top of page
tituboorskime

Houdini Opening Book Ctg 18l: Unlock the Full Potential of the Houdini Chess Software



Hello there, the book has been updated to versuon 2.1, including only computer games from 2014 and grand masters rated at least 2500: the result is a lighter book (124 MB) with stronger variants, my intuition was correct.


Yesterday I've made a public challenge on Twitch again the latest Cerebellum_light, the once stronger opening book available, my book has won by 2 points. Here you can find the latest part of the games and all the games databases archives:




Houdini Opening Book Ctg 18l




This time it's a very leap of strength up, being +23 ELO points stronger than goi 2.3 and +51 ELO points stronger than the latest Medulla.A reduction of entries in spite of a greater quality of game lines turned this new book into another very strong one, goi 3 has successfully passed every test in order to be published.


The opening book is 143 MB sized and it was built by compiling only high-end computer games with the most powerful engines, the datas were selected using Chessbase 14 and compiled by Polyglot 64-bit. Goi.bin 3 is suited to be a book for chess engines competitions, but you can also install it on Droidfish on your mobile device.


Hello again, the book has just switched to version 3.2. The further days a CTG version will be prepared, in the meanwhile there's the BIN format, for those using competition engines like asmFish or Brainfish and/or not having Chessbase softwares. On the free Arena interface you can load the book in the Engine settings. You can also add the book to the Droidfish app as default opening book if you use an Android device.


Welcome to this new product, the elite of the chess opening books, its strength is unmatched. Built using only Stockfish 9, asmFish and CFish on very powerful high-end servers. If you want the best you must get this book. Much more powerful than Chessbase PowerBook or Hiarcs 14y.


After several thousands new lines from my server providers I was able to make a compilation worthing a new version.Goi 4.4 features variants took from opening books built by Komodo 11.3.1 and Houdini 6.03 games and a major extension on winning lines on the other hand.


With this combination, the free asmFish and my opening book goi 3, the program was stronger than Fritz 16 + Powerbook 2018 of more than 400 ELO points! (at the end of the video the games results are shown).


The second update of the 4,4 version is a killer: it outmatched any other top book, such as Cerebellum, Medulla or Vitamin kid. The Goi book, with previous hence weaker versions, has already won against other engines opening book like Chessbase Powerbook 2018, Chessbase Strongbook 2018, Hiarcs 14x and 14y.


This new release includes a new way to make the book: for the first time I didn't only make and select powerful games through very powerful servers, but I also use them in analysis to correct some wrong/weak line. The result is impressing, as you may see in the benchmark subsection.


I have to admit that wasn't an easy work both because of the strength of my opponents and for a strange instability of the engines in selecting moves in deep search: the most of the times I came back with the lines and they gave totally different moves, but after a while going forth and back I've always found the point in which the score was in my favour, so I was able to correct all the mistakes in the lines and I've even discovered that there were weak lines also in the Cerebellum book. Streethawk07 gave me the idea, sending back some wrong lines of the past versions. After having passed all the tests, Playchess 2500 ELO engines room score included, I've decided to release the book.


After the hard defeat on the last SCCT tournament, and seeing through it, I've decided to completely change the way in which I produce the Goi book.Cerebellum lines are by now obsolete so I've decided to get rid of them. The previous Goi book wasn't a merging of Cerebellum + minibook, it was the product of a dissected Cerebellum (BIN to PGN) then reassembled, reordered and selected in some criteria, hence not a mere Cerebellum clone as someone accused...


With the upcoming of Attacker Pro and the growth of Vitamin in strength plus the Playchess best players private books I've realized the Goi 5 technique wasn't enough anymore to be on the top again. Goi 5.1 CTG was an experiment, I wanted to see how a book full of lines performed... it performed very bad at the SCCT tournament despite my huge dedication on it; it was a good book for Playchess, though. Anyway, the new 1.c4 killing lines on Playchess didn't permit Goi 5.1 to survive anymore, so I've decided to restart everything from scratch again and create a totally new book, hence the new version 6.


All these years of bookmaking and testing brought me to realize that the game is decided since the very first moves, as the engines have almost the same strength. Or better, a good book can allow you to win or at worst to not be beaten. The idea of a light book with incredibly powerful lines came into my mind then.


Try to image my book like an onion: it's composed by a multitude of layers, first the top won games, then the top draw games, then my top players kibitzed games, then the engine tournament games > 2500, then my played games with the prototypes, then again kibitzed games, then again other engine tournament games > 2500, and so on. Create a new mini-book and merge it with the base prototype, then modify the lines where the errors lie. This for something like more than one hundred times. It was amazing but it required times, as you may have noticed. Many lines used by best players were improvable or even wrong, and this was inherited in my first beta version, so I had to manually correct them and test them myself both in books tournament and in the battlefield.


I had the idea to start a betatesting made not only by me, having few time, but anyone wanting to help me... Clearly I couldn't allow the book to leak out so easily (otherwise all my efforts would have been vane), so I decided to set up a reduced price for the betatesters to try it.


This testing thing didn't work, though, as people just "bought the book" without giving me almost any feedback or lost game, exception made for my partner Deiby L., which I thank so much. He sent me hundreds games played by his AMD ThreadRipper 2990WX from both Playchess gaming and private tournaments with other books.


This time the fun was great because by now I have fully grasped the mastery of chess bookmaking, so it was easy for me to correct any flaw I encountered and accentuate the lines in which my book was beating the other players/books.


Last but not least, since the stupid Fritz bug to play sometimes moves outside the book and making lose the game, I had to use an unconventional (and syntactically wrong) way to mark the moves on the lines I have analyzed (!green -> move to do, ?!red -> any other move to not do). For this reason Goi 6 is not a book for study, be CTG or BIN, this is just a book for competitions.


The fact that I forced the moves allowed me to vary the book parameters: variety of play is not anymore the minimum, but 70. In this way I got rid of infinitely drawing against Vitamin. Learning strength is different now too, 500 (the maximum), in the hope the book can learn from its evencutal losses.


I recommend you to NOT use Houdini, Komodo or Lc0 with this book as they are very weak compared to the Stockfish clones, unless you have a very, very powerful hardware and Lc0. The recommended engines for rapid games and average-high end CPUs are asmfish, Eman, Brainfish, Stockfish, Stockfish Polyglot and so on. I guess Goi 6 will have great result also with Cfish, CF_EXT, CorChess, McCain.


  • Shaping the analysis tree. Aquarium has a number of settings that have a significant influence on the shape of the IDeA analysis tree.The "Tree width" parameter determines how wide the resulting IDeA tree will be.

  • The maximum length of variations can be defined. This allows you to make the tree denser near the root and avoid very long variations.

  • Analysis can be limited to positions within a certain evaluation range. Positions outside the specified range will not be explored further.

  • The number of alternatives generated for White/Black can be limited.

  • A "preferred side" can be set; e.g., for developing your opening repertoire.

  • A list of "thematic moves" can be given. You can use them to ensure that certain moves are always analyzed. For instance, in an attacking game you might want to make sure that the moves Bxh7+ and Ng5 are considered.

  • "False alarm check" immediately performs additional analysis of new moves that look "too good to be true," preventing IDeA from spending more time on them if they turn out to be worse than they first looked.

2ff7e9595c


1 view0 comments

Recent Posts

See All

Comentarios


bottom of page