You are right about sbmm, but it doesn’t make for a good progression system imo. If my hidden elo says I’m diamond or w/e I should cruise through early ranks instead of being matched with other diamond elo players climbing from bronze. It should only feel like a slog/be super competitive when the external rank matches the hidden elo. Not the entire time IMO.

Some games get this right, others (looking at you Apex Legends Arenas Mode) does not and it just winds up being a terrible slog.

> If my hidden elo says I’m diamond or w/e I should cruise through early ranks instead of being matched with other diamond elo players climbing from bronze.

Or maybe you should stop making new accounts and stop smurfing?

I've been a beginner in plenty of games, and it sucks to be matched against someone far above your skill level in most games.

------

I get it. A lot of people want to be matched up against weaker players so that they have an easy win. That doesn't make it fair for the losers in this situation. Furthermore, it weakens the community and turns it more toxic. Expert players (or even advanced/intermediate players) shouldn't seek out weak players and stomp them.

50% win ratios, for everyone, should be the default goal. You climb the Elo to keep 50% win/loss, and as you lose, you fall to 50% win/loss as well, or so the ideal is. That way, everyone is on fair grounds.

If you want someone to "feed you wins", maybe play a PvE game like MMOs or something? But a competitive game is zero-sum by default. Half the players lose, and half the players win each game. So community managers need to spread the wins and losses around in a fair way.

> If you want someone to "feed you wins", maybe play a PvE game like MMOs or something?

Yeah this is easily solved - just go ratting in high sec in a nice specced out Drekavac or something.

> I've been a beginner in plenty of games, and it sucks to be matched against someone far above your skill level in most games.

I used to think this way until I started to understand what it takes to become competitive at high levels. The most efficient way to improve is to play with and against people better than you. If you take your ego out of the equation and pretend everyone is there to teach you, it's hard NOT to learn something.

Tangential to that, a lot of people would be honored to play against a GM in chess. Yet when placed against actual GMs doing some variation of a "3000 elo speed run", they throw a fit. What? Maybe its cultural. For a while in SC2 GSL, there was a side show where a random amateur for the audience got to play against a pro. You could clearly see the reverence most of the amateurs held for their opponent.

> The most efficient way to improve is to play with and against people better than you.

Unless the opponent is purposefully throwing a "teaching game" and/or giving an appropriate handicap of some kind, this is absolutely not the case.

In Chess, games at the 1500 Elo level are won/lost due to a knight fork, a bishop pin, a skewer, and other such traps. GMs literally will give you _zero_ opportunities to find such mistakes from them, because they play nearly perfectly.

Finding such mistakes (in yourself, and your opponent) requires an opponent who makes such mistakes at a roughly the same level as you. GMs who can see maybe 2 or 3 moves ahead of you will prevent such mistakes from even occurring, giving you no chance to learn.

A "show match" vs a stronger player is to be honored, but its not a lesson learned. Its not a teaching game.

------

Sure, GMs make mistakes. But the mistakes they make are invisible to 1500 Elo players. There's no hope to ever see it, and no one expecting to learn to play the game studies GM-level chess.

1500 Elo players will improve far faster with skewer puzzles, fork puzzles, outposts, etc. etc.

Why can't you learn from the forks and skewers that you expose yourself to, and that your GM level opponent will never miss?

Because that's not how the human brain works.

You learn skewers by practicing them a thousand times. Not by having a GM show you your mistake once over the next hour. Chess, at 1500 Elo, is decided by one mistake like that.

------

There's also the chance that the GM performs a checkmate in 7 vs you. This is too far ahead for a 1500 Elo player to see, but something easily within a GM's grasp in the midgame. In this case, you learn nothing (you shouldn't be looking for checkmate in 7 or that depth as a 1500 Elo player)

This is why playing vs Stockfish (superhuman AI) is useless. It doesn't teach you how to think like a human or how to get better at chess as a human... all of Stockfish's moves are so deep that they're AI only.

