Overview of NNUE Evaluation
The world of computer chess has been revolutionized with the introduction of NNUE (efficiently updatable neural network) evaluation in Stockfish, the strongest open-source chess engine. This new evaluation system has been described as a game-changer for Stockfish and has significantly improved its playing strength. In this article, we will explore what exactly NNUE evaluation is, discuss its benefits, and take a look at how it works.
What is NNUE Evaluation?
NNUE evaluation is a new way of evaluating chess positions in a chess engine, which has replaced the traditional evaluation function based on handcrafted terms. This new approach involves using a neural network to evaluate a position and has greatly improved Stockfish´s ability to understand and play chess.
The NNUE evaluation system was developed by Tord Romstad, the creator of Stockfish, after experimenting with various neural network architectures and training methods. This new evaluation system has been implemented in Stockfish since version 12, released in October 2020, and has quickly gained popularity among chess enthusiasts and professionals alike.
Benefits of NNUE Evaluation
The introduction of NNUE evaluation in Stockfish has brought several benefits, making it a game-changer for the chess engine. Let´s take a look at some of the major advantages of this new evaluation system.
1. Increased Playing Strength
The most significant benefit of NNUE evaluation is its impact on Stockfish´s playing strength. When compared to the previous evaluation function used in Stockfish, NNUE has shown a significant improvement in performance, making it the strongest chess engine available for free. In a computer chess competition held in early 2021, Stockfish with NNUE was able to beat Komodo, the previous strongest engine, at a 54% win rate.
2. Faster Evaluation
Another major advantage of NNUE is that it is much faster than the traditional evaluation function. This is because the neural network evaluation requires fewer operations compared to the handcrafted terms evaluation. As a result, Stockfish with NNUE can analyze positions at a quicker pace, making it more efficient and effective in gameplay.
3. Less CPU Usage
The use of NNUE evaluation in Stockfish has also reduced the engine´s CPU usage by almost half compared to the traditional evaluation function. This not only makes Stockfish more energy-efficient but also enables users to run it on weaker hardware without compromising on its strength. This benefit has made NNUE evaluation a popular choice among amateur players who do not have access to powerful hardware.
How Does NNUE Evaluation Work?
At its core, NNUE evaluation involves using a neural network to assign a value to each chess position. This value represents the strength of the position and is used by Stockfish to make strategic decisions. The neural network used in NNUE has been trained using a dataset consisting of millions of chess positions and their associated evaluations. The network learns to assign a value to a position by analyzing features such as piece placement, pawn structure, and king safety.
The process of using NNUE evaluation in Stockfish involves first loading the pre-trained neural network into memory and then passing the current position´s features to the network. The network then outputs an evaluation, which is then combined with the handcrafted terms to produce the final score. This process is repeated at each turn, and the network is updated after every move to improve its accuracy.
Conclusion
In conclusion, the introduction of NNUE evaluation in Stockfish has been a game-changer for computer chess. This new evaluation system has resulted in a significant increase in playing strength, faster evaluation, and less CPU usage for Stockfish. With the continuous development and improvement of neural network architectures and training methods, we can expect even more significant advancements in the field of computer chess in the future.
