Chapter 1-2: Versus「 」

21.3K 749 561
                                    

Lelouch's POV

As it turns out, it seems I had greatly underestimated Blank. I've learned a great deal about these two siblings just through this chess match. I opened my eyes, temporarily breaking out of the deep concentration I had been in up until now and looked over my desk at my opponent.

Sora's sly grin has still not left his that face of his, and Shiro was as stoic as always. Shiro was a person who operated based on cold, calculating precision. Her ability to think was extraordinary; throughout this game she had flaunted her immense calculating ability that puts even chess engines to shame. If it was any other game but chess, I'd think she has some kind of precognition - it's almost as if she has a map in her head which shows all possible continuations with terrifying depth and accuracy.

That being said, I think I would be able to beat her. If anything, her being like a supercomputer is both her biggest strength and weakness. Identifying traps and tricks, and intuitively knowing which kinds of moves - like an early b6 push by black for example - can result in long term weaknesses which will be exploitable down the line. Shiro, as well as most computers, can avoid this for the most part by analysing the position at a depth of around twenty moves; if a move won't create a big weakness within that amount of moves then a computer will judge it as a good move, especially if it accomplishes a short-term goal or prevents an opponent's tactic.

There is a problem with this, however. In fact, it has a name: The Horizon Effect. Also known as The Horizon Problem, it's a name for the issue wherein a game such as chess has so many possible game states that computers and AI can only feasibly evaluate a small percentage of them. 

Human beings can overcome this weakness by using their intuition and pattern recognition. A human chess player for example will easily be able to identify a variety of traps and tricks in a game by virtue of recognising them. Once a player falls for a trap once, they won't do so again because they'll remember it next time. A computer however has no simple way of doing this - the only way a computer can identify a trap is by evaluating the game state many moves down the game tree.

I had fallen behind a bit but when I realised all of this, I figured out how I can win. By exploiting tricks and traps that would have a negative effect mainly on the positional play of the game - something Shiro can't easily see, if at all. 

"Knight g6" I said confidently.

Out of the corner of my eye, I noticed some of the observers look surprised for a second. Namely, Ayanokōji narrowed his eyes when he heard my move - it was the first time a noticeable change appeared on his face since he entered this classroom. There are two possible reasons a person would have this reaction to my move and my evaluation of Ayanokōji would change drastically depending on which reason he gave for his reaction. I made a mental note to ask him later.

On one hand, he could be surprised because on the surface it looks like I'm simply giving away my knight for free. However, this was the trap I knew that Shiro would fall for. If the pawn on f7 takes my knight, it will open up a diagonal which the pawn previously blocked which I can use to force Shiro's castled king from out of its safety into the centre of the board. From there, it will be easy to attack and will eventually result in checkmate; even for a supercomputer, recovering from a position where your king is isolated is near impossible.

Despite this, the move I had played - knight to g6 - was still objectively a bad move. Why? It's because black can respond with a passive move that covers the diagonal. By doing this, it accomplishes two things - it makes black's pieces more active and developed and also forces me to retreat my knight back to where it was which effectively means I wasted two moves meanwhile black has been developing their pieces and building an attack.

The True ElitesWhere stories live. Discover now