Calvin is a superhuman chess engine written in Java.
It features a a classical alpha-beta search algorithm paired with an NNUE evaluation function.
Calvin can play standard chess, fischer random chess (FRC) and double fischer random chess (DFRC).
Calvin is rated roughly 3400 elo (~62nd place) on the Computer Chess Rating Lists leaderboards.
My aim with this project was to combine my passion (playing mediocre chess) with my profession (writing mediocre code). My secondary goal was to learn about chess programming. I have certainly learned a great deal, and I hope that my code is well-documented so that first-time readers can learn too. If you find some information is missing or poorly explained, don't hesitate to let me know!
Like most engines, Calvin does not implement its own user interface. Instead, it communicates using the UCI protocol, meaning it can either be used directly from the command line, or via any popular chess GUI, such as Arena Chess, Banksia, Cute Chess, or En Croissant.
Please refer to the Releases section for the latest binary downloads.
To run Calvin, you will need Java installed locally (minimum Java 17). Start up Calvin by executing the command:
java --add-modules jdk.incubator.vector -jar calvin-chess-engine-5.1.0.jar
From there, use the "help" option or refer to UCI documentation for further information on available commands.
The table below tracks the estimated strength of previous Calvin releases, both on the CCRL leaderboards and on Lichess.
| Version | Release date | Estimated | Lichess | CCRL Blitz | CCRL Rapid |
|---|---|---|---|---|---|
| 5.1.0 | 2025-01-16 | 3460 | - | - | - |
| 5.0.2 | 2024-11-16 | 3360 | - | - | - |
| 4.3.0 | 2024-10-05 | 3300 | - | 3332 | 3277 |
| 4.2.0 | 2024-09-19 | 3230 | - | - | 3224 |
| 4.1.0 | 2024-09-04 | 3150 | ~2850 | 3171 | 3161 |
| 4.0.0 | 2024-07-30 | 3000 | ~2700 | 3011 | 3029 |
| 3.4.0 | 2024-05-19 | 2500 | ~2580 | - | 2492 |
| 3.3.0 | 2024-05-10 | 2450 | ~2550 | 2453 | - |
| 3.2.0 | 2023-12-09 | 2250 | ~2400 | 2233 | - |
| 3.1.0 | 2023-12-05 | 2220 | ~2390 | - | - |
| 3.0.0 | 2023-12-02 | 2200 | ~2380 | - | - |
| 2.6.2 | 2023-11-12 | 2175 | ~2300 | 2173 | - |
Calvin features a pretty traditional chess engine architecture. The engine can broadly be split into three parts: Move Generation, Search, and Evaluation.
Every chess engine requires an internal board representation, in order to track the position of the pieces, the move history, and so on. From there, for any given chess position the engine needs to be able to generate legal moves for that position, to be used during exploration of the game tree during search. As with everything chess-engine-related, the faster the movegen the better!
The search algorithm is all about exploring the possible positions in the game tree, in the most efficient manner possible. To achieve this Calvin uses a classical alpha/beta negamax algorithm.
- Transposition table
- Principal Variation Search
- Iterative Deepening
- Quiescence Search
- Zobrist Hashing
- Aspiration windows
- Lazy SMP
- Null-Move Pruning
- Futility Pruning
- Reverse Futility Pruning
- Delta Pruning
- Late Move Reductions
- Check Extension
- History Pruning
- Razoring
- MVV with Capture History
- Killer Heuristic
- History Heuristic
- Continuation History (1-ply and 2-ply)
- Hard/soft time bounds
- Best move stability scaling
- Score stability scaling
- Node TM scaling
- Calvin communicates using the Universal Chess Interface (UCI) protocol.
- Pondering, where the engine thinks on the opponent's move. Can be disabled using the 'Ponder' UCI option.
- Hash size and number of search threads are also configurable via UCI.
- Calvin is connected to Lichess where he plays regularly in the engine pool: https://lichess.org/@/Calvin_Bot
For any given chess position, the engine needs a method of obtaining an estimate of how good the position is for the side to move. Chess engine evaluation mechanisms can be split into two camps: traditional Hand-Crafted Evaluation (HCE), and Efficiently Updatable Neural Networks (NNUE). Since version 4.0.0, Calvin has switched to a neural-net based eval.
The neural network was trained using the excellent bullet trainer on a dataset of 1.2 billion positions taken from the Leela dataset, that I re-scored using Calvin's own search and evaluation.
I am hugely grateful to the many people who have shared their knowledge and resources with me throughout Calvin's development. Without them, Calvin would be nowhere near the level it is at today. To name just a few people, places and things that I am thankful for:
- The Engine Programming and Stockfish Discord servers, both amazing resources for chess engine knowledge and full of patient, helpful people.
- Jamie Whiting, the author of bullet, an incredible tool for training NNUEs among other things, which I have used to train all of Calvin's neural networks.
- The Chess Programming Wiki and TalkChess forums, both a huge help in the early days of Calvin's development.
- Other engines - I have drawn inspiration from countless others' engines, including but not limited to: Chess Coding Adventure (whose Youtube video inspired me to write my own engine); Stockfish (the queen of all engines); Leorik (whose author keeps an excellent devlog on the TalkChess forum); Lynx; Simbelyne; Stormphrax; Viridithas, and many more.
If you would like to contribute to Calvin, or just talk about chess/chess programming, get in touch!
