Chess engines

Onno

About Onno chess engine

Onno Chess engine was developed and created by Onno Garms. Onno Garms is German mathematician who was enthusiastic about chess in his youth. Onno was fascinated by Commodore Chessmate computer of his father, and at age of seven he learned to play chess on his father’s computer. After Garms get the doctors degree on university he again started his old project about chess engine. Onno was written in C++ and its first release was in May 2009, after 3 years of developing. Executables for 32 bit processor register and 64 bit processor register for Windows and Linux are well in communication with chess Graphic User Interface through the Universal Chess Interface.

Onno’s time progression

As we mentioned before the very first release was on May 2009, version 1.0.0, but Onno Garms started this project 3 years before. Next year, at the same month Onno Garms released second version of his chess engine version 1.2.71 with included multiprocessing in the code of the chess engine. On first view Onno chess engine was very successful but on second view point it did not achieved his developer expectations. First version of Onno found itself on the seventh place of the world ranked top list, few ELO point below the fourth place. In the fallowing months competitors get far away with their progress and Onno staid far behind them, so Onno Grams decide that he could not fallow that degree of progress and he stop development and distribution in May 2011.

Features of Onno chess engine

This chess engine use variation of fixed shift on Magic bitboards for determining attacks for figures on the board. It also uses alpha – beta evaluation of exchanging figures and they are tuned automatic by Genetic algorithm that Garms include in his code for Onno engine. Search functions in Onno engine includes iterative search together with Principal Variation Search and null move pruning and also with verification search. Garms in his code for Onno engine is also using technique called bad pruning which is similar to the one named razoring. This idea include reduction of the search along with reduction window and works like this: if the engine search at some depth found that one loses more than one pawn, the search for that move is stopped and do not search more depths for it. Onno also has technique called parallel search with use of Young Brothers Wait Concept following Rainer Feldmann’s Ph. D thesis with name “Game Tree Search on Massively Parallel Systems” published 1993 with use of virtual messaging.

Node kinds in Onno

Onno engine is expecting to use node type to do internal iterative deepening for PV – nodes and also for Cut-nodes. Garms says that root node is PV – node and the first inheritor of the PV – node is also a PV-node. Following inheritors are depending on scout search like CUT – nodes. In the code of Onno engine Principal variation search researching will be done by a PV – node. First node of bad pruning technique will be a CUT – node.

 Senpai

About Senpai

Senpai chess engine is created by Fabien Letouzey also famous for his other chess engine Fruit. Senpai was released on 17 of March, year 2014. This engine has Universal Chess Interface UCI and is created under the open – source license like open source chess engine. This engines is written in C++11 by the scratch and its released under the GNU General Public License version 3. Whole engine have only one file senpai_10.cpp and structured for namespaces and executables so could work on different platforms like Windows, Android, Linux, Mac OS and etc. Interesting fact about the name Senpai is that this name comes from Japanese ancient times when there exists strong and scaring school for martial arts, and the meaning of Senpai stands for mentorship, also the name of the high ranked students of martial arts, the ones that get to the black belt. This is great analogy when we talk about chess engine which can teach you very strong and great techniques about chess and its play.

About the engine

Senpai is based on bitboard engine and the maps are consecutive bits. This particular engine is using Blockers and Beyond techniques loop approach to decide between attacking sets for all figures except for the pawns. This engine is also equipped with parallel search with special one master thread and a big amount of helping threads. Fabien Letouzey took this concept from the Young Brothers Wait. Also Fabien was using Principal variation search based on alpha – beta algorithm along with transposition tables inside the aspiration windows. The Author of this engine also added few more pruning techniques like late move pruning and aggressive futility pruning, beside the basic techniques of pruning which Senpai has.

Senpai vs Top ranked

In 100 round gauntlet matches Senpai plays against the top ranked chess engines and it does not show well. Computer used in this matches was AMD Quad cored processor with 4 GB of random access memory and Windows 7 operating system, not much. Results were not pleasing for Senpai and Fabien, because Senapi could not break the top ranked list of the top ranked chess engines and did not do what Fabien want from it. Senapi take a score of 31.5 points in 300 played games or that is 10.5 percent average. Senpai take won only on 5 matches, 53 draws and he lost a big number of matches, 242 lost games. Senpai could not break the top 3 ranked players Stockfish, Houdini and Komodo.

