We lived to see it. Finally, it is here – new version of Hiarcs Chess Explorer Pro and updated author's chess engine. After many years of waiting, is it worth to give a chance to the new Hiarcs Chess Explorer Pro? What has changed in the program that the note Pro appeared ? Does the new version of the engine bring significant changes in the style and strength of the game? I will try to answer these and other questions in this review. Hiarcs Chess Explorer Pro (single-core version) was run on the Mac version using a MacBook Pro with an Apple M1 processor. First, let's take a look at some of the data in the table below. Table of Hiarcs Chess Explorer Pro specifications. Interface, Appearance, Access to Program Features. Hiarcs Chess Explorer Pro (HCEP) has a highly customizable and readable interface. The elements-windows and toolbar-can be active or inactive and moved relative to each other. If we so desire, the HCEP look can be minimalist... ...With several panels active... ...With a chessboard of various sizes... ...And colors... ...And in a light or dark theme (the dark theme is available in HCEP exclusively on macOS). Of course, there is also the option to change the theme of the chess board and pieces. However, I have my favorite set & dark theme, because I spend a lot of time working in front of a computer monitor and it makes my eyes less tired. You can also use several open windows and tabs, depending on the function or database you are using. Access to the most frequently used functions is provided by the toolbar at the top of the interface. Using the drop-down menu, you get use of all the other functions and capabilities HCEP. Some windows / panels have context icons that when clicked, expand their menus or open windows. The interface and all HCEP messages have been translated into several languages, which will certainly please users who prefer to use this program in their native language :-) To sum up this part of the review, my feelings are very positive. The interface is neat, clear, functions are easily accessible and dedicated windows or panels have useful context menus. One thing that may bother users with poorer eyesight is the size of the help icons located in the window bar. They are relatively small and certainly if there was an option to enlarge them, it would have a beneficial effect. I make no secret of the fact that I was very comfortable using the Hiarcs Chess Explorer Pro interface. Despite the many options, the program is easy to read, interface elements can be changed (position, size), access to program functions is intuitive and easy. Dark mode in the macOS version "does a great job" bringing relief to tired eyes and when working with less light. Power of play. The previous version of Hiarcs engine number 14 was released 11 years ago, in 2011. Even then, the strength of the Hiarcs game oscillated around 2900 Elo (depending on the hardware). Compared to the playing strength of the strongest grandmasters, that's about 100-200 more Elo. The consistent development of this engine resulted in several World Champion titles, including Hiarcs becoming World Computer Software Champion in 2013 ! It is worth noting that Hiarcs on slow hardware played at a relatively high level compared to other engines, which even on old computer processors allowed this engine to top the ranking lists. So let's check how the latest Hiarcs 15.0 performed against two class opponents and in a tournament with the strongest chess engines. First, Deep Shredder 13 engine, whose playing strength is some reference / comparison to other engines. DS13 is an engine with basically no major weaknesses, playing with strength in excess of 3100 Elo on modern computer. Each engine had 1 CPU at its service, game time 1 minute per game. Banksia GUI, Perfect 2021 opening book. Hiarcs won that match, winning 25 more games than DS13. A confident victory for Hiarcs 15.0 and a show of strength. 74 Elo points difference in this match between these two engines. Next, result from match consisting of 100 games Hiarcs 15.0 against the Komodo 13.2.5 engine using the MCTS algorithm. Each engine had 1 CPU at its service and 1 minute per game; Hiarcs Chess Explorer Pro GUI, internal opening book. And two images showing what a match between two engines might look like. A very elegant and minimalist look. Result: 28 Elo points difference, but as many as 10 wins more Hiarcs. That's a pretty convincing win for Hiarcs in this match. And a little more challenge for Hiarcs 15.0 and more computer opponents. There were 15 engines in the tournament. Each engine had 1 CPU at its service, game time 1 minute per game +1 second extra for each move. Banksia GUI, Perfect 2021 opening book. Hiarcs 15.0 took an excellent fifth place in competition with the strongest chess engines. It obtained a ranking of 3225 Elo, 225 Elo higher than its previous version (Hiarcs 14 WCSC). Hiarcs 15.0 playing strength in games