Humans vs Machines
Howdy! It's The Chess Friends!
As you might expect from the title, in this blog we are going to talk about how chess engines and humans battle and collaborate.

Previously, we discussed how different chess engines come up with their moves and how we, The Chess Friends, use computers in our chess studies. Today, we will cover the history of human competition against machines, and the ideas behind correspondence chess matches, when a human player consults with an engine during the game.
The Man Benji
All About History
Before starting with history, I would like to talk about … please bear with me …
Drosophila
…also known as fruit flies: those pesky insects that appear seemingly out of nowhere when bananas on your kitchen countertop sit for more than a day.
One specific species of fruit flies, Drosophila melanogaster, is a legendary Nobel-Prize-winning fly because it’s a model organism in genetics and other scientific fields. In 1933, Thomas Hunt Morgan won The Nobel Prize in Physiology or Medicine for his discovery (using Drosophila) that, inside cell nuclei, chromosomes store genes, the factors responsible for inheritance of different traits.

In 1955, The Nobel Prize in Physiology or Medicine was awarded to three scientists who discovered genes that control formation of early embryos of Drosophila. In 2011, Jules Hoffmann shared The Nobel Prize in Medicine for studying how Drosophila activates its innate immunity. In 2017, The Nobel Prize was shared once again, for discoveries how all living organisms (starring Drosophila of course) adjust their biological rhythms to different phases of the day.
I hope you are convinced by now that Drosophila is highly regarded in science. After all, thanks to this little fly, we understand so much more about human physiology and medicine.
You might think: it’s all good, Benji, but what in the world do these flies have anything to do with the history of computer chess?
Well, there is this saying, attributed to Aleksandr Kronrod, a brilliant Soviet mathematician and computer scientist: Chess is the Drosophila of Artificial Intelligence.
You wouldn’t believe when he said that: In the 1960s! Truly a man ahead of his time! Allegedly, he compared chess to Drosophila of artificial intelligence when theoretical physicists at his institute were complaining that he was using up precious computer time to play chess. Only he wasn’t playing some games. He used chess like other scientists used Drosophila: to experiment, test his theories, collaborate, and, ultimately, teach computers how to think. Back then, not many people paid attention to those experiments. Computer power was rudimentary, the algorithms needed improvements, and human chess players always won.
How times have changed! Nowadays, chess engines can take down the world’s top players, finding tricky ideas in just a span of a second. We take our engines for granted, relying on their computing power to spot tactics we missed in our game, to come up with a plan in the opening, and to sharpen our technique. For us chess kids, it’s hard to picture a chess world without chess engines, and maybe even harder to imagine that once, the machine was the underdog in the battle against the human.
People have been fascinated by the idea of a machine playing chess for a long time. In 1769, the first chess computer, aka “The Mechanical Turk” was invented by a Hungarian engineer Wolfgang von Kempelen. Only back in the days, it wasn’t a true computer but a cleverly engineered device with a secret compartment from where a human, a very good chess player, operated the game. We talked about the invention in one of our early blogs. The Mechanical Turk’s secret was well guarded, and people lined up to play what they perceived as a chess robot!
A few other chess automatons similar to The Mechanical Turk came and went, but in 1912, the first true chess machine capable of making moves without any human assistance was invented by Spanish engineer Leonardo Torres Quevedo. Called El Ajedrecista, it was a carefully constructed electromechanical device that used simple algorithms to win a number of endgames against any opponent.
Once computers were invented, though, no one could stop humans to dream of, build, and keep on improving chess engines that would win matches after matches against any human being. In the process, humans have learned more about themselves, how their decision-making works, and what is the intelligence.
Colossus was the world's first electronic digital programmable computer, created at Bletchley Park during World War II in Britain. After the war, in 1948, Alan Turing, a famous codebreaker who worked on Colossus, started writing a chess computer program Turbochamp. He wrote his algorithm, but the problem was that at that time, the most powerful computer didn’t have enough power to work through the complicated chess code. It was an start to modern-day chess engines, even though the algorithm wasn’t the best chess player around the block as it calculated only 2 moves in advance.
Still, Alan Turing, determined to test his algorithm, set up a match against one of his co-workers, where he would calculate the program’s each move by hand using his algorithm, taking 30 minutes per move! In that match, the chess program didn’t quite work because its queen was lost to a pin. Another match was played, against a novice player, with a move favorable outcome for the program.
Meanwhile in the USA, Claude Shannon, another mathematician and codebreaker, worked on his version of the chess program. Based on the game decision tree, he calculated that there are 10120 chess games that can possibly arise, a number impossible to sift through by any computer at that time. Thus, the work began to optimize the computer program’s decision making.
The work was also underway to advance technology; computer calculating potential was becoming deeper, and more attention was tuned to perfect chess computer programs to test the up-and-coming computer power.
One curious example from the 1970s was chess engine Mac Hack, which was the first computer chess program to participate in a human chess tournament. USCF (United States Chess Federation) graciously allowed Mac Hack’s developers to test out their program in many tournaments. It even received a membership and a rating! It started with a rating of around 1200, and progressed to 1500s level play, winning against some tournament-level players, setting the stage for future Man vs Machine battle.
Enter Deep Blue.
In the 1980s, Fengshiung Hsu, an electrical engineer and computer scientist, teamed up with tech company IBM, in order to create a chess supercomputer, which they lovingly called Deep Thought. Building a supercomputer is no easy feat, and even though lots of help was involved to make it, the project took over many years.
The IBM team entered Deep Thought in a few computer chess championships, usually winning them all. It had even become the first ever chess engine to win against a Grandmaster (Bent Larsen) in a tournament game.
Not stopping there, the IBM team upgraded the engine, renamed it Deep Blue, and in 1996, it was ready to challenge the world #1 at that time, GM Gary Kasparov, in the first official edition of Man vs Machine, the long-anticipated battle that took place in Philadelphia over best-of-six matches.
Starting from the beginning of the match, Deep Blue showed that it was not going to be an easy opponent by winning the first game. Although, with the pressure of representing “Man” on his shoulders, Kasparov managed to bounce back and win the match convincingly 4-2, drawing 2 and winning 3.
The Deep Blue programmers didn’t give up. They went back to the lab and upgraded their supercomputer more, doubling its efficiency by allowing it to calculate 200 million moves per second! Then the rematch of Man vs Machine was set for 1997.
Once again, with the pressure to defend humankind intelligence on his shoulders, Kasparov traveled to New York to greet his old rival.
Kasparov won the first game in only 45 minutes, due to sloppy play by Deep Blue, so all seemed well for Team Man. The tide started to turn in game two, as Deep Blue played a great game, overpowering Kasparov. As it turned out, though, Kasparov could have saved a draw, by perpetual check. Can you find the game-changing sequence?