Summary:

Although, this chess engine was created and developed from scratch and is unique with some basic things, Senpai need big improvements and corrections. In meantime Senpai stays to be used for various type of platforms. We think for Senpai like a baby in battle of the chess engines so it has to be more learnings and upgrades. Maybe one day in future Senpai will make boom on the chess engine stage but that is not now.

 Spark

About Spark

First intention of Spark chess program was to be played online. Creator of this chess engine is Armand Niculescu, and was developed for Media Division. Code for this chess program was written entirely in Java Action Script. Because of the intention to be played online, the engine was created in big platform variation, so could be played on Windows, Mac OS X, Linux and also could be played on tablets and phones which runs on iOS, Android and Tablet operating systems.

History of Spark Chess program

Very first version of this engine was released in year 2002 and his name was Flash-Chess. It was written also in Action Script, intended to be demo technology for Macromedia. The 3D view was added one year later when was released new version. But because of the limits of this programming language the chess engine has defects and missed chess rules, and the biggest defect was that the chess engine was incredibly slow. Because of the defects mentioned before there was a need to be rewritten from scratch, so they did this and the new version of the program which was released in 2007 was much faster than his ancestor. Because Adobe Systems already had branded the name Flash, the chess program and chess engine was rebranded in SparkChess. In 2009 this chess program was featured to be used on Chrome Operative system and one year later it was featured for iPads and Android tablets. When version 5 was released in 2011, it came with online multiplayer support and High Definition graphics and version 6 one year later came with improved speed, enhancements and more important it was integrated on Facebook. This year 2014 was released version 7 of the SparkChess and this version came with new interface which was based on feedback of players, also it was added a new intermediary level and much more

Spark engine

This engine is using alpha-beta pruning along with principal variations, also includes iterative depending and also mentioned aspiration window. Spark engine also use transposition tables and history heuristics. The engine has database with openings with 7000 moves. Limitations mentioned above also do not allow to be used bitboards and end-game tables.

SparkChess Features

SparkChess includes chess engine and graphical interface for the players. Can be played in two modes, single-player and multiplayer mode. Board is designed also in two different modes, it can be played as a 2D board (eagle perspective) or can be played as a 3D mode (human perspective). Every game is saving automatically but there are also slots for saving the game. The game can be also exported on file in PGN format. Spark also has implemented PGN player which release comments and variations of recursive annotations. Also you can import and export board configurations and you can edit the board, imports and exports are made to be in FEN format. Online multiplayer mode is built in Union Platform and the server for it was written in Java. There are also many other features that we will let you to find by yourself.

 Spike 1.4

About Spike

Spike is chess engine first time released in 2004. Creators of Spike chess engines are Ralf  Schäfer and Volker Böhm. Spike chess engine was developed from start but they also includes some of the ideas that were already tested on the two other engines developed by the same developers. The interest question is how did they came with the name Spike. At that time both creators loved same Tv show and one of the characters in that show, if you are questing was having the name Spike. When they told this story about the name they emphasis that Spike was very cool and interesting character. Despite the tv show they gave this name to the chess engine because they think its aggressive and strong name and will as they say spike other engines. Very first released version of Spike was ver.0.7, but this was not the first version created they start with ver.0.1 but they did not want to release bugged and unclear version of their chess engine so they decided to release version 0.7 as a first version of Spike chess engine.

History of Spike

The very first created version of Spike, version 0.1 had played on online tournament called Chesswar organized by Oliver Deville. Creators Ralf  Schäfer and Volker Böhm want to see how will their engine work on tournament. This version played at tournament only once and they upgrade it in version 0.2. With the next version 0.2 they decide to send to Leo Dijksman for WBEC Ridderkerk, because they want him to help them to found the bugs they made in the code and help them to correct the bugs. The fallowing version of the engine managed to win in his group in qualifications. Version 0.5 was more successful than his predecessors and won on WBEC in third division qualifying itself for the second division.