with longer thinking time also comes in at just over 3200 Elo for the version using 1 CPU during play. Source: CCRL We can see that the latest Hiarcs 15.0 plays with a strength of about 3200 Elo, which is the same as the information from the website of the manufacturer of this chess engine. It is gratifying that Hiarcs 15.0 plays with so much playing power also having a short time to think, because during analysis or chess training his suggestions will be at a very high level. Style of play. Hiarcs has "always" had a reputation as an engine that plays chess in a style that is similar to the human style of play. Has the increase in playing strength by over 200 Elo points caused a change in playing style ? Let's find out through various examples. The most natural way to recognize Hiarcs 15.0's style of play is to play games against it. Which I did :-) The attachment contains 31 games played in ranked mode with adaptive strengths enabled. At first I planned to play with half an hour, after 30 minutes I still wanted more. In the end I played 31 games which took over 3 hours. In the first dozen or so games, I tried different techniques, plays and behaviors and predefined game styles (solid, aggressive, active) to see how Hiarcs 15.0 would do. From about the 20th game on, I started playing more and more seriously. Hiarcs plays fantastic chess with a very natural - human style of play. In my opinion his game is even more like a human opponent than the previous 14 version of this engine. Of course this is my subjective opinion and not everyone may have similar feelings. I remember that at one time playing in the same mode with the Shredder 13 engine, I had the impression even playing against an engine with a strength of about 1600 Elo, that I was "banging my head against the wall"; to each game I had to approach with maximum focus. Playing with Hiarcs 15.0 I had the impression that, depending on the style of play I chose, I was playing with my friends at the chess club and could play relaxed. Even if I make a mistake, my opponent can also make a mistake or play inaccurately. Overall, I have a very nice feeling after playing with Hiarcs 15.0. I'll talk more about this game mode in the section describing the training possibilities. It's one thing to play with Hiarcs 15.0 in ranked mode, but how will the engine do when analyzing games played by people - very high level games ? I have chosen games from games played by humans, with difficult positions for chess engines. My goal is to test the Hiarcs 15.0 engine on 10 positions giving it a maximum of three minutes of thought per move. I have assumed that these will be positions played by people who have allotted up to a few minutes for a move. Therefore, I did not take into account positions created in correspondence games. We start with a relatively simple position for a man to solve - the only effective option is to break the black pawn barrier by sacrificing the Queen. Hiarcs 15.0 found the correct move after 68 seconds. In the following final position, the elegant c5 wins! This time Hiarcs found the best move after only 2 seconds. And another ending, this time much more complex. It took Hiarcs 2 minutes and 37 seconds to find the winning Rook move. The following position is from a match between then World Chess Champion Garry Kasparov and the computer Deep Blue. G.Kasparov surrendered the game in this position, although the draw was led by Qe3!! The correct continuation was noticed by Hiarcs almost immediately, Another position involving titled players. A tactical Queen play to the d3 square Hiarcs finds after a few seconds of thinking time. In the following position, on the other hand, the winning move is Knight's "quiet and inconspicuous" move to field c7. For many years this was an unsolvable position for many chess engines in a reasonable amount of thinking time. Hiarcs took 80 seconds to find the correct move. In the following position many chess players are able to point out the strongest move Qf6+ although it is harder with justification :-). Hiarcs managed it in just 36 seconds. One of the most difficult positions for chess engines. To successfully launch an attack on the black King, white must clear the lines for Rooks by maneuvering with the King. Hiarcs 15.0 perfectly "feels" this position and almost immediately proposes the strongest move Kf2!! In this position, what can white do with one less piece and a black king in the middle? Of course, destroy the black King's position by hitting the Rook on e7 ! The chess player notices that after removing the black pawn - the defender of the enemy King - another white Rook, Bishop and Queen joins the attack. The attack of these three strong figures guarantees at least a draw. For a human this maneuver is relatively easy to see, for a chess engine giving up one more piece (Rook) is not so obvious. In this chess game 15-year-old Carlsen did not play well and the game