57 ms·

Although this article doesn't make the point explicitly, this is really about how we can't discern right tails. The Olympics nicely illustrate this. In most cas

by gizmo 2y ago

Although this article doesn't make the point explicitly, this is really about how we can't discern right tails. The Olympics nicely illustrate this. In most cases we can only tell which of these world class athletes are best by having them perform side-by-side. But otherwise who can tell by eye if an athlete is top 100 or top 10 worldwide? Even good chess players can't tell by the moves alone what the Elo rating of more skilled players are. We don't know how bad things are is one way to put it. Another way is that there are people of extraordinary ability, walking among us mere mortals.

joe_the_user 2y ago
Even good chess players can't tell by the moves alone what the Elo rating of more skilled players are. That's an interesting claim, is there any verification for it? (I'm honestly curious).
- wavemode 2y ago
  Humans can't analyze chess exhaustively. We do so largely by pattern recognition. As a result, we mostly cannot accurately analyze the chess of players who are significantly stronger than we are. In other words, when a human says "this is a good position for white" or "that was a good move" (roughly equivalent to saying "that move leads to a good position"), we do so not by checking every possible sequence of moves (which would require analyzing billions of possible continuations) and ensuring that all of them are good. We instead do so by intuition - we assess what the position looks like in comparison to positions we've analyzed before that we know lead to a win for one player or the other. As a concrete example: if we see connected passed pawns for white, we're probably going to say white is winning, unless mentally we can find a continuation where black is able to capture them or blockade them. So all it takes for a chess position to be good without a human being able to tell, is for there to exist some sequence of moves that leads to a different result from what we expected, and we are unable to find that sequence. In the example above, a human would wrongly conclude that white is winning if there actually does exist some deep and complicated series of moves black can play to permanently blockade the connected passers. You see this most commonly with chess engines (i.e. chess AI). The engine will make a move, or suggest a move, and the human (who may even be one the top players in the world) raises their eyebrow. Why, because the move visually seems to make no sense and/or serve no purpose. The engine is making that move based on some billion-position analysis it performed, that the human just isn't capable of. Only after the engine ultimately wins the game, does the human go back and start to get a sense of what the purpose of that strange-looking move was.
  - joe_the_user 2y ago
    Your argument relies too much on "I know exactly how people think". I'd like evidence.
    - wavemode 2y ago
      It's not an argument at all, it's just an explanation based on my experience playing and studying chess. My purpose was not really to argue or debate. If you're -curious- about why weaker players can't accurately analyze the chess of strong players, I've laid out for you why that's the case. On the other hand, if you're -skeptical- that weaker players cannot do this, I'd encourage you to do your own research on the topic and highlight where you believe I've stated something false.
goatlover 2y ago
> The Olympics nicely illustrate this. In most cases we can only tell which of these world class athletes are best by having them perform side-by-side. Maybe for the judging events. Race vents have times. You can google how fast someone has run or swam before.
- deltaknight 2y ago
  Over time this is likely true, but the reality of the top-end of sport is that tiny differences in tracks/environments have a big impact. So if your goal is to find the best out of a group of people at a point in time, I’d suggest the only fair way to do it is to pit them against each other at the same event. For an example of this, see how the Paris Olympic pool was considered a slow pool due to its depth. Or how wind impacts running times, and can’t always be fairly accounted for (e.g. gusts). There’s also the phenomenon where athletes perform better when competing against someone vs competing by themselves.
TheAceOfHearts 2y ago
I've played this game called Heroes of the Storm for many thousands of hours and I still struggle to identify a player's skill level. The only time I'm certain that I'm playing against a Grand Master-tier opponent is when the match-up allows for a lot of counter-play, these players manage to make you feel incredibly frustrated at every opportunity. It's also one of the best ways to learn, because it teaches you what kind of behavior is allowed and possible, and how to punish your opponent for their mistakes.
- empiricus 2y ago
  In games with elo ratings the skill level is quite clear. If I play vs a GrandMaster then I will lose 99% of the time. I can watch the replay and see what he does better, but even then I will probably understand just a part.
  - TheAceOfHearts 2y ago
    It's not as clear in a team based game, since draft and teamwork both play a large role in the match outcome. Many lower ranked players are often very skilled at micro but fail to macro correctly. There are streamers that do seasonal Bronze to GM runs, and sometimes viewers will accuse someone of being a smurf because of how well they're playing. But then after the match their profile gets checked and it turns out they're actually playing in their correct rank, and they just got a very lucky draft and overall game. Finally, even if you're playing vs a GM they might not be playing their main role / heroes, or going the most optimally tryhard builds. Even if someone has the core fundamentals correct, mastering a hero and learning all of their interactions perfectly could take dozens to hundreds of games. An example from another game: in Brood War there was a time where the most legendary player, Flash, held two of the top 10 spots in the ranked ladder; once for Terran (his main race) and once for Protoss (his second best race). But while his Zerg is also very good, he's not a top 10 in the world player of that race.