That’s right! The moves that save the game for Black is…
45… Qe3!
After that, we can expect White to play 46. Qxd6. Then, Black’s one saving move is…
46… Re8!
This move stops all of the sequences White can play against Black’s king! Now, if White plays 47. Bf3, the key perpetual check sequence starts with 47… Qc1+, and from here, all Black has to do to draw the game is to be careful and keep checking the White King. White cannot escape a repetition.
Kasparov didn’t find this clutch game-saving sequence. In an interview after the game, he said he was shocked that an engine could play so well, so human-like. It was definitely unfamiliar territory for him, going against a strong opponent without any preparation and previous games to study.
Deep Blue brought something new to the chess table no one has seen before: a relentless mind that is never tired, never hungry, and never has to go to the bathroom. Plus, the computer had no pressure playing, while Kasparov had to defend human intelligence and prove to the world that he is still the best-ever chess player!
For game three, Kasparov chose an unusual opening, trying to take Deep Blue out of prep. This strategy worked well; Kasparov was pressing for the win, but the chess computer skillfully defended the position, and the game led to a draw. Similar results happened in the fourth and fifth games.
During the last game, however, Kasparov, probably really feeling the weight of all humankind, blundered a tactic early and gave Team Machine the last laugh and a 3.5-2.5 win!
If you want to relive these historical moments and get a peek on mindsets of both sides of this high-stakes match, I would recommend reading the graphic novel Mind vs Machine by a collaboration of an esteemed writer and a legendary grandmaster:
Mind vs. Machine: The Chess Match That Changed the World by Steve Sheinkin, Maurice Ashley, and Thien Pham (Illustrator)