was won by Svidler. In this position there is theoretically a draw. Hiarcs 15.0 saw the best move after 74 seconds. And finally, a position in which GM Anish Giri played a beautiful move with a pawn on a4. This is an extremely difficult position for chess engines, as the solution requires sacrificing the pawn and then exchanging Rooks. This seemingly contradicts the chess rule of avoiding sacrifices and exchanging pieces - if the opponent has a material advantage. In this position, the sacrifice of a pawn and exchange of Rooks by white is justified because the binding of the black Knight by the white Bishop is inevitable without the loss of that Knight - leading to black's defeat. Hiarcs indicated the best move after 15 seconds and the correct continuation after another 10 seconds. To conclude this section on the Hiarcs 15 engine, I find that the play style of this engine has not changed compared to previous versions. Hiarcs continues to play chess in a "human" style by evaluating positions in a way that is understandable to me and not artificially inflating or deflating its evaluation. As a reminder, Hiarcs 15.0 was tested in a single-core version using only one CPU at most. Surely the Deep Hiarcs 15.0 version run on e.g. 4 cores would provide solutions to the above chess positions in less time. But... do you really need a solution 5 second, 15 or xx... seconds faster? If not, then it's worth giving the single-core version a chance; you can have an excellent computer chess training and playing partner, whose moves and analysis will be in a very "human style". Training with Hiarcs Chess Explorer Pro The training capabilities of HCEP have not changed much from the previous version of this program. I described these possibilities in the article "Tools in a chess player's workshop - Windows - Part 1". In this part of the review, I will mainly focus on the changes in this version of the program. What may be important for many chess players, HCEP allows you to: play with engine strength adjusted to the human opponent (Match player strength), manually set a rating calculated in Elo points from 750 to 3300 (Custom), play with Hiarcs chess engine on maximum strength. One new feature is the ability to select a specific opening book to be used during play, including the popular CTG format books used by ChessBase's programs. Another added option is the ability to choose to play chess in the Chess960 (Fischer random chess) variant format. A nice option is to play a training game from the current position, e.g. from the chess game you are currently analyzing (Starting position: Current). You can also select a specific chess opening to train. The selection of openings is large; below is just a section that makes up the larger whole. When you play with the computer coach turned on, it continuously comments on the chessboard situation. The student knows if he is playing well, if his position has deteriorated, and if he makes a mistake the HCEP will indicate the reason and propose a move back. Of course, we can ask for a hint at any time during the game. In order to show the work of the computer coach, I have collected below in pictures the different kinds of messages it gives during the game. During training / practice, we can always turn on a window where we can see how our opponent (the Hiarcs engine) "understands" and analyzes the position -> It is very interesting that we will then see a line with analysis on the ranking level according to which Hiarcs plays -> if for example the engine plays at 2180 Elo then the analysis line shown will be at exactly that level of play. What does this mean? This analysis is not done by the engine at its highest level of play (like Kibitzer in ChessBase programs). This is a great example of how the Hiarcs chess engine "thinks" and plays in a very human-like way. Some examples. We can track our performance in games in a variety of ways. The statistical data is presented clearly; you can click to e.g. open a selected chess game and add your annotations or analyze it. Multi-Tasking & Automated Games Analysis Hiarcs Chess Explorer Pro is multi-tasking program, which is extremely useful. While you are playing a game, training, working with a database, a match between engines, or performing any other activity, Hiarcs Chess Explorer Pro can perform all other tasks in the background. Below a ranked practice game and in the background an analysis of one of the other games. In this example, I marked 3 games for analysis and ran the process. During the analysis of the 3 games by HCEP, I worked using the Mega database 2022. We can analyze games using HCEP according to many different criteria (see image above), e.g. in the mode of pointing out all errors or pointing out all inaccuracies and errors, etc. Once the analysis is complete, HCEP adds its annotations along with a summary. Annotations are readable whether you use dark or light mode. To make