How much does fast thinking matter in Lichess games?
emiruz.com
emiruz.com
It’s almost a different game. Yes, thinking “fast” is good, but it’s not about finding an ideal move, it’s about finding any move that isn’t an obvious blunder as quickly as possible.
There’s also different strategies that emerge with the time limit. Running out the clock of course. When time gets low just move quickly. Also, confusing moves. Anything that makes your opponent go “Wait, what?” can give you an advantage in time.
I’d argue that faster speed relies more on pattern recognition and intuition, versus strategy.
- Detect when your opponent is using premoves and use that against them (for example, if you think an opponent probably has a premove for their next turn, you can plop an unprotected bishop right next to their queen. If you guessed wrong, you lose the bishop. If you guessed right, you take their queen.)
- In an endgame with seconds left, you can make “confusing” moves that in any other game would be blunders: Say a rook-pawn endgame, if I notice my opponent has a second or two left I’ll plop my rook right next to their king. Check, the premove is invalidated, they scramble to move their king and run out of time.
So much meta-gaming potential in Bullet.
It’s also the variant where internet latency starts to matter a lot. When you’re making a move per second, or better, the difference between 100ms round trip ping and 30ms starts to add up.
Edit: according to this page lichess does compensate network lag to make play fair for players on slower connections, though it doesn't specify how. https://lichess.org/lag
Blitz and bullet chess are much more fun because precise play is impractical and conservative strategies don't work as well. You can drag a wild, tactical game out of the most timid opponent.
For the vast majority of players, this simply isn't true. Even for the top players in the world, it's only sometimes true when they are playing with the black pieces, or a win doesn't do anything for their standing in the tournament.
Aggressive play puts a chess player at a marked disadvantage. So people don't do that. They play conservatively, because that gets them the best results.
The intent may ostensibly be to win, but the observable behaviors are driven by loss aversion.
Again, the vast majority of players are not strong enough to even know any lines, let alone refute the endless variations that may be thrown at them. Unstable positions typically end in a blunder, or a simplification down to an end game. No matter how risk averse one of the players is, the other player usually has the power to change the nature of the game pretty easily. And at non-classical time controls (rapid, blitz, bullet, etc...) the game becomes extremely interesting at all levels of play.
> By “fast thinking” I mean any thinking process which is advantageous in chess and which can be employed in short time intervals.
> I’d venture therefore that fast thinking is more prominent in quicker games.
Isn't this pretty much true by definition? Fast thinking is the only thing you can do in quick games, so it has to be more prominent.
But then the author continues:
> That is, if fast thinking is the dominant factor determining outcomes in games amongst normal players, it should imply that more time makes the faster thinkers advantage more prominent
And now I'm just confused. How do we know that fast thinkers aren't using "slow thinking" or anything else when they play at slower time controls? Maybe good fast thinking is correlated with good slow thinking, but this doesn't mean it is the dominant factor!
PS: I've never seen a player with a high bullet rating but a low blitz/rapid rating, so I'm not sure how to interpret a negative result either.
I believe that the author is just restating the NH, which is the hypothesis that slowing down time controls will make the dominant player more dominant (or at least maintain the advantage). The alternate hypothesis is stated afterwards
> AH: NHC is false. On average, given some rating difference \(x\), if \(A\) beats \(B\) at the quick time control with a probability \(p\), then \(A\) beats \(B\) at the slower time control with probability \(p' < p\).
Right at the beginning of the article the author says the main goal is to:
> investigate the extent to which fast thinking is the dominant factor affecting game outcomes at any time control.
To achieve this goal, I think we'd need an hypothesis that isolates the effects of fast thinking, and not one that could confound correlated factors with the effects of fast thinking itself.
By just comparing probabilities of winning at various time controls, we're evaluating players as a whole and not isolating any aspect of their thinking, so this hypothesis doesn't seem helpful.
Perhaps if the study looked at individual moves and how much time they took to make that could help?
If you think of chess as two people with different capacity CPUs calculating against each other (many people do) then time should only ever make the outcome more certain. The author basically says that if that was true, people that beat you at fast chess should be even more likely to beat you slower chess, but that doesn’t seem to be the case.
I don't have a "high rating" in bullet, but it's higher than my blitz/rapid ratings.
Why ? I'm impatient. When I wait for my opponent to play their move, I lose focus. When they take 30s to 1min to think about a move that seems obvious to me, it just makes me lose my train of thought.