The rest, as they say, is history. Though team IBM discontinued Deep Blue, it will be remembered as the supercomputer that formed the stepping stones for modern engines that could crush any human in their sleep (or in their shut down state), like Stockfish or Leela.
The Myth Sarang
All About Correspondence Chess
For those unfamiliar with correspondence chess, it’s an amazing way for people to connect all over the world and enjoy learning through chess. In the old days, moves were sent by mail. Players were given up to two months sometimes to make a move! Now a lot of people can play correspondence chess online, playing many games at a time to better positional awareness and approaching every position with a fresh mind. At one point I had around a hundred correspondence games going at the same time and it was great. Every time I logged in, I’d have a dozen to fifty “fresh” positions to look at and even though I may not have remembered my plan perfectly in every position, I could see things with new perspectives.
While online correspondence chess on sites like GameKnot, Chess com and such follow faster chess rules of no consulting books or engines as they are considered cheating, the International Correspondence Chess Federation (ICCF) actually encourages such consulting. I’m going to talk a bit about why. Simply put, the goal of correspondence chess is a great game and learning. Classical and speed chess tend to be tests of thinking, speed, creativity, instinct, and perhaps above all else, memory. Correspondence chess allowing books and engines removes the memory and speed parts in a sense. You can take the time to learn, analyze, go deep into a position on your own time, and find each move in a very intentional way.
Correspondence chess remains valuable because it turns chess into a long, transparent project: with days or weeks per move, players can investigate positions deeply, test ideas against engines and databases, and learn why certain plans work. In the modern ICCF setting, it is openly a human-plus-engine contest, not a test of who can avoid using a machine.
Here are some reasons why correspondence chess is valuable.
Tactical oversights are largely removed, so the challenges are evaluation, planning, pawn structure, piece placement, endgame technique, and understanding long-term compensation. This makes it great training for the deeper parts of chess.
Correspondence players have historically expanded opening theory by exploring lines far beyond normal tournament preparation; novelties tested by correspondence players often later appear in faster over-the-board games. This idea of growing the collective knowledge is really cool to me, kind of like being a science researcher in a lab.
A single game may last months or years. That format rewards methodical candidate selection, blunder checking, written analysis, and willingness to revise an initial judgment. Unlike speed chess, it’s much less about preparation and instinct and much more about deeper thinking, analysis, and utilizing the current collective knowledge and resources to potentially change the game further.
Let’s look at a game between Jon Edwards and Trygve Hagen in the semi-final round of the 32nd World Correspondence Chess Championship in 2022. You can read his thoughts and see some of his games here: SIM Jon Edwards wins 32nd World Correspondence Championship | US Chess.org. Jon went on to win the event and become World Correspondence Chess Champion! I’ll focus on some of the critical human thinking that overlays the computer engine analysis.

Edwards’ plan was to expand on the queenside and create a passed pawn, while arranging his pieces and even his king to support that expansion. The resulting pawn imbalance, he believed, should make a queenside passer. In this position out of the Sicilian Najdorf, he wanted to gain space and restrict Black’s queenside. a4-a5 would establish the advanced pawn while 12. Na5 put pressure on b7. If Black were to advance the b-pawn to b6, Edwards intended to settle the knight on c6 which would be an outpost square.

This was a pivotal moment in the game. Edwards wanted to keep the a1-rook available for the pawn advance because if he had castled queenside, the a-pawn would become very weak because his rook would be on d1, away from its intended work. Edwards played, in order, a4-a5 locking down Black’s queenside pawns then Ra4-b4, activating his rook in support of his queenside operation. In a position like this, if Black relies on optimal engine moves, a natural choice given the accessibility and the option, White will break through. Black doesn’t have a clear long-term PLAN. This is what engines struggle with. Planning. If Black plays the top engine moves, sometimes the top moves are moves that no human would understand. You have to UNDERSTAND the long-term plan of the top engine moves to play them.