------

Its different if the GM is explicitly performing a "teaching game". (Ex: Hey, you can be pinned with this move, lets go back 2 moves so you can try to avoid it). But if the GM is there to troll you with shitty strategy because they're way stronger than you (ie: most online smurfs who are +200 or so Elo above you), then you learn nothing and simply serve as the opponent's entertainment for the night.

------

Same thing with fighting games: go into training mode, and perform your combos another hundred times. You learn nothing from the opponent comboing you though.

Same thing with RTS games: go into training mode, and practice your build orders another hundred times. You learn nothing by the stronger opponent executing their build order perfectly.

Same thing with League of Legends. You learn nothing when the stronger opponent team combos / ganks you out of position. You need to practice your ganks (which are likely a different character, so you'd need a different set of spells used at different timing).

Etc. etc. Training / getting better at a game is very different from actually _playing_ the actual game.

Or maybe you're new to ranked and hidden elo takes in to account quickplay? I've never smurfed in a game so shove it.

Actually, I think matching you against equal players would be much better in such a situation, but you should gain much more visible elo for wins than you lose for losses (thus going up even with a 50% win rate).

Lol so true. Why am I being matched up against top 500 players when I am quite literally making my way through bronze.

Just double my queue time until we find a good match at that point.

For you to win more than 50% of the time requires that others lose more than 50% of the time. Which sucks.

But what's more important, is that if you apply the same logic to other players from 1 league higher than your real ELO - you should be the one losing over 50% of the time to make them happier :)

The issue is that riot has designed it so the 50% has become the target rather than the metric.

The result is that you'll go on win and loss streaks and if you play long enough it becomes obvious it's being manipulated that way.

Winning 1 or 2 and losing 1 or 2 are just fine.

winning 10-15 followed by losing 10-15 just feels bad and isn't worth it. And if you've played long enough you can start feeling when the tipping point is about to happen. The wins will start getting more and more difficult as more people on your team obviously don't belong in the game. Or vice versa, and you'll start getting people on your teams consistently such that nothing you do matters because you'll just win.

Then you add in a dash of streamers and people doing "bronze to plat" challenges and it's just not fun. Either they're losing on purpose to tank their rank or they're pubstomping people far below their skill level.

The end result is that it all just feels bad and the only people that end up playing a significant number of games are actively addicted to it.

Yeah, I've absolutely noticed those long win/loss streaks and it reduces the value of the game by far. It's like, by the time I get a win, it's such a total stomping, it's not even fun. I slog through like 8 losses in a row, and then suddenly get a totally uneven matchup (in my favor) and the game is like the shortest possible game, because the matchmaking finally tossed me a bone (or whatever). Great, I finally won a game, and it was basically just a free win. I don't find it fun, and the losing team definitely doesn't either. The concept of "winning"/"losing" is diluted or made irrelevant by a crappy underlying system. After this long you'd think there would be a better system devised? I'd rather wait longer than have to get these constant poorly-balanced matchups.

Goodhart's Law!

There can still be issues. For example, how does the ELO system view previous wins/losses when calculating ELO gain/loss from future wins?

Anecdote: My friends played the first few ranked seasons on my Xbox Overwatch account, they started at Bronze and eventually got to Silver. I decided to pick it up and play ranked with a different friend, we won 9/10 of the placement matches. I got ranked Silver, he got ranked Platinum. A month later, I make a new Overwatch account on PC and rank Platinum with ease.

In these systems, if you gain a lot of skill in a short amount of time, or have bad luck early on in your ranked career, you can be held back in rank gains to the point that you are essentially stuck in place. Creating a new account generally allows you to advance significantly faster through the ranks.

And I actually only truly grasped this recently(the concept makes sense). My paranoid self always thinks that in a game like League, there are artificial matchmaking things going on to keep you at 50%, apart from a stagnation of skill. In truth, I probably just need to get better(at whatever, doesn't have to be League).