the notation of the analyzed game even more attractive and clearer, you might want to have HCEP place diagrams at key/critical points in the notation. Working with chess databases In comparison with the previous version, the biggest and probably the most important change in the newest Hiarcs Chess Explorer Pro 1.0 is the implementation in this program functionalities found in software dedicated to handling chess databases such as Scid vs. PC, Chess Assistant or ChessBase. This is a very brave step by the creators of HCEP, because in doing so they have challenged the above mentioned programs that for years have been gaining and establishing their position in a very specialized area of the computer chess field. Can the features that HCEP offers be satisfactory or even exceed the expectations of users using competitive software ? Let's check it out. In addition to supporting databases in the popular .PGN format, HCEP introduces in the Pro version support for .EPD and the widespread .CBH, .CBV and .CBZ formats that are used in ChessBase programs and chess databases. In addition, a database format with the extension .HCE has been implemented, which is the new default format in HCEP. The first time you open a database with a supported extension other than .HCE, HCEP suggests converting it to the default .HCE format. The conversion is done only once and is instant or very fast for databases of up to several hundred thousand chess games. When converting very large databases such as Meg Database 2022 of over 9 million games, the conversion on my MacBook Pro M1 took 2 minutes 36 seconds. Once the conversion is complete, we get the chess database in the default format of Hiarcs Chess Explorer Pro. According to my tests, in HCEP, working with .HCE databases is the fastest compared to working time with databases in other formats. Access to previously opened databases is from the Databases window, where we can also search for databases of interest. Here's how Hiarcs Chess Explorer Pro might look with several chess databases and windows open. Analogically to other windows, the database window is operated in the same way - using icons, shortcuts and functions in a dedicated Menu: Database. As expected, all operations related to copying, pasting, working with the Opening Tree, searching for games by header data, etc. work very well and as they should in a chess database program. Creating databases, deleting them, deleting duplicate games and those to be deleted, filtering, using simultaneously the chess engine and or the online/offline book of openings, automatic game analysis - it all works perfectly. What literally struck me is that Hiarcs Chess Explorer Pro even when working with a large amount of data in a database of millions of games - works... surprisingly fast compared to other programs for chess databases. For example, by the following criteria (Header search): HCEP searched for chess games in a database of over 9 million games in just 10 seconds ! HCEP has new database function called Find position, with which we can search for more or less complex positions using various criteria. And this is one of the features that is usually found in chess database programs. Like the aforementioned features, this one also performs admirably and surprisingly fast. A certain shortcoming is that these search criteria are simply not many compared to SCiD vs. PC, Chess Assistant or ChessBase. At this point, I'll give one of the examples I've graced in the courses area. I am interested in chess games that contain a motif of white attacking the black King's position with the help of the white pawns g4 and h4, the white rook on h1 and after castling made. On the black side with the King defended by Bishop on g7 and supported by Rook on f8 and pawn on h5... ...and meet the following conditions: Players with any rating Elo. Sicilian Defense: dragon A chess game takes a minimum of 15 moves, a maximum of 29. That is, we are interested in a game with not too many moves. Results: 0-1 and 1/2-1/2 Other criteria such as the number of total pieces, I omitted because HCEP does not yet offer such a search criterion. We can enter the first four points as search criteria in the Search header function. Then, when we go to Find position, whatever we select in the Find: filter, we do not get the required search criteria. The only things I was able to search for were games where the material according to the diagram was found - but not the specific positioning of the figures and pawns. But that's not what I was looking for! It should be noted that games with such criteria (and even more complex ones) exist in the Mega Database 2022. Below is one of many examples from that database -> chess game that should be returned as a result of the search. In the practice of a correspondence chess player, a professional chess player or a chess publicist - this may not be enough to find specific positions, maneuvers, motives, games