, ELO isn't always accurate of reality.

For example in my experience the toughest way to play is real-time Arena, it's tougher than turn-based, and both are significantly tougher than non-Arena games. If you play say a 400 ELO player in Arena they tend to be stronger than a 400 ELO player outside Arena, even though logically that shouldn't be true. I suspect it's because in practice the crossover between groups is relatively low. It's also likely to do with you can't lose ELO below 100, and there are far more 100 ELO players in non-Arena games vs Arena.

There can be other principles too, so take Heat, you can play Championship mode which is effectively 4 races in 1, these will lead to way lower variance in your ELO, because you are far less likely to lose to someone ranked significantly lower than you over 4 races than 1. It's where ELO wouldn't really be telling, because say you are 80% to beat someone, the system will say you're still 80% to beat them in Championship mode, in reality you will be much, much higher. So in theory Championship mode would be far easier to gain ELO, in reality most the top players are not playing Championship mode so you will run into a lot of lower ELO opponents, and games obviously take 4x as long to play.

I have not found players that also play in Arena to be any more difficult than players that don't. While generally filtered Arena games are less competitive than filtered regular games, the main reason for that is the filter difference, so you can avoid the lower rated opponents easier in regular mode than Arena. In other words, Arena can't always avoid facing 100s-200s.

I'm actually surprised by your championship mode comments, given the advantages you claim of championship mode I would expect it to be favored by the higher rated players.

[tbhp]

Also because player pools tend to split via various modes to play a game, ELO isn't always accurate of reality.

Pretty much. This is why I've always found it weird that Arena can be played either real time or turn based. Comparing the score of someone who plays all their games real time and the score of someone who plays all their games turn based doesn't seem like a very pertinent comparison. Either it should be two different rankings, or it should be the same setting for everyone, decided by gurus every season.

In contrast, tournaments are either all real time or all turn based, and it would be weird if players could pick on a game by game basis.

People often act like Elo is a fixed thing, but in reality it's impacted both by its natural statistical variation (people tend to be aware of that) and also by the types of variants and the types of players that someone plays (people tend to not be aware or not acknowledge that). In other words, it's also your "gaming habits" that determine where your Elo will hover around.

[RazorOz]

I have not found players that also play in Arena to be any more difficult than players that don't. While generally filtered Arena games are less competitive than filtered regular games, the main reason for that is the filter difference, so you can avoid the lower rated opponents easier in regular mode than Arena. In other words, Arena can't always avoid facing 100s-200s.

I'm actually surprised by your championship mode comments, given the advantages you claim of championship mode I would expect it to be favored by the higher rated players.

I guess it also maybe game dependent, at Heat, it's for sure the case. And I don't even really use filters (because imo they're hard to use when you want to play 4-5 player games as you just never find a table that way). My ELO nearly always would start going up when I play regular games for a bit, and while this is partially to do with the average player in Arena being stronger, I will also win at a higher rate outside Arena, and that's despite outside Arena I will play 5 player, whereas Arena is 4 player, and yet I'd still win like >10% more outside Arena.

Championship gets shunned for a few reasons, #1 is time, for a real-time player, you need so much time to play even 1 Championship game, because you are looking at probably 3 and a half hours to play one game. But even at turn-based that is having a significant impact, if your game that was before 10 days long, is now 40 days long. #2 It does introduce random luck elements that I don't think are popular with high rated players, firstly you can be at a significant advantage purely from where you started in the first race, because you only draft 1 upgrade for the first race, whoever gets 1st pick can have a massive advantage over those with later picks, it might not matter too much if there's a huge ELO gap, but if you are say rated 700 ELO, and playing someone 600 ELO it will, and it's particularly bad when you combine with the time it takes to play, eg. no-one wants to play someone of similar ELO to them for 3 and a half hours in a game they effectively lost or were put at a significant disadvantage in on turn 0 from luck of the draw (order of the grid). The sponsors I also think add a variable that aren't great competitively, part of it is in a regular game of Heat you know all the variables, you know what cards your opponents have to play around etc., once sponsors are in the mix you don't know what they have, and while the sponsors themselves all have uses, at certain times some are clearly better to draw than others.

I think what Championship does, and this ties in to FrankJones point where he said he'll play like a 200 ELO player and feel like he's 100% to win even though the %'s say otherwise, so at Heat you might feel that way, but in reality in a normal game you will always at some point just get so screwed over that you lose, which is why you end up like 95% or whatever. But if you played those players in Championship mode where in a normal game you are say 90% to win, in Championship mode it's going to be up close to virtually 100%, because in a one-off race anyone can get completely screwed, but over 4 races? Highly unlikely.

[RazorOz]

People often act like Elo is a fixed thing, but in reality it's impacted both by its natural statistical variation (people tend to be aware of that) and also by the types of variants and the types of players that someone plays (people tend to not be aware or not acknowledge that). In other words, it's also your "gaming habits" that determine where your Elo will hover around.

This is a key point, that the majority of games have some kind of variance. Like at Heat in the last year, I believe my peak was 824 ELO, my low was in the 560s, the win % calculator will say my chances of winning when I was ranked in the 560s were significantly lower than when in the 800s, but in reality my true ability probably hasn't changed that much in that time, that's just the variance of the game, you're never going to get perfectly accurate predictions for that reason.

A game like Heat even adds further random variables that ELO can never truly judge, and I would imagine these exist at other games in some form. The main one being what track gets selected. The lowest win rate track I win around 25% less than half the tracks, because that track has by far the most variance. It's another thing that ELO can never truly judge, other than working off the long run average, it can't establish that in this particular game you are actually much more likely to win because it's track A, than this other game because that's track B.