Yeah, if you win 50% of your games it means the matchmaker is working perfectly. People really have a lot of trouble understanding that.

Also, to improve, you probably have to lose a bunch of games while you're unlocking that new knowledge. The toxicity of teammates kind of steers people away from that. Thus people get stuck at where they place and never go up or down significantly.

I’m confused as to what folks think the alternative is.

Winning is a zero-sum game. If they target something higher than 50%, there would have to be a reservoir of players willing to take less than 50%…

It's not hard to come up with designs where you effectively cycle types of queues to where the end result is 50/50, but the pathway to get there is psychologically manipulative(win streaks/loss streaks/etc...).

Now, winning a single game of LoL is zero-sum, but you can't say that the ratings themselves are. Without knowing what the actual system is doing, we can't assume that players are losing/winning the same number of ranking points in a game.

Ranking should be more or less stable across even a small number of games unless your skill level actually changes.

Trying to draw parallels with singleplayer content where you can progress regardless of actually getting better is a lost cause.

Unless of course if a team full of me would have a 70% winrate, and the system is compensating (because hidden ELO) by sticking me with teammates that each have 45% winrates.

It still averages out to 50% winrate. Just because you can force a point of convergence does not mean it is the correct point of convergence.

Yeah, I definitely feel this variance in matchmaking. play games with my friends every Saturday night and 8 hours in after 5 beers, I doubt our MMR is quite as high as it was when we started the session. (The first game probably isn't great either!) But the matchmaker doesn't really know these subtleties, so sometimes the enemies get an easier than expected game. So it goes.

The competitive ladder climbing meta is really about recognizing when you shouldn't be playing and not playing then. Interesting angle to optimize.

Yeah, if I want to climb a ladder I have to play first thing in the morning (with physical exercise as well as warmup games), and make sure I quit at the right time.

I don't think the system should even care for these externalities because they will eventually balance out across a player base and the net effect should be zero(everyone will eventually play against a team with a drunk person on it). But it creates a prolonged cycle and general frustration amongst the player base.

Eventually is the key word. What eventually looks like to the player could be 10 matches, or 1000 matches, depending on how well the ELO system is implemented.

Achieving an equilibrium is step 1. Allowing for upward mobility is step 2. Enabling that upward mobility to occur (correctly) over fewer and fewer games is step 3.

Alternatively, you could also skip step 3, and profit off of smurfing.

Back when I was in game dev, I really wanted to find a way to make them fun even when you lost. I never managed to grok that magic, but I did see it in e.g. Settlers of Catan.

This becomes a different beast when you introduce competitive play. In non-ranked there is this feeling of freedom to try new things, have fun, etc... but in ranked it feels like things are on the line(despite those things being so trivial as a few points up or down).

Settlers is so fun because you really feel like you're on a journey the whole game, and while it is nice to win, you remember the interactions and experiences in the game as the valuable thing(IMO).

Unless the dice fuck you over extra hard x)

Play with a deck of cards instead of dice! 36 cards, six sevens, five eights and sixes and so on.

I actually did a similar thing using a python script! shuffle(all_combinations[: -10]), and repeat forever. Change the -10 to whatever you want: take everything and you've got your solution, put -35 and you've got random dice again. -10 is a nice middle term, also avoids that at the end of the deck results become too certain.

How is that supposed to help? A bad run is a bad run no matter the randomizer. Also cards need to be shuffled which wastes time if done correctly

A deck of cards behaves the way the gambler's fallacy expects everything to behave: after enough bad luck, you're sure to get a good result.

I've taken to calling that the gambler's fallacy fallacy: the misapprehension that a randomness can only arise in situations where the events are uncorrelated. The randomness in Tetris is another good example: when you think you're due for an I, you actually are.

Tetris actually uses two entirely different algorithms, one classic and one modern. Classic is randint(1,7), modern is shuffling "decks" of the 7 pieces and repeating.