in the several million chess games databases. Undoubtedly, the Find position function represents a large and as yet unexploited development potential for HCEP developers. Looking at all the database functions perfectly implemented in HCEP as a chain of functions making up the whole of possibilities, Find Postion is, in my opinion, a unit that should be further strengthened through its further dynamic development. Player Explorer In chess database programs, of which Hiarcs Chess Explorer is one, a key feature is the Player explorer (in other programs called Player Report, Dossier, etc.). With this function a chess player can prepare for e.g. a game against another player by taking into account detailed information obtained from games of his future opponent. HCEP shows almost instantaneously in the Player Explorer window the statistical data of the selected chess player, in this case Mr. Spanton Timothy. When you click on the opening, HCEP shows a window with the related games. HCEP also allows you to view a player's ranking over time via a graph. In a separate window we can also browse the games of a particular chess player using filters (rating, result, date). You can also see the aggregate statistics. And that's pretty much it. Player Explorer is another function, as well as functions for handling databases, which are much less numerous and have much fewer possibilities than in other previously mentioned programs for chess databases (Scid vs. PC, Chess Assistant, ChessBase). What does this mean in practice? In short, this means that if, for example, a chess player wants to prepare for a tournament or a match, with the help of HCEP he will be able to do so in a ... little above basic. For example I would like to: review my future opponent's chess games against players with an average 1900-2200 Elo rating in the last 10 years. see in these games the most common openings chosen by my opponent along with the moves from 1 to 10 after which he lost his games. Unlike the previously mentioned competing programs, using HCEP in the reviewed version 1.0 for this type of task, it would take a lot of time and manual labor to achieve results. Databases provided with Hiarcs Chess Explorer Pro Depending on the version purchased (Single-Core, Multi-Core, Premium), the Developers provide access to the following databases: Top quality database with 350,000+ Master games, IM/GM annotated games, tactical exercise Online access to up over to 1,000 GB of endgame tablebases and opening book database Local book with latest GM Theory and access to much larger online books In my opinion, the noteworthy chess games with added annotations are of a high standard. Below is one of those annotated games. Summary Hiarcs Chess Explorer Pro on the creators website is described as, among other things: the ultimate chess database, highly advanced chess database, world class chess database features. Source: Hiarcs Chess Software website (2022-03-26) Thus, HCEP is positioned by its creators as a program for chess databases - competitor to e.g. Scid vs. PC, Chess Assistant, ChessBase. So, are the key capabilities offered by HCEP in the reviewed version 1.0 in terms of chess databases at the same or higher level offered by competing programs ? The answer is no. HCEP's key chess database capabilities like search, preparation and reports are at a level below what competitors offer - as I demonstrated with examples in the review. Undoubtedly, the speed of all HCEP functions should be emphasized, including searching in a database of millions of chess games. HCEP is very easy and comfortable to use, the interface is eye-catching, it is highly customizable. Hiarcs engine is a gem of the program; it supports training tools and Fisher chess mode offering very good quality of analysis and unique - human - style of playing chess. The multitasking of the program has a positive effect on efficiency and can be a great added value for the user who can simultaneously e.g. work with several databases, analyze chess games and work on the repertoire of openings. Unfortunately, while working with the databases, Hiarcs Chess Explorer Pro stopped responding several times or just shut down by itself. These kinds of problems occurred only when working with chess database functions. For whom is Hiarcs Chess Explorer Pro ? Certainly for someone who needs a bit more than basic functionality for chess databases and does not need the very advanced features offered by the free Scid vs. PC or commercial Chess Assistant or ChessBase. Particularly noteworthy are the training features, which the HCEP offers at an above-average level and can certainly be useful for chess player who wants to develop their chess knowledge or enjoys playing with a friendly computer partner. HCEP allows matches between chess engines but it is a pity that it does not offer the possibility to play chess engine tournaments. Certainly fans