Links

Wednesday, March 1, 2017

Version 10.11: improving analysis of a game

All changes in the development version 11 have been included in the version 10, except for new engines, and the support of Linux. 

Analysis of a game:

  • Elo perfomance: it is a formule created by Michele Tumbarello that calculates the elo of each move and of the game. This formula is designed to do the more accurate results when using DeepFish as engine and a minimum time of 5 second per move.
  • Stability control: a tab has been added that after finishing the analysis of a position with the parameters of time and depth indicated, it continues analyzing until in the last depths the calculated move is the same and the difference in the evaluation does not exceed the number of centipawns indicated. 
  • Lost points in the analysis graph, now can be hidden (Giovanni di Maria)
  • Right button on the graphic, to save as a pgn (Giovanni di Maria)

Captures: added a new option "Material advantage". (Carlo)

Some internal improvings in transiberian (coded by Alfonso Solbes) 

  • improving detection of draws
  • little performance improvement checking human moves

More control of the colors: Options/Colors/General


In addition, thanks to Max Aloyau, Ed Smith, philmath, vga, jojoB, Giovanni di Maria, Patricio Cainzos and MR's feedback.


32 comments:

  1. Hello!

    Some suggestions for improving Lucas Chess even more:

    1. In playing Blindfold, there could be an option that the board position should lag n half-moves behind. That is, when n=1 the start position should not be updated by white's move until after black moved, then black's will be visible after white's next move, and so on. That's a very effective way to learn blindfold chess. (I know there is something a bit similar under "Resources for Zebras", but in actual play it's much more effective and enjoyable).

    2. The Scanner tool is amazing, but it's sometimes inconvenient to select manually the borders, especially when the board in the image is not exactly a straight square. So maybe it's not too difficult to add the possibility to auto-detect the borders within the selected area (and adjust for non-flat scans), even better would be for it to be able to read from image files (not just screen-captures). Once all these are implemented, there could be a batch processing mode where the scanner processes all image files in a directory and saves them to one fns file, presenting the positions one-by-one to the user for corrections. A simpler feature would be to mark all the squares where it's unsure (below a certain threshold), so that the user will quickly see all the necessary corrections. It seems that Chessboard Capture (https://sites.google.com/site/fredm/) does some of these.

    3. Not all engines are available under Lucas-Elo, I wonder why. In particular between Snake and Bikjump 1 there is a wide gap in strength that could have been filled by Acqua, Material, Steven, and other Irina personas.

    (continued)

    ReplyDelete
    Replies
    1. Hi, I´ll try to respond your questions:

      1. I use a very simple way in Blindfold, no more taht change the image of pieces. Your proposal need more proccess each time, not easy to implement. I take note.

      2. There is a recent option = Tools/Save positions to FNS/PGN to work with a lot of images. I like the idea to mark unsures (I´ll try it).

      3. Previous to publish version 11, I want to update Lucas-Elo, and include all of these engines.



      Delete
  2. 4. Speaking of engines, having 47 engines is wonderful, but it wouldn't hurt to have even more. Many of the available engines are very similar or of the same family, and many interesting engines are of yet unavailable. Playing against different engines is not only more interesting, it might also help your tactical thinking to face contrasting approaches to chess strategy and it better emulates the experience of playing with human opponents. Following is a sampling of some free UCI engines that are interesting or unique in some regard (the descriptions are not all mine):
    * Houdini (version 1.5 is free): Good combination of tactical abilities and positional understanding. Evaluates well closed positions and two of his main characteristics are long combinations involving piece sacrifices to recover the piece and even win a couple of pawns after 10 to 15 moves and a tremendous defensive capability, very good at twisting and turning in positions where it stands worse, being able to hold a draw in seemingly lost positions. Often comes up with the best defense. Strong in endgame and leading up to endgame. Very tactically solid. Sees narrow paths deeply. Best at figuring out surprising tactical points at blitz time controls.
    * Spike: Aesthetically enjoyable games. Configurable. Learning.
    * LambChop: Active playing style. Can be entertaining. Slow but accurate.
    * Zappa: Very solid and "feets on earth" playing style, very good "Insanity Check" engine for seeing if dubious ideas work or not. Unique evaluations, sometimes proving wrong all other engines. Great analysis strength. Super-stable and consistent in its style. Also noted for playing in a sharp style.
    * Chiron (early versions free): Will present a move that is just a bit different. One of the engines most like bobby fischer. Very balanced.
    * Thinker: Especially its Active personality. Few other engines can match its elegant play. Noted for playing in a sharp style, willing to go into losing positions if it deems them enough complex so humans don't have any chance.
    * Grapefruit: Very nice and active engine (based on Toga II). Also neuroGrape, using back propagation neural network (machine learning) for evaluation.
    * Gandalf: Most like bobby fischer. Highly selective search. Plays aggressively.
    * IvanHoe: One of the engines most like bobby fischer. Active fork of the Ippolit and RobboLito series of programs and base of a lot of further derivatives.

    (continued)

    ReplyDelete
    Replies
    1. I have an imaginary limit of 50, candidates are Cdrill, Houdini 15 and Spike 14.
      I like to include Zappa, and Thinker but it is not possible, one is commercial and other not UCI.

      Delete
    2. Why the limit? Most engines don't take up much space, and disk space is planty nