No Skills Necessary: The Dunning-Kruger Effect in Chess
Hi!
Arguing with idiots is like playing chess with a pigeon. No matter how good you are, the bird is going to knock over the pieces, take a crap on the board, and strut around like it won anyway.
- Unknown origin
Most of us have at some point in our lives dealt with people who come across as unusually confident. They have an attractive vibe and they can amass a good flock of followers in the online world. But it's not harmless to mistake their appearance for reality. The loudness and boldness with which they portray an aura of authority may very well be an act of overcompensation. Behind the screaming mask you might find a complete lack of substance, and their superiority is predicated on your not being able to see behind the front they fed you.
Underneath a thick layer of quasi-knowledgeableness, you may very well encounter enormous depths of stupidity. The world of chess is anything but short of these examples. In this blog I'm going to show you several of these nut cases.
Introduction: the perfect response
Writing with artificial intelligence
Introduction: The perfect response
It was an early morning in February 2019, and the ride made my wish to be back in bed even stronger. On this day I was to play in a chess team match; we were the visiting team, the distance was significant, and I was to sit in the same car as my least favourite teammate. This guy, who believes that he should be handed the GM title on a silver platter even though he had stagnated and plateaued for several years, was up to his usual monologues about the incompetent people he's oh-so-cursed to have to work with. He was filling up the entire car with his snide remarks here and there, while I was debating whether to try to keep my eyes open.
As 50 Cent said in The 50th Law, "the kid in the schoolyard who doesn't want to fight always leaves with a black eye." And sure enough, halfway through the ride, when he ran out of other targets for his venom, he decided to single me out. "Hey, how come you had this rating surge lately? You don't even know how to play chess!"
I was fuming on the inside. Can't this dude just leave me alone for this one sec? I've so had it with this guy, with his unearned...can't I just...I want to...
...
...
Do you know those perfect answers that always come to you at 1:30 a.m., when you've already lost the argument and the moment to reply is long gone? How you wish for that extra time that you needed, like "if only I had come up with this answer there and then"? This day, for the first time in my life and the only time that I can recall, the right answer finally came to me at the right moment.
I sat up and asked my teammate:
Do you know what the Dunning-Kruger effect is?
As it happens, shortly prior to this day I had listened to John Cleese explaining this strange psychological phenomenon that hitherto I'd never heard of before. It was fresh in my memory, and upon my teammate's "No" I decided to claim my moment and cite John Cleese almost verbatim:
"And that's the case with you, I'm afraid," I added.
His reaction was the best I could've hoped for.
In a nutshell, the Dunning-Kruger effect is the connection between someone's competence in a field and their self-esteem in the same field. When the term "Dunning-Kruger" is used colloquially, as it is in this blog, it typically means that people with inadequate understanding and insufficient experience grossly overestimate their skill level and think that they're on top of the world, while in reality they hardly know what they're talking about.
I've yet to meet the first person who's immune to this phenomenon. I'm not above DK either: when I learned the first bits about music theory, there was a lot of music that I immediately began to disqualified as cheap, easy, and not nothing special, simply because the composition appeared simple. I knew virtually nothing about music back then but I presumed I knew everything. When one of my classmates said that a G# and an Ab sound differently, I confidently and boldly proclaimed that that was nonsense. And I was completely wrong.
In a graph, Dunning-Kruger looks like this:
One of the first times I encountered a clear case of Dunning-Kruger in chess was when I was flyering at an exhibition with my club. The grown-up son of one of our members wanted to play a game with me. He played what appeared to me to be random moves, so I won without troubles. He then asked me to elaborate a bit on the game, and I explained the three basic principles of the opening to him. In the subsequent rematch he played much better, and I only won because of an endgame mistake. I complimented him on having played a good game, to which his response ("Well usually I get through the opening fairly well, but the endgame, you know...") made readily apparent that the poor soul had already forgotten the rest of our exchange.
If you want to see Dunning-Kruger in action, I'd advise you to visit public comment sections like Twitter or the chatbox next to lichess. Go with the flow, absorb their points, and come to the conclusion that Caruana is an idiot for not finding the win in that endgame against Carlsen, that players who play the London are lazy draw-seekers, that Kurt Cobain's role in music history pales in comparison to that of Justin Bieber, that pineapple on pizza...
STOP! NOW!
For your health's sake, don't spend too long in these internet pitholes. They're one-dimensional echo-chambers that make you think that only intellectual outcasts and mavericks would ever disagree with their version of common knowledge. You get stuck at that DK-peak while you really believe that you're having meaningful exchanges of ideas.
The clearest signs of Dunning-Kruger are when people say that something is easy, when in fact it's not and takes much more effort than it might look at first. It takes great skill to make something look easy. Not everyone can be a successful actor, comedian or writer, although many people believe that they can.
Only those who are open to new information and/or experience will learn anything. So if you pretend to know everything, that's true in the sense that you won't be knowing any more than you do now.
If you really want to understand something, the best way is to try and explain it to someone else.
- Douglas Adams
Teaching can be a great way to cement your knowledge and understanding. On top of helping you clarify the information for yourself, you now have to apply it. Simultaneously, it adds an additional layer of experience with how things work. In turn, this experience can help refine the teachings and make you a better teacher. So a tip for improvers: start teaching. And a tip for teachers: let your students explain the material, to you and to their peers.
Experience will help refine your knowledge, and at some point you might begin to see that a lot of the advice and dictums that you've been given are much more nuanced than they seemed at first. Look at this position, taken from Silman's How to Reassess Your Chess, 4th Edition, Part Four: "Psychological Meanderings":
Is the rook really worth 5 points in the ensuing position above?
Case study 1: bishops and knights
Minor piece imbalances are some of the most frequent imbalances, and they're often misunderstood. Explaining the many differences between these could easily have been its own blog entry, but much has already been written about it.
Some general pieces of advice that you often hear are that "Bishops want open positions, knights want closed positions," "The side with the bishop pair has to open the position," "Knights want pawns on just one wing, bishops want pawns on both wings" and so on. While these aren't necessarily wrong, they're one-dimensional. Applying them without any regard to the specifics of the position opens the door for some very detrimental decisions during your games.
Along with their atrocious use of clock-time, my opponent lost this game because of a complete misunderstanding of the desires of black's pieces, most specifically the knight. It's clear that it didn't like the closed nature of the position because it had no action whatsoever.
What separates experts from DKs and other amateurs is a much more accurate and complete understanding of these principles and their nuances. I don't want to pretend like I know everything, but I've come to see the battle between knights and bishops as much more dynamic and lively than just "Open position = Bishops +1, Closed position = Knight +1." If we disregard the instances in which king safety takes precedence, a more accurate description in my view would be this:
Both bishops and knights want activity.
Bishops need open diagonals.
Knights need outpost squares.
There is a rationale behind the idea that the side with knights to want to close the position down, but it's not for the knights. It's against the opponent's bishops. Following the Kasparov quote, a closed position doesn't accommodate the opponent's bishops, so in a closed position the knights are better able to outperform the opponent's bishops.
Speaking of outperforming...
Naive as I was, I once used to believe that it's an admirable trait to know what you're talking about. But in the digital space of the 21st century, that's hardly the case anymore. It's more likely that you'll be seen as an insufferable know-it-all than someone whom others can rely on as a good source of information.
In the world of chess streaming, popularity is the name of the game. What sells matters more than what makes sense. As such it's no coincidence that the most popular chess streamer on Twitch is also the greatest example of Dunning-Kruger in chess: world #2 Hikaru Nakamura, who openly bashes people for not being able to checkmate with knight and bishop even though he can't do it himself either.
So the proof is in the pudding: you don't have to be a professional chess-player to create chess-related content and do well with it. Literally anyone with an internet connection, a Twitch account and streaming software can get going.
With more and more streamers being allowed to set up their streaming equipment during OTB tournaments, the number of commentary streams has also skyrocketed. The Maia Chess Festival in Portugal was a great example of this: 18 streamers playing on fixed boards with a camera pointing at their games. Some of this commentary is actually pretty decent. But a lot of it, well...let's just say it fits this blog.
Case study 2: inhumanly accurate commentary
In round 7 of the Andorra Open 2024, Australian streamer Hannah Sayce had come out of the opening with a fantastic position a piece up, but her conversion was suboptimal to say the least. The following position emerged after black's 29th move:
White, while being a piece ahead, is quite plainly losing. There's literally no white piece that's doing anything at all, while black is ready to crack open the little shelter that remains around the white king and attack with each and every piece from multiple different angles.
When Sayce played
30.Re2
her commentator shot the following snippet of prose into the ether:
Oh no! Now it's forced checkmate in 9!
Commentary like this is priceless. Most important of all, it's 100% true: the engine dropped from -5.5 to #-9. Next to that, it wonderfully allowed for the commentator to mask the fact that they had absolutely nothing to contribute of their own. They didn't even bother backing up these words by showing the mate sequence. It's a beautiful performance to parrot the engine so accurately.
I won't embarrass this commentator by revealing who it is, but in the following position (over a year later) they missed the following mate in 1 in a classical tournament:
White did win the game eventually, but this missed mate in 1 is the moment that people will remember. Good for content. Bad for chess.
If one niche of online chess content lends itself perfectly for showcasing the Dunning-Kruger effect in all its glory, it has to be PogChamps. A collaborative enterprise of Chess.com and Twitch, PogChamps is an online streamers tournament with stars from lots of different fields of expertise. These content creators, each with a very respectable fanbase of their own, compete in a chess tournament hosted by Chess.com, with a total prize pool of $100 000. It's a great way to make chess more accessible for a wider audience and establish crossovers between multiple different worlds of online content creation.
Because the contestants are typically novice to the game of chess, they get paired up with a chess coach who teaches them just enough to get their games going. By knowing just enough to feel some degree of confidence, the participants often find themselves precisely at the DK-peak. Combine this with the explosive theatricality that many of these public figures are known for, and you can have yourselves hours upon hours of Schadenfreude by watching PogChamps coverage on Twitch and YouTube.
I'll share two clips that were also featured in this highlights video.
The thumbs-up at 0:17 says it all: Ludwig has no idea what's about to happen, but the commentators certainly do. The cruel thing is that if Ludwig had started the staircase with the other queen, it'd absolutely have worked. Bad for chess, but absolutely perfect for content.
And then this moment, where MoistCR1tikal feels on top of the world for executing what would only be possible in this competition (though admittedly it's good preparation and guesswork from his team):
Note that in this last clip, the participants aren't the only ones to suffer from some sort of Dunning-Kruger, so they're with like-minded people there.
Personally I'm not a fan of PogChamps. I find it the trainwreck of online chess entertainment. And I wouldn't have bothered looking into it if it hadn't been for the infamous cheating incident that scarred Season 6.
Cheating is the unwanted dark shadow of competition, and PogChamps isn't immune to it. A streamer who goes by the name of DrLupo beat the top seed in his pool 2-0. It was quickly revealed that he had received outside assistance from an engine, and he was subsequently removed from the tournament and the site.
Since this incident has received a lot of media attention already, and the analytical efforts of Eric Hansen and Fabiano Caruana suffice here, I won't analyse the moves of the game in-depth. However, the language that DrLupo used during the second game of the match is interesting. I'll be quoting directly and add a few remarks of my own:
DrLupo displayed textbook Dunning-Kruger in his commentary while playing most of the game at engine-level accuracy. This should raise eyebrows. Several of the hallmark trades that you can see from his commentary were that he:
- has an inflated sense of confidence (move 12 and the rest of the game after move 32)
- was critically unaware of what was happening on the board (move 6, 22, 25, 29-32), plays multiple great moves that seem to come out of nothing (15.Rd3, 30.Ne6+, 31.Rxg7+) and that he can't properly rationalise (14.b4, 22.Nc2)
- starts thoughts and never finishes them (throughout)
Upon playing a nearly perfect game, DrLupo didn't know that it was checkmate until it appeared on his screen. DrLupo tried to act as if he knew how the game works, but he really had no clue.