of inter-engine matches would welcome the option to use engines that communicate through a Winboard/Xboard interface - in the next version HCEP, because it is not currently implemented. The release of Hiarcs Chess Explorer Pro for Mac is certainly very good news for Apple computer users. Undoubtedly, HCEP has strengthened its position as one of the most prominent programs in the macOS chess software segment. Hopefully, future versions / updates will introduce native support for Apple M1 processors that power modern and very fast Macs. Hiarcs Chess Explorer Pro is a program... uneven. Next to features refined to the smallest detail, we have functions operating at somewhat disappointing level - certainly not at the Pro level. Hiarcs Chess Explorer Pro version 1.0 is a program that offers a wide range of possibilities, but still requires fine-tuning (eliminate program crashes) and development of database functions. The creators behind Hiarcs Chess Software have accustomed their customers to regular updates and development Hiarcs software. I will certainly be watching the further fate of Hiarcs Chess Explorer Pro in the future, hoping for its further development, development at the Pro level. Hiarcs Chess Explorer Pro 1.0 score

I am pleased to present you the first chess engine tournament for Mac. I invite you to watch the video on the YouTube platform. I hope you enjoy your time watching the chess engines struggle in this tournament :) Below are games to download in .PGN format and... ...a table with results. Summary of this completed tournament The winner of the tournament is probably no surprise: Stockfish version 15 - an engine that premiered a saw days before this tournament was played. Congratulations to the whole Stockfish Team ! Stockfish, colloquially speaking, smashes almost all their opponents, not giving them much chance to win or even to draw. 16 points scored out of 17 possible is a very good result. Over 200 Elo difference between him and the 4th engine in the table is a result showing how incredibly strong Stockfish is today in fast chess games (1 minute per game without extra time). Second place went to the Dragon engine version 2.6.1 by the Komodo Chess Team. A very big surprise is the great result achieved by Dragon getting the same number of points as Stockfish. The achieved result shows that this newest version of Dragon, although older than Stockfish by over 4 months, is a super powerful engine with a phenomenally fast compilation. Big applause for Komodo Chess Team :) Third place went to Fire, who has been at the top of the chess engines for years and in this tournament showed that he really needs to be reckoned with. A word of appreciation to Norman Schmidt – author of the chess engine Fire ! It is worth noting the good result of Hiarcs 15 engine, which at the end of 2021 (after many years of waiting by fans) lived to see the next version. Hiarcs 15 does not use NNUE (Neural Network architecture) which shows that a good engine with a traditional position evaluation function still can be a challenging opponent.

How will the latest version of the Hiarcs 15 chess engine perform in a match against its predecessor ? I invite you to watch the video on the YouTube platform. Below are games to download in .PGN format and... ...a table with results. Of course, due to the small number of games played (30), this match and result is only indicative for a possible estimate of the strength of Hiarcs 15 game.

30.03.2022 - An unexpected and free update for the Fritz 18 chess engine A nice surprise from ChessBase, which has offered Fritz 18 users a version of the engine of the same name with the addition of Neuronal. I invite you to watch the video on the YouTube platform. The Fritz 18 Neuronal, according to the manufacturer's website, is expected to be about 120 Elo points more powerful than its previous version. So let's check if indeed the update of this chess engine has led to such an increase in the strength of play. Of course, due to the small number of games played (30), this match and result is only indicative for a possible estimate of the strength of Fritz 18 Neuronal's game. Below are games to download in .PGN format and... ...a table with results.

Here's how the title chess engine fared. I invite you to watch the video on the YouTube platform. Below are games to download in .PGN format and... ...a table with results.

According to the ranking lists Hiarcs 15 using 1 CPU is quite a strong chess engine. Let's see how it performs playing in a tournament with nine good opponents. I invite you to watch the video on the YouTube platform. Hiarcs 15 took third - an excellent place. It is worth noting that it is one of the few contemporary engines that do not use NNUE (Efficiently Updatable Neural Networks), which probably preserved its unique "human" style of playing chess. Below are games to download in .PGN format and... ...a table with results.