Spike 1.4

After four years without new version of their chess engine Ralf  Schäfer and Volker Böhm released new version Spike 1.4 in February 2011. In the pause of four years the creators did not stop with their work on Spike chess engine so when finally new version was released there were a big amount of upgrades and changes. Here are the new features that we could manage to write for you: firstly they simplified the evaluation system and then they had rewritten the evaluation of the passed pawn and had thrown away some of the doubtable values, they added to the engine new terms and values which will be used for pawns in rook endgames and also for the pawn endgames, they manage to sort the material table. They also improved utilities of pruning technique and add the more selectiveness for extensive usage of the lmr. With this version Spike is upgraded to support 12 processor cores and also they manage to fix big amount of bugs.

Summary:

Previous version of Spike engine maybe was not enough good but this one made big promises for further tournaments and further progress in upgrading.

Texel

About Texel

Texel is chess engine developed and created by Peter Osterlund with origin from Sweden. Osterlund released Texel engine as User Chess Interfaced compliant version with open source chess engine under the GNU General Public License. Osterlund firstly released his engine in March 2012 written almost everything in C++11 programming language and a part of his older engine CuckoChess. The engine got his name after a HMS Cuckoo, ship that was wrecked on the Haak coast of Texel on April 04, year 1810. The engine same like other engines made in C++ and uses magic bitboards, it is based on iterative deepening, aspiration windows and quiescence search with Static Exchange Evaluation pruning and with Most Valuable Victim - Least Valuable Aggressor moving order. Also the engine is designe with hash tables and history heuristics and recursive moves. Texel engine also includes futility pruning and late move reductions. Texel can be used on computers with Windows, Linux and Android base platforms.

Texel History

Texel 1.0 version has added evaluation term to do not enters in wrong corner of KRKB endings. Also this particular version has implemented new revers futility pruning and implemented late move pruning for the first time.

The next version of Texel, Texel 1.01 was released in March 2012 and came with more cache alternative and implemented new passed pawn race technique of evaluation and also technique called passed pawn as bonus for pawns which are ranked with 6 and 7.

Next version Texel 1.02 was released one year after his ancestor in June 2013. This version came with big amount of little changes and with added new evaluation features. Mentioned version was speeded up because of the changes in search and added one check in quiescence search. Texel 1.02 has line aligned entries on transposition table and cache optimization with magic bitboard constants. There can be found also reduced aspiration widows with widening fail – high and fail-low at the very root.

Texel 1.03 inheritor of Texel 1.02 was released in January 2014 with big improvement in ELO points, hundred ELO points above the previous version of Texel 1.02 and with use of only one core. Texel 1.03 version also includes specific and unique recursive parallel search technique with use of threads. Version 1.03 implies the power of technique called Dynamic Tree Splitting and the simplicity of ABDADA, also there can be found verification of the null move and high aggression LMR. This version precludes the use of Botvinnik-Markoff extension.

 The last and latest version is Texel 1.04 released in May 2014 shortly after Texel 1.03 and above the previous version 1.03 for 70 ELO points above it on developer’s computer with using only one core. This version almost all changes are made in improving the functions for evaluation.

Summary:

Peter Osterlund is doing pretty good job with his engine Texel. He did not stop with his creation and development and improvements in every single version that he released are better and better with each one of the engine.

 Thinker

About Thinker

Thinker chess engine was developed by Lance Perkins. Thinker collection of programs includes Windows User Interface named Thinker board, also includes Pocket PC, ICS client and a Book engine Book Thinker. Thinker is private engine and has his own Graphic User Interface. Since 2004 Thinker is participating in CCT Tournaments and since 2008 it participating in ACCA Americas Computer Chess Championship. Newest version of Thinker was developed by Kerwin Medina in 2009 and was named Thinker 5.x, in mean time first developer of Thinker was occupied with the Thinker’s chess collection to add into it Pocket Pc and also chess program targeted for Nintendo DS. In 2010 Thinker along with Medina played at WCSC 2010 and also that year in WCCC 2010 Thinker found itself on the third place and was named joint champion together with Rondo after disqualification of Rybka in June 2011. Also worth to mention is that Thinker won at ACCA 2010 and at ACCA 2011. Thinker was utilized with bitboards and was written in C++ programing language.