Case study 3: Eugene Varshavsky
In 2006, a player by the name of Eugene Varshavsky partook in the 34th World Open tournament in Philadelphia. In round 7, he was paired up against the Israeli GM Ilya Smirin who was solidly in the top 50 of the world at the time. The odds were heavily against Varshavsky when he played this game:
Smirin recalls:
[A]fter I resigned, it was clear from his reaction that he didn't expect me to resign. Then, Varshavsky made the comment that 'those doubled pawns were the reason for your defeat.' Then I became suspicious.
- Ilya Smirin, cited in Winning the World Open, Joel Benjamin and Harold Scott, 2021, Chapter 24: "Cheating in Chess"
Someone with a very basic understanding of pawn positions might believe that doubled pawns = bad, because the doubling reduces their attacking power and some defence from neughbouring pawns. But, once again, it's much more complex and nuanced than just the observation that the pawns are doubled. I'll explain why Varshavsky's remark is an indication for DK in this specific instance:
The doubled pawns provide white with a semi-open a-file against a potentially backward pawn sitting on a6 (and later a5). On top of this, the pawns present white with extra control over the a- and c-files, on which some action may be expected to happen in the future. For example, they can prevent much of black's queenside expansion. So they served as a functional strengthening of white's control over the queenside.
What's more telling is that those doubled pawns were under no real threat at any point during most of the game. The square in front of the pawns was occupied by one of black's pawns, so they were never subject to a frontal attack. Nor was their apparent weakness any real factor in the strategic middlegame battle, which took place in the centre and on the kingside. Only when white had lost that battle did those pawns become vulnerable.
Making a comment on these doubled pawns being the reason for white's downfall shows that Varshavsky had no idea what was going on in the game. Their falling weak was the result of white's collapse, not the cause of it.
Although hard evidence of his cheating has never been found, the circumstances have indeed been highly suspicious. Michael Aigner asserted that Varshavsly's final 25 moves matched Fritz 9 perfectly, and someone with his utter lack of chess comprehension would be hard-pressed to find them under tournament conditions. On top of that, his two follow-up games (the first being one in which he held himself up in a bathroom stall for 45 minutes) were simply abysmal:
The first time this guy Eugene Varshavsky was mentioned again in a press release was when he got himself entangled in a cheating scandal in sudoku.
For further reading on the World Open 2006, suit yourself:
http://chessninja.com/dailydirt/2006/07/06-world-open-concludes.htm
Picture © by FM (now IM) Eric Rosen, taken at the World Open in 2014.
There's one other branch in which this method of emulation without understanding is the name of the game:
Writing with artificial intelligence
When the Carlsen-Niemann beef erupted in 2022, this video popped up in my algorithm in which Niemann mentioned that Alireza Firouzja "hates being attacked." I remember having a discussion with an acquainted IM who claimed that Niemann did his psychological research well, while I maintained, "Well, doesn't this line just sum up every chess-player?" I tend to call this type of language "Astrology" language. It's vague, emotionless, and so generic that it could apply to almost everyone. It sounds clever and based at first but it's completely hollow and empty.
AI-generated text belongs to the same category of porcelain text. It looks profound, but has no substance to it and falls apart if you put it to the test. Its learning method is that it consumes a lot of pre-existing texts and condenses this into an estimate of what fits your request. That's a long way from knowing how it really works. And in fact it doesn't know how things work. This is also why it can't draw a glass of wine filled to the brim:
I'm 100% sure that you will be able to outperform the AI image generators on this one. I already did so on a different occasion:
Case study 4: GPT explains checkmating with knight and bishop
I believe I made my point when I wrote my post about checkmating with knight and bishop: I can deliver it, and I'm skilled enough in it to explain how it works. But to see what it comes up with, I decided to have an exchange with GPT about it. I can share the entirety of the interaction with GPT in a separate Appendix entry, but for now I'll highlight a few points that it came up with:
Phase D: The Mating Net
Once the king is trapped in the correct corner:
-
-
- Your king cuts off two adjacent escape squares.
- Your bishop covers diagonals.
- Your knight delivers the final blow, giving checkmate.