The latest version of the Orion 0.9 chess engine boasts a rating of 3121 according to CCRL. "More than one year since the last release already: time flies ! I have been very busy these last months, without a lot of time to dedicate to Orion, but the next version is here (it is actually ready since... last November !), with: support for SMP a 24x smaller neural network speed improvement and memory consumption reduction correction of a small bug in 3-pieces endgame evaluation Support for SMP I'm really happy to announce the support for SMP : Orion will now be able to think using several CPUs/threads in parallel, hopefully resulting in a stronger play ;-) This required a lot, what am I saying, a ** LOT ** of work: I had to redesign the main parts of the engine, to ensure thread-safe execution, split, refactor, simplify, rearrange the code to avoid problems when computing in parallel. On the contrary, I was surprised by the simplicity of the Lazy SMP approach, that's brilliant ! Smaller network architecture The other big change is the architecture of the neural network: it is now much simpler than the previous one, for a more or less equivalent strength (~20-30 elo weaker in my own tests). I replaced the 40960x2 weights in the first layer by a simple 768x2 scheme (6 types of piece x 2 colors x 64 squares = 768). This probably hampers accuracy in some complex positions, but globally speeds up evaluation as you don't have to recompute all the first layer when the king is moving (this is really helpful in endgame positions where kings have more mobility). This choice resulted in a 24x smaller network (421 kB vs 10 MB)...! I'm really happy with the result: It seems possible to compress chess knowledge a lot ! Important note about originality I know that some people are looking for originality: do not forget that engine creation can remain for some of us (including me) a hobby and/or a way to learn programming and A.I. ! This has always been my goal: develop a 100% original engine, not only in terms of playing style (that's not the case for the moment) but also in terms of code: Orion is not derived from any other engine, I wrote 100% of the lines of code, in my own way, always after having taken the time to understand what I was doing (the most recent example being the NNUE experiments I led in 2020). For example, Orion is based on a "only-legal" move generator, using flags embedded in each move representation to help sorting and pruning moves during search. Its transposition table also uses the number of pieces on the board as a criterium to replace old-entries. But then comes the issue of the data used to train the neural network with the NNUE approach. As for the 0.8 version, the provided network has been trained on positions that were statically evaluated with the nn-82215d0fd0df.nnue network which is the one embedded in StockFish 12. The StockFish engine itself was not used at all in that process: I took the network, reused the code I developed for my NNUE experiments to read the weights and evaluate a bunch of positions that I collected from CCRL games, and then trained my own network with my own (shared) Cerebrum library (note: this time, I was able to use only 128 million of positions, compared to the 360 million used for 0.8). Finally, from this perspective, I think one cannot consider that Orion is - at this stage - a 100% original work, as it uses knowledge coming from another engine. Please note that starting from v0.4, it has actually been the case: I was previously using StockFish 8 (static) evaluations to tune parameters of my handcrafted evaluation function. But, for sure, this remains the goal, and I already started to work on reaching that objective... The "zero" approach I think the most exciting challenge now that I know how to design and train neural networks is to find a way to train a network from zero, i.e. only using results of games (win / draw / loss). Inspired by an idea proposed by Connor McMonigle (Seer author), I tried to train one of such network, without success so far. The idea is to consider endgame positions (3-4-5-6 pieces), use the results provided by Syzygy tablebases, train a network on these positions, use the engine to evaluate 7-pieces positions with the trained network (after a depth 'd' search), re-train a new network on these labelled 3-to-7 pieces positions, and then restart all the process for 8 up to 32 pieces positions. The beauty of this approach is that the network is trained only using the endgame outcome, and shall learn how to "retropropagate" to middlegame positions the expected result. Next steps This is my current effort: try to improve the way to train such a "from zero" neural network, only relying on game results. That's a very difficult challenge ! Be patient ;-) " Woooooooow !! What an admirable passion in creating a chess engine :) I took the liberty of quoting the words of the author of the Orion engine Mr. David Carteau to fully reflect his thoughts. I am happy that there are people with such a chess passion and giving us - the chess community - a cool and powerful engine. I watch with admiration the development of the Orion engine. Below is the strength of the Orion game according to the published versions. Source: CCRL

Well, also the latest version of the Toga chess engine has received an NNUE rating enhancement. Source: chessprogramming.org It is worth mentioning that the Toga engine is derived from its famous prototype Fruit engine. A long way has passed Toga to get to the club of engines playing with the power of over 3000 Elo ! Here's what we find out from the Rebel website: "TOGA IV now with Rebel 14.2 NNUE evaluation. Toga IV 1.1 is about 45 elo stronger than Rebel 14.2 and above that offers you: Multiple threading up to 64 threads; MultiPV up to 10 moves. TOGA IV 1.1 earned a rating of 3323 on the Gambit Rating List. Offered to you in cooperation with Chris Whittington."