Back to the game. White has executed his first plan and now wants to get his king to safety without disturbing the rook. With Kd1 and Kc1, he effectively “castled by hand” or “manually castled”. This solved the king-safety plan while leaving the rook where the plan needed it. Kingside castling, by contrast, would have placed his king on the side where Black had already advanced … h5 on.
Edwards’ plan was not just to put a rook on b4 and capture b7. He explicitly stated that he could not have imagined, when playing Rb4, that b7 would later fall without a fight. This part of the game gave him an advantage that was enough for him to go on to win the game and make the finals.
In summary, correspondence chess that utilizes engines is a great way to grow the collective knowledge of the game. Having full access to all the current computer and human knowledge in planning each move gives us a platform to expand how we play great chess. We can use the deep analysis of engines coupled with the deep thinking and planning of humans to create novel positions and situations that will motivate classical over-the-board play. Now, let’s turn it over to my bro The Legend Vivaan for more!
The Legend Vivaan
All About Humans
Thanks bro. This reminds me of an article we wrote about a year ago: Chess: Art, Science, or Sport? In that piece, Sarang discussed how he views chess as a science, where he can deep dive into each position and seek out the optimal move. I talked about how I view chess as an art, where I can be super creative and innovative and come up with crazy ideas, even if they’re suboptimal. I enjoy leaning into psychology. I enjoy assessing how my opponent is thinking and crafting how I play around that. I like to trust my instincts and sometimes in the moment I just want to sac a piece or two even if it is unsound, knowing I can outplay my opponent in the situation that comes and understanding that my opponent is human and won’t come up with the “best” move every time.
While I see Sarang’s points on learning and finding the best solution to each problem or position, I think it can also lead to everyone playing the same way. It can make chess a bit boring. In some ways this is my issue with long classical chess games too. Rapid, blitz, and bullet chess force time pressure, force you to think fast and be creative, and allow you to take chances and make mistakes and have an amazing learning psychological dance with an opponent in the moment. It also allows for me to emote through my play, release whatever I’m thinking about or feeling in that moment, through the pieces. It needn’t be the same every time. One game can be super different from the next. It’s authentic, unprepared, raw, and intense.
I study chess a lot. I love learning from the greats, learning from my old games, and even learning from engines from time to time to help me point out the moments where I could’ve thought a bit deeper. That said, when I play a game, I like to be fully present in that moment and run with my ideas, however crazy or suboptimal they are. One of the fun things I like doing when playing computers, bots, AI, engines, doesn’t matter, is thinking about how they play differently from humans and where their coding breaks.
Simply put, computers play as they are coded or as they are trained to adapt based on human play. They are materialistic, which usually works, but can struggle in locked positions and after sacs that trick them into believing they’re much better. You have no idea how many bots I’ve beat with tricky sacrifices and bad play from the machine getting too passive after gaining a material advantage. On my many long flights with seatback video, I enjoy playing lines like the Fried Liver and Scotch Gambit, sacrificing material for initiative and attack. Usually they work against the bots, even ones labeled “Advanced”. Now you’re probably saying, ok ok but those are not engines. You’re right.
Give this article a read when you have some time: 10 Positions Chess Engines Just Don't Understand - Chess.com. Check out the blitz game between Rybka, an engine supposedly much better than a human, and Hikaru Nakamura. Hikaru creates a super locked position with no open files. He allows the engine to win two exchanges and enters a dead drawn endgame where the engine has two rooks and eight pawns while Hikaru has a bishop, knight and eight pawns.

Rybka, because it is +4 materially, wants to push for a win, and doesn’t recognize the position is dead drawn given its rooks are useless behind all the locked pawns. After a lot of wiggling and the threat of a 50-move-rule draw, it sacs a pawn, which gives Hikaru a passer, but it’s still theoretically drawn. Then Hikaru does something even more brilliant. He puts the engine in a situation, where it’s +3 materially but if it doesn’t accept a 3-fold repetition draw, it’s dead lost eventually, as he creates an additional passer. Confident it will win with the material advantage, the engine persists, gives up the extra passer, and proceeds to lose. Sarang talked about how humans have the edge in long term planning and thinking. This is a great example of that. It’s different for an engine to calculate many more moves deep than to truly understand the long-term WHY!
At the end of the day, engines do a great job of giving us new ideas and analyzing how we play. The problem is, we as people, over trust the engines and lose our individuality, style, and creative minds in the process. Correspondence chess, as Sarang mentions, offers a chance to do a bit of both, which is cool. For me though, I prefer good old over the board speed chess where me and my opponent can engage in an epic, fun battle of our minds, where we can both use our creative brains to the fullest and don’t feel any pressure to play optimally or how computers would.

At the end of the day, computer chess is here to stay, and we can learn as much from them as they can from us. We can be each other's metaphorical Drosophila melanogasters. Hopefully this blog has given you something fruitful to think about as the flies enjoy your leftovers.
Three Cheers, Fellow Future Master Chess Friends!
The Man Benji, The Myth Sarang, The Legend Vivaan



Comments