Between a battle of chess engines, the one with the stronger fixed depth search is likely to win although not certain. Other algorithms are surfacing which may have lower depth but can beat a stronger one, like the neural chess computer.
Computers have a reputation for being the strongest chess-playing entity that has been witnessed by mankind.
It is supreme, it’s hard to see the strongest human beating the strongest chess computer anytime soon which is a testament of their power.
This is why some beginners may ask the question what happens if two computers play chess? After all, it will be very interesting to see the outcome.
I have watched a lot of battles between two chess computers so I think I am qualified to answer this question, and I tell you they put on a good show. That is what I’m going to explore today, let’s start.
Do chess computers have different strengths or are they the same?
Chess computers are not equal, some are stronger than others as they are ranked in the world chess championship. Engines like Stockfish, Komodo, and Houdini are the top dogs while neural computers like Alpha zero and Leelachess might even be better than them.
It used to be that chess engines utilized the same methods (brute force calculation). Neural networks came and changed everything, I have dedicated a full article to this change. Today, chess engines have different strengths and weaknesses.
Chess computers just like people are very different, they are a product of various creators with their own algorithms. The term chess computer just generally refers to a broad term where each engine has a unique functionality of its own.
The point being that there is a hierarchy even among chess engines, they are not created equal. Some are way stronger than others and would just crush weaker ones without even doing major considerations.
How do we know this? There’s an actual tournament exclusively for chess engines (World Computer Chess Championship) where the top dogs are ranked.
Some of the strongest engines are Stockfish, Houdini, and Komodo, where one from the upper-division will naturally beat a weaker one.
If Stockfish for example played a game against Bobcat engine, Stockfish will most likely crush the latter. They have some unique differences yes, but most of their other algorithms run on the same idea.
This means that an engine with a stronger processing power will not always win, but most of the time will win against a weaker one. What’s interesting is there are rules in the event where computers are even allowed to resign or draw their games.
This means that we can have a look at their uniqueness when one declined a draw since we know that not all of them run on the same principles.
This adds a little bit of computer psychology that engines are not necessarily always agreeing with each other’s evaluations.
Is the result of the battle between chess computers always a draw?
The results of two chess computers playing would be a draw approximately 80% of the time as seen in the world chess computer championship, though 20% are decisive where one side gets the victory.
More or less the results in the World Computer Chess Championship are a draw at least 80% of the time, which is why most people don’t watch it. Statistically, it would not be insane to mark the battle between two engines as most likely a draw, but not always.
I mean, even in human chess, the sheer number of drawn games is becoming a problem amongst games played between super grandmasters.
Of course part of the problem is the engines being approximately the same strength, but that’s not all of it.
The fact that human chess also involves a lot of draws means that the game has turned into something more draw-oriented, not necessarily because of strength.
However, it cannot be denied that the drawn games of engines are still more in number even in comparison with human players, since they can calculate better.
But don’t think there isn’t anything decisive between matches of computers, since even with the same strength one could still win and not draw.
If two computers of the same strength (same depth) would play the result would most likely be a draw yes, but will not always be the case.
In fact, even if it was against a lower division computer like the Stockfish against Bobcat example from before, Bobcat can still win some 1 out of 100 games.
It’s not because one will blunder, but more of the opening choices, the endings, or differences in their strategic schemes.
Will the one wielding the white pieces win in a battle between chess computers?
In a battle between two chess computers, the one with white does not always win and has more or less equal chances as the computer playing black.
Has been seen with the results of the world chess computer championship, the outcomes are more or less equal for white and black.
The one with the white pieces will not always win since there are instances where the one with black gets the victory.
This does not mean to say that the one with white will not have an advantage, since the position that emerges does favor white with having more space.
There are just too many factors in play like the computer depth, opening choices, or pawn structures, the color is a factor but not the entire story.
Even a weaker engine playing black has demonstrably been capable of beating a stronger computer playing white. Though I have to say, even with the difference in numbers between the seasons, white is likely to win but not always.
Believe it or not, there are in fact advantages to playing black even in a match between engines.
Certain openings played with the black pieces are actually better than some openings played with the white pieces for example.
Will the one with stronger depth always win in a battle of chess computers?
A chess computer with a stronger depth search will beat the one with a lower number, although if the margin is small the one with smaller depth can win.
Every computer plays around with the idea of a fixed depth search, however these things keep getting better and the results are becoming unpredictable.
In a nutshell, computers calculate various variations in the future of the search tree, the depth is how far they can see.
If you want to learn how chess engines evaluate chess positions, you should read my other article about this. It will introduce you to the concept of a search tree.
This is why computers of significant strength will not blunder, since they can literally see the positions in the future before it has been played. This is a critical consideration when determining the results of an engine match, who has the better depth.
Different computers will have different depths they can see into but as have been stated before, it is not the only factor (depth) that is affecting the outcome though it is important.
Engines are revolutionizing year after year and we are slowly moving away from relying on depth searches for the strength of computers.
Otherwise, if the depth is the only factor then it would be impossible for a weaker computer to beat one with stronger depth, which did happen before.
Though statistically since the one with stronger depth has more wins than a weaker one, we can say that they are still stronger than the one who relies on other things (other than the fixed depth search) to make them strong.
Will neural chess computers always prevail against traditional ones?
The strongest neural chess computer Alphazero has demonstrated to beat stockfish 8 and 10 most of the time, although its sister Leelachess have been struggling with stockfish 10. This implies that it is still not certain if neural chess computers are better than traditional ones.
There’s a new batch of neural networks (a.i) among the likes of Alphazero and Leelachess that actually learns over time. This is not the traditional engine that relies on fixed depth search and other conventional methods, this one plays similar to a human.
So who would win against a common engine and a neural network? it’s interesting to see who would come up on top.
It has been demonstrated that neural computers pack some serious punch against other engines, Alphazero in particular has been regarded as anonymously the strongest engine ever.
And Alpha zero is a type of a neural computer that taught itself chess in about 4 hours, already able to beat stockfish 8 at that point. After Stockfish 10 has been released, their battle has been reinstated and Alphazero has dominated the matches completely.
Similar success has been found to its sister neural computer Leelachess, although it was not able to beat stockfish 10 in the finals of the world computer chess championship.
The major downside of these neural computers is the time it takes to learn, the traditional engine can become strong right off the bat.
There was a campaign before where people could download Leelachess and play it in order for it to gather experience and become strong over time.
So far the results of the matches have been in favor of the neural computers, and they could beat common engines most of the time.
But traditional engines have improved as well like stockfish 10 for example, so it might be hard to distinguish which is better over the other.
Do you now know what will happen if two computers play chess?
Chess really is a complicated game that requires various strategic processes to play well, chess computers have nailed those down. So matches between computers have become the wonders of the new world, something that we are not able to enjoy before.
This does not mean that it will replace human chess but will definitely give us an interesting perspective on the game we love. They are getting better and better everyday, fixing their flaws and expressing insights to solving the game of chess.
Their battles are unexpectedly decisive than what you may have imagined a battle of machines would be (which would be drawn), I love it, sleep well and play chess.