Seer - one of the few chess engines that plays with a strength of over 3400 Elo (!) has been released in the latest version 2.5.0. Source: chessprogramming.org Here is what engine author Mr. Connor McMonigle wrote about this release: "I've released a new version of Seer here: https://github.com/connormcmonigle/seer ... tag/v2.5.0. This release adds around 100 Elo in self play. This release adds roughly 100 Elo in self play. The majority of this strength improvement can be attributed to a variety of miscellaneous optimizations and an improved network trained by way of several further self play training iterations. Additionally, time management was revisited with this release which should notably improve performance at shorter time controls. As with the prior release, the embedded network was trained solely on data originating from Seer produced by way of several self play iterations starting with the initial "retrograde learning" network bundled with v2.0.0-v2.1.0 as a base." Compiled by Darius Compiled by Connor McMonigle, author Seer chess engine

A significant and free update Fritz chess engine, increasing strength by about 120 Elo points at shorter time controls. Source: ChessBase From the developer's website we can read, among other things: "The classic Fritz 18 engine has a manually optimised rating function and achieves a playing strength of about 3300 ELO. In the new Fritz 18 Neuronal engine, the rating function has been replaced by a fast neuronal network (NNUE), enabling play 120 ELO points better, (i.e., achieving about 3,400+ ELO). (...) The special feature of NNUEs is that they are relatively small, and can be efficiently evaluated incrementally during the search. NNUEs do not require a GPU - the instruction set extensions of modern CPUs (SSE2, AVX2 or AVX-512) are sufficient. The fastest version of Fritz 18 Neuronal, which uses the AVX-512 instruction set, achieves approximately the same number of nodes (NPS) as Fritz 18; the AVX2 version (all modern CPUs) is only 5% slower. A total of approx. 5 million games were played, and from these, 400 million rated positions were generated for training the net. The first net was trained based on evaluations from 200,000 self-play games of Fritz18, then in a further 80 training iterations, a total of about 300 intermediate versions of the net were tested and the best net was chosen step by step." Fritz 18 can be purchased from the company's ChessBase store.

Five participants. Though games between the first four. The winner was determined after the Armageddon formula struggle. Hearty congratulations to Mark Lefler, Larry Kaufman, Erdo Gunes and the whole team. To its collection of championship titles, Komodo Dragon adds another one from 2022. I always emphasize that I am very pleased with the good results of chess engines that are developed and regularly made available to a wide range of users. Among such chess engines is Komodo Dragon. Final standing. Conditions: Double round-robin tournament. Time control 60 minutes with 15 seconds per move. Hardware: Source: ICGA A very good result was achieved by Ginkgo (World Chess Software Champion 2022), who finished third, only half a point behind the winner. Noteworthy is the nice result of Shredder, which played on hardware using ARM processors. Let's hope that the new version, which we are looking forward to, will be even stronger. All tournament games (original notations). Source: ICGA The games of the World Chess Software Championship are played in Messe Wien, Press Centre, in Vienna, Austria from Tuesday 26 July, 2022 until Friday 29 July, 2022. Some photos from the game room. Source: ICGA Source: ICGA Source: ICGA Source: ICGA Source: ICGA

The latest version of the most powerful chess engine from Cuba! Here is what the author of the Nawito chess engine Mr. Ernesto Torres wrote: " Greetings friends. I announce the new version of the Nawito chess engine. In terms of strength, I think it is superior to the previous version. For more technical information check the homepage on GitHub Engine version: 22.07 Available versions: Windows/Linux/Android, 64 bits Communication protocol: UCI ELO estimated: 2690 License: GPL " Source: TalkChess Thanks to the author, Nawito is written in such a way that it is easy to compile this chess engine for all popular operating systems. Below to download Nawito source code and compilations for: Android Linux x64 & ARM Mac Intel & Apple Silicon Windows All compilations by Mr. Ernesto Torres except Linux ARM and Mac by Darius.