Because of this, I play mostly only bullet games and puzzle rush. I'm at ~2400 in tactics on lichess, and ~1900 in bullet.
I think you might be implicitly defining "quick games" as ones where you only have time to think fast, so it comes across as a tautology to you. Meanwhile, the author writes in terms of continuums quicker and slower. I think the emphasis is on time pressure crowding out slow stuff that might make a difference. I.e. if the slower stuff made a difference, people that beat you at quick games might not beat you at slow ones ceteris paribus.
Today I'm just over 2000 on Lichess rapid. My inability to think quickly leads me to play like either an old or a drunk man (in fact I'm usually neither). I eschew tactical battles in favor of taking the position to a drawn endgame, where I can slowly gain an incremental advantage. Closed, maneuvering positions are great for me. My strategic ability is the only thing that saves me. Fortunately, tactics tend to favor the player with better coordinated pieces.
After some time you will notice that you don't have to calculate anymore for many simple tactics, you just spot them immediately after a few second, because you recognize the pattern. Now, you might still be a slow calculator, but pattern recognition is fast and you don't have to fear tactical battles as much anymore, because you can trust your brain, that'll find the simple tactics with a very high success rate.
Of course, there are also deeper tactics, e.g. a not so obvious mate in 7, but chances are high that neither you nor your opponent are able to spot them and you will only know afterwards with engine analysis. That's okay and also happens quite frequently even on GM level.
It tickles me to think that you're sometimes randomly old.
And if you are 2000 in rapid your inability to do well in bullet likely has little correlation with your raw thinking speed and very likely just pattern recognition and/or lowering your standards and/or mouse speed
If this were true statistically, I think that should be reflected in the Elo spreads: Players winning more in blitz should mean they’re able to separate farther from the middle of the rating distribution.
This doesn’t seem to happen at the very top end. Looking at the Elo range of ranks 1 to 100 to get a feeling for the distribution:
Blitz: 2850 to 2610
Rapid: 2847 to 2620
Standard: 2864 to 2645
(Source: https://ratings.fide.com/top_lists.phtml)
Looking at the difference between the top-ranked player and the 100th, we get 240 points for blitz, 227 for rapid, 219 for standard. That suggests that among the elite, the best players are indeed more dominant at blitz than at longer time controls. However the effect size is small and you'd want to check for other effects causing this: eg pretty much all of the best players will be on the standard list, but some of them may not be on the blitz list as that is restricted to players who have played a rated blitz game over the board in the last two years.
What I discovered about chess is that it is largely pattern recognition that makes people good at chess, which is visual thinking.
The other thing about fast games is the type of game you play, if you want an early result you generally go for early sacrifices to get positional advantage and queen side castling to open up the game.
Come to think of it, fast chess is probably a different game all together
But then people end up speculating on what an opponent will pre-move. Maybe you do something unexpected, so their premove doesn't work and they have to do another move and lose some. Maybe you can assume they have pre-moved something, so you can do something that's objectively a bad move, but get away with it since they can't punish you for it, having already decided on their move etc.
Quite a different game, but fun to watch.
I should play bullet a bit more often, really is fun.
I don't know why, but when I realized this it was a bit of a turn off for me. I still play chess, but the idea of climbing ranks became less appealing because it's simply a matter of volume of games:
You play more games, you recognize more patterns, you climb.
So now I'm content with my rank, because it's the rank I'm supposed to be with the volume of games I play.
That's not true. A chess master calculates lines to a much deeper depth than an amateur player. They also have much better heuristics for evaluating positions, true. And they are much better at pruning irrelevant lines, as you say. But at the end of the day a master-level player is evaluating many more positions than an amateur before moving.
"From my own experience I can say that grandmasters do not do an inordinate amount of calculating. Tests (notably de Groot’s experiments) supports me in this claim. If anything, grandmasters often consider fewer alternatives; they tend not to look at as many possible moves as weaker players do. And so, perversely, chess skill often seems to reflect the ability to avoid calculations. It is, in truth, not clear that chess is a game of calculation" – Artur Yusupov
I am not necessarily talking about playing fast. For example, it is very rare for me to see a Fried Liver attack in 5/3 --see it all the time in 10/5. When that happens, people playing 5/3 rarely fall for a Traxler counterattack, while, again, in 10/5 they fall for it all the time and get absolutely destroyed. The distinction is such that I have completely stopped using Traxler in 5/3 games.