-
For example, with a light-squared bishop, checkmate in h8 looks like this:
-
-
- White bishop controls g7.
- White knight delivers mate from f7.
- White king blocks g8 and h7.
-
The enemy king has no squares left.
If we take GPT's descriptions and apply them to a real chess-board, it would look like this:
Nailed it! Perfectly!
AI chess writing typically includes no board images, which makes it hardly more than flyer text. In most cases, chess-educative text can't stand on its own: it simply needs diagrams. In the first chapter of Move First, Think Later, author Willy Hendriks makes the following point with which I fully agree:
Positions are not examples illustrating more general principles – they constitute the actual learning material.
- Willy Hendriks, Move First, Think Later, Chapter 1: "First move, then plan, then judge"
(N.B.: If you think that this line looks AI because of the em-dash and the juxtaposition in the text, please observe that this book was released in 2012, long before any of the present-day AI writing tools were developed.)
However, GPT did suggest to provide diagrams, so let's see what it came up with. According to the great 21st-century writing machine, this constitutes a checkmate in the right corner with the light-square bishop:
To explain why this is checkmate, it came up with this gem:
- Knight on f7 delivers checkmate.
- Bishop on f1 covers g2 and h3, but more importantly, it controls g6 (one escape square).
- White King covers g7 and g8.
- Black King on h8 has no legal moves.
Checkmate!
Earlier this year, an MIT study was conducted into the use of artificial intelligence in writing and the effect that this has on people's cognitive abilities. The study found that after having created their text, 83% of the research participants couldn't recite a single line they had just produced minutes earlier. The group of test cases wasn't significantly large enough to draw any conclusions (just 18 test subjects, of which 15 failed), but it does point in the direction that outsourcing your own research and writing to external entities doesn't make you any more skilled in the field.
Artificial intelligence doesn't make people dumber, but it certainly does make for a lot of Dunning-Kruger content. And so the cycle goes. Although it must be said that unlike my teammate in that car back in February 2019, your AI machine definitely knows (i.e.: is able to reproduce) what the Dunning-Kruger effect is.
The Dunning-Kruger effect is a very common phenomenon. It has some hilarious manifestations, but it's more than just a laughing matter. There can be some very serious consequences to blindlessly following the nonsense of others simply because they come across as confident and knowledgeable.
I firmly believe that teaching and explaining is a great method to reinforce what you know, so I'm very much in favour of commentary streams. I dislike looking at the mouthpiece of an engine, though, so I will put them to the test with questions that you can't use the engine for, like "On which wing of the board would you plan to mount an attack in this position?" or "Would you trade queens here and why?"
Not all streamers are DK-dummies. There are several titled players among the streamers who provide solid and accurate commentary. Although I love to poke fun at him for botching the knight and bishop checkmating, I'm very happy that as the strongest regular chess-streamer on Twitch, Hikaru is also the most popular one. There are several other strong titled players, but I won't be doing a roll-call here.
Staying with Twitch for a second longer: DrLupo's story may have a good plot twist to it. Having realised how much he has ruined things for himself, he has offered to cover the entirety of the expenses for PogChamps Season 7. It'll be up to the community at large to decide what to do with his case, but this attempt at repentance and redemption is better than nothing at all. It's not like he scammed his fans for a $12k poker seat and then gaslit his fans for calling him out for his nonsense.
There's one more thing that I owe to tell you. After rubbing the Dunning-Kruger effect in my teammate's face, he started laughing. He recognised it as a well-construed burn, gave me a high-five, and essentially stopped bothering me as much after that. In terms of social resilience, standing my ground in the face of animosity like this was perhaps my greatest social victory to date. It taught me something that I learned about youth players as well: if you cut what you fear to size and see it for what it is, what you fear is not nearly as bad as you thought it would.
The team won the match with a two-point lead, and both he and I drew our respective games. I don't speak to him that often anymore these days, but I'm quite sure that he's incorporated the DK question into his verbal arsenal.
I'll finish with a clip of Michio Kaku telling my favourite Einstein joke. It has a nice component of Dunning-Kruger in it too.