by gizmo 2y ago

joe_the_user 2y ago

Even good chess players can't tell by the moves alone what the Elo rating of more skilled players are. That's an interesting claim, is there any verification for it? (I'm honestly curious).

wavemode 2y ago
Humans can't analyze chess exhaustively. We do so largely by pattern recognition. As a result, we mostly cannot accurately analyze the chess of players who are significantly stronger than we are. In other words, when a human says "this is a good position for white" or "that was a good move" (roughly equivalent to saying "that move leads to a good position"), we do so not by checking every possible sequence of moves (which would require analyzing billions of possible continuations) and ensuring that all of them are good. We instead do so by intuition - we assess what the position looks like in comparison to positions we've analyzed before that we know lead to a win for one player or the other. As a concrete example: if we see connected passed pawns for white, we're probably going to say white is winning, unless mentally we can find a continuation where black is able to capture them or blockade them. So all it takes for a chess position to be good without a human being able to tell, is for there to exist some sequence of moves that leads to a different result from what we expected, and we are unable to find that sequence. In the example above, a human would wrongly conclude that white is winning if there actually does exist some deep and complicated series of moves black can play to permanently blockade the connected passers. You see this most commonly with chess engines (i.e. chess AI). The engine will make a move, or suggest a move, and the human (who may even be one the top players in the world) raises their eyebrow. Why, because the move visually seems to make no sense and/or serve no purpose. The engine is making that move based on some billion-position analysis it performed, that the human just isn't capable of. Only after the engine ultimately wins the game, does the human go back and start to get a sense of what the purpose of that strange-looking move was.
- joe_the_user 2y ago
  Your argument relies too much on "I know exactly how people think". I'd like evidence.
  - wavemode 2y ago
    It's not an argument at all, it's just an explanation based on my experience playing and studying chess. My purpose was not really to argue or debate. If you're -curious- about why weaker players can't accurately analyze the chess of strong players, I've laid out for you why that's the case. On the other hand, if you're -skeptical- that weaker players cannot do this, I'd encourage you to do your own research on the topic and highlight where you believe I've stated something false.

goatlover 2y ago

> The Olympics nicely illustrate this. In most cases we can only tell which of these world class athletes are best by having them perform side-by-side. Maybe for the judging events. Race vents have times. You can google how fast someone has run or swam before.

deltaknight 2y ago
Over time this is likely true, but the reality of the top-end of sport is that tiny differences in tracks/environments have a big impact. So if your goal is to find the best out of a group of people at a point in time, I’d suggest the only fair way to do it is to pit them against each other at the same event. For an example of this, see how the Paris Olympic pool was considered a slow pool due to its depth. Or how wind impacts running times, and can’t always be fairly accounted for (e.g. gusts). There’s also the phenomenon where athletes perform better when competing against someone vs competing by themselves.

TheAceOfHearts 2y ago

I've played this game called Heroes of the Storm for many thousands of hours and I still struggle to identify a player's skill level. The only time I'm certain that I'm playing against a Grand Master-tier opponent is when the match-up allows for a lot of counter-play, these players manage to make you feel incredibly frustrated at every opportunity. It's also one of the best ways to learn, because it teaches you what kind of behavior is allowed and possible, and how to punish your opponent for their mistakes.

empiricus 2y ago
In games with elo ratings the skill level is quite clear. If I play vs a GrandMaster then I will lose 99% of the time. I can watch the replay and see what he does better, but even then I will probably understand just a part.
- TheAceOfHearts 2y ago
  It's not as clear in a team based game, since draft and teamwork both play a large role in the match outcome. Many lower ranked players are often very skilled at micro but fail to macro correctly. There are streamers that do seasonal Bronze to GM runs, and sometimes viewers will accuse someone of being a smurf because of how well they're playing. But then after the match their profile gets checked and it turns out they're actually playing in their correct rank, and they just got a very lucky draft and overall game. Finally, even if you're playing vs a GM they might not be playing their main role / heroes, or going the most optimally tryhard builds. Even if someone has the core fundamentals correct, mastering a hero and learning all of their interactions perfectly could take dozens to hundreds of games. An example from another game: in Brood War there was a time where the most legendary player, Flash, held two of the top 10 spots in the ranked ladder; once for Terran (his main race) and once for Protoss (his second best race). But while his Zerg is also very good, he's not a top 10 in the world player of that race.