My rationalization is that people playing 5 minute games must play at least 3x more games and have seen more variations, which, therefore, means they've seen enough to become better players.
I don't usually play games shorter than 5 minutes. I usually play in bed just before going to sleep, so I am mentally tired. Still, I find it relaxing enough to play a few games before ending my day. Shorter games are not a good fit for a tired mind.
The other thing I sometimes do is use chess to tell me what to focus on during the day. I might play a game or two while having a cup of coffee. If I do well, it generally tells me my head is in good shape for deep focus work. If I suck, I tend to avoid work that requires deep concentration and might instead go work in the shop, do some machining/prototyping/hardware testing, etc. Didn't bother to search around and see if anyone has done any research on this phenomenon.
I realize this is a single data point. I definitely have days when I could not win a game to save my life. Not sure why. I also have days when I will win ten out of ten games. Something is different in my head in each of those instances. I think using this information to decide how to go about the rest of your day might be useful.
They do talk of it being a proxy and confounds etc. but it just seems like there isn't much you can expect out of that test even beforehand.
The significance of the interaction terms in the analysis means that the Blitz rating difference does a significantly worse job at predicting who will win at Rapid.
Whether its on lichess.com or some other platform like chess.com, the guarantees no longer exist for there to be any value playing the game.
Specifically there is no longer a guarantee you are actually playing against another human, and cheating is largely not punished.
The vast majority of the time, and this depends entirely on whether its a weekend or not because its typically worse then (90-100% of the games), You won't be playing a human, you will be matched up against a bot, which may or may not have a human moving the pieces.
The platforms of course say they combat cheating, but in reality their efforts are ineffective, and that communication is entirely disingenuous if not intentionally deceitful or fabricated.
As of last weekend on average, a solid 70-80% of the sampled games were people using an engine to play for them.
Background, I'm rated around 1900-2000, have studied for decades, and my opening repertoire has been designed so there is a high deviation between engine supplied moves and known theory and strategy.
In most cases by move 17 there have been enough cumulative and unique moves to have shown whether they are using an engine. Specifically, there is no strategic plan for them but they choose engine supplied moves at various depths.
There are a number of branches where if it appears they are using an engine I may induce certain lines that are not obvious but may ultimately result in a loss to evaluate their responses.
Typically heavily technical and complex positions with no strategic theme for their side but multiple possible trees and tempo to account for with a winning advantage of 0.8-1 with perfect conversion, and when they ultimately convert perfectly (usually in seconds) around move 30 its pretty damn obvious to anyone what happened.
When there is no credibility between people that they will follow the rules. There is no point in playing the game.
You can only improve when playing against another person, if you are playing against a computer, there's no point, it is just a giant lookup table that's deep enough and accurate enough to memorize all possible moves, even down lines which haven't been documented. Winning is not making mistakes, and minor mistakes are what computers with perfect iterative evaluation capitalize on.
In these current circumstances there is no point, all a game of that nature provides is the person cheating is stealing your time, psychologically abusing you (and generally being deceitful), often because they themselves have a significant mental illness.
The ease of new anonymous accounts makes the pool of potential victims...games highly likely that if you follow the rules you are volunteering to be a victim for the sake of their illness. The only solution to such an environment is not to play.
The failure of the platforms and also the two main chess organizations to address this, only make this game defunct whether they want to admit it or not publicly (due to vested interests).
There is value in teaching certain problem solving patterns to developing minds, and as soon as that's done, chess can provide no further value.
Ultimately, the Computer scientists that developed and continue to develop chess engines have killed the game by making it meaningless and their actions have damaged the livelihood of those that choose that as a profession.
Why would anyone invest the time and resources today in becoming a GM when you can play better than a GM with an engine at your side with no prior experience.
Why would anyone buy products for chess that teach you how to play better to win (the right way) when an engine can win for do it for you and the majority of your opponents are just engines.
Lichess and Chess.com both like to say they do a lot to remove cheaters, and that's only a small part of the problem.
It comes down to time to detection, and time to re-entry.
If the first is large, and the second is small, water will fill the boat faster than you can bail it out and this has been a problem for decades.
AI won't magically solve the problem either, lichess already knows their siamese network can be thwarted with a minor tweak in inputs that are necessarily generated on the client. They chose to censor the discussion about it instead of correction, like that kind of discussion can't take place off their service by malicious actors, its just being an ostrich and burying the head in the sand.
Every security person knows not to trusts the client.