Features intergraded in older version

Firstly we will review older versions features. Thinker Chess engine version 5.4A has better and re-written code for move-ordering than his previous versions and also has material evaluation during endgame only for inert version. TinySeal version 0.4 developed for Linux is better than his ancestor with improved efficiency with better allocation of resources and better use of them which is very important featured. Pocket version 0.2.0026 has added settings for squared displays that its previous versions did not, also there was problem with promotion of the queen which was corrected in mentioned version and last thing there was also defect in evaluation of computer opponent so that was also corrected. Pocket 0.2.0027 ancestor of the newest version has rewritten code for the validation of board and fixed problems with seek/sought window, and for the end this version also include every single improvement which was made in desktop version 5.4A.

Thinker Chess Collection

Thinker standard chess engine version 5.4C has big upgrade on the code with which Thinker is doing parallel searching, also algorithm for evaluation of king safety was re-written and upgraded for the passive version. The passive version also has been corrected in handling with KBPKP endgames. Into the inert version was made smaller changes on the evaluation on few endgame positions. Newest collection also has BookThinker 2.4A with upgraded fix for the problem that could cause picking bad moves from the book and also was allowed to do very deep opening sequence. Collection also have Pocket PC ThinkerBoard version 0.2.0028 which takes upgrades from the desktop engine of Thinker version 5.4. Pocket PC ThinkerBoard also have added new engine called Danasah developed by Pedro Castro and have fixed and upgraded problem of castling state of the games which were resumed. Rating display for online gamers and chess players also get through fixing. TinySeal version 0.5 for Linux WINE was improved with faster load timing, so the game should load much faster than its previous version.

 Zappa Mexico II

About Zappa

Zappa chess engine was developed by Anthony Cozzie and is a Universal Chess Interfaced engine. Few of the oldest and earlier released versions of Zappa are free for use but not for changing, they are not open-source software. Latest version of Zappa named Zappa Mexico can be found at Shredder Computer Chess to be purchased. The name Zappa comes from the movie Austin Powers 2 in which in one scene Dr. Evil talks about how he split the moon on half and to each part gave a name, first part of the moon was called Moon Unit Alpha and the second one Moon Unit Zappa, so the name comes from this scene. Project Zappa has been ended so there will be no more tournaments for Zappa and will be no more new releases and upgrades. After all standard for tournaments for Zappa stands at 67 wins, 37 draw and only 11 losses and that is 74.3 percent. Zappa has 3 titles, one title from championships and 2 match wins. Also Zappa has won 10 000 dollars prize in money. If we do not calculate internet tournaments for fun Zappa has never been defeated if the opponent was not Rybka.

Zappa Mexico I and II

Zappa Mexico was developed for Windows and Linux operating systems with incredible big amount of processors and the numbers go up to 512 processor cores. But the implementation of parallel search was based on sharing memory, unfortunately. On the other side this gave it ability to the user to connect to computers with Ethernet cable and use their united power. This versions of Zappa also use multi Principal Variation analysis and Nalimov tablebases. His Universal Chess Interface can be used on almost every standard chess graphic user interface. Interesting fact in the name of the multiprocessing and multi computational power is that no computer has get fired up while Zappa was running, so this means that you are safe for this optimization. Zappa Mexico 2 was released by Anthony Cozzie on February 1, 2008. This version of Zappa Mexico has two variations one for 32 bit processors and one for 64 bit processors. The new release of this engine includes new feature contempt and this is parameter that values which contempt is higher. This new release is faster than its ancestor Zappa Mexico and also there was fixed few small bugs which were found in the evaluation system. Zappa Mexico II has his own auto detector for threads and processors. There are also options to make Zappa to print less information and to print principal variations and to hide fail highs.

Summary:

Zappa Mexico II maybe is not the best chest engine ever developed but still there is something worth in it. We have never seen before a chess engine with such power and support for processors like Zappa Mexico has, the support for 512 processor cores or 512 deferent processors is remarkable. If you have a chance for try, do not be shy, connect some computers, and do the test.

Reported for plagarism

From where?