What Makes A Good Chess Engine?

Today, whether we like it or not, computers are the best chess players.

Through advanced and various programs, chess engines can consistently dominate human players. Our question today is: How do they do it? And what makes some chess engines stronger than others?

Chess Engines: The Basics

Before looking at what makes today's engines so effective, let's take a quick look at the history and workings of chess-playing computers.

A Very Brief History

Chess-playing computers have been around for almost as long as computers have. Notably, Alan Turing created a rudimentary chess program as early as 1948. Before that, machines like “The Mechanical Turk” were presented as playing chess, but they were hoaxes – there was always a hidden human operator. 

Since then, chess engines have matched the rapid progress of technology over the last half-century. By 1997, a chess computer (IBM's Deep Blue) was capable of beating World Champion Garry Kasparov under tournament conditions. Massive supercomputers were soon shrunk down to the size of chess programs that could run on a laptop. The engines themselves rapidly progressed, becoming more and more challenging to beat.

Today, there is not a human alive who can match the top chess engines. 

How Do Chess Engines Work?

We know that chess engines can perform – but how do they do it so well? Until recently, chess engines worked using pure calculating power. As a game of chess develops, there are an immense number of possible variations, and engines are capable of “playing out” and evaluating these mathematically staggering possibilities. They do this by creating a “tree” of variations, which allows them to predict a series of moves stemming from each possible legal move in every position. 

They evaluate promising variations based on positional and material factors, like:

• Piece value
• King Safety 
• Pawn Structure
• Important Squares
• Tactical Possibilities

Gradually, these evaluative and search functions will be able to narrow down the best move in a position.  

This calculation-based method is being combined with, and in some cases replaced by, a machine learning technique based on neural networks. Rather than following the input criteria for evaluation, neural nets learn how to play chess through self-play and deep learning. Some argue that it has given them a more creative and human approach to the game. 

AlphaZero was the first neural net engine. Now, most of the top engines incorporate machine learning approaches for the evaluative functions. This includes the world’s best chess engine: Stockfish.

Further reading: Everything You Need to Know About Chess Engines

Top Chess Engines

Chess-playing programs are now commercially available through a variety of platforms. Many are also open-source, meaning their code is freely available. Let’s take a look at a few top chess engines to see what sets them apart. 

Stockfish

Stockfish is a free-to-use, open-source chess engine. It also happens to have the highest playing strength in the world. Through dominance in competitions like the Chess.com Computer Chess Championship (CCC) and the Top Chess Engine Championship (TCEC), it has consistently proved to be the most accurate and powerful engine. 

Initially, a purely calculation-based engine, Stockfish now uses a neural network system for its evaluative functions. 

If you want to try out Stockfish, it can be downloaded or accessed through a user interface on Chess.com (though this may not reflect the engine’s full depth of analysis).

Leela Chess Zero

Leela Chess Zero was developed based on the neural network model established by AlphaZero. Now, it often takes the second spot in the Computer Chess Championship. 

This powerful chess engine is also an open-source project and can easily be played against online. Bots available on Lichess, for example, are powered by Leela Chess Zero.

Komodo

Another forerunner in the chess engine world is Komodo. This engine has been around for much longer, but has continued to develop and grow to keep pace with top programs. 

In recent years, Komodo released the Dragon development, which incorporated neural network technology. Dragon took second place in many Computer Chess Championship events. 

Since being purchased by Chess.com, it can also be played on that platform. 

Why Are Some Chess Engines Better Than Others?

This brings us to the main question of our article: What makes an engine good?

A chess engine is a calculating machine. For many, while they can win games through deep analysis of positions and massive computational power, they lack something of the creativity, psychological competition, and strategic thinking that make the game beautiful. 

However, machine learning technology, used in all top chess engines today, is closer to the human way of studying and playing chess. They learn by processing games and playing against themselves. The result is that they are extremely powerful calculators, but arguably have a more detectable style. These new elements of creativity, combined with the immense calculating power, are creating stronger chess-playing computers that are affecting the way top human players approach the game. 

With this in mind, what are the most crucial elements of a good chess engine? 

Calculating Power

For pragmatic people, the purpose of a chess engine is to be good at chess. This means their main goal needs to be finding the best move in any position on the chessboard. As such, the most fundamental ability of a chess engine is calculation.

Due to the vast number of possible variations on a chessboard, engines require immense raw computational power. While a human player has to read opening books, engines can search through millions of positions every second. The key element to their playing strength is how effectively and quickly they can process information. 

That said, some engines focus more on evaluation. By quickly eliminating unsuccessful or unhelpful lines, they can streamline the calculation process. These engines have more limited search trees, but continue to find the best moves.

Creativity and Strategy

Traditional chess engines can calculate more accurately and deeply than any human alive. What they lacked was the creativity and nuance to adapt their strategies. Engines can still win against just about anyone through raw computational power. However, to be the best, today’s engines are using machine learning to create engines that can strategize. 

When chess engines play each other, the differences between wins and losses often hinge on subtle positional ideas. This means beyond computational power, engines today need to excel at advanced pattern recognition, strategic ideas, and understanding when to apply specific evaluative principles. This is where neural network engines are exceptional. 

While a traditional engine may rely on a principle like piece values, a neural network can draw on millions of games to create subtler and more flexible evaluative functions. So, for older engines, a rook may be worth five points, but for a new one, it is worth precisely what it is capable of in a specific position.

In practice, these engines can limit counter-play from their opponent extremely early. They are exceptional at keeping their pieces active and ruthless in exploiting any minor weaknesses. Overall, not only their calculations, but also their strategic principles are advancing beyond what humans can achieve. 

Today, one of the things that makes chess engines effective is their ability not only to be precise, but also to generate new ideas. 

Usefulness 

With chess programs advancing so far beyond the level of an average player, some question the point of advancing them. Others even question the point of playing chess.

While incredibly powerful, chess engines have not solved chess. Despite being a 1500-year-old game, there is no certain way for a computer to win every game. Improving chess engines is not only a technological challenge, but also a frontier that explores what is theoretically possible in the game.

As for the reason to continue playing chess, the game still holds the same interest and the same potential for creativity as it always has. For amateur players who worry that they’ll never beat the computer, it's worth remembering that they never would have beaten Bobby Fischer either. 

At the professional level, chess engines are expanding how players study and play. Professional players are now:

• Preparing much further into the openings, sometimes up to 30 moves!
• Playing more accurately in middlegames and endgames. 
• Exploring riskier variations inspired by subtle insights from engine games.
• Becoming more strategically complex.

While it is often said that engines are making the game less interesting, the truth seems to be that the highest level of the game is becoming more creative. As such, another way to define a good engine is the extent to which it can create new ideas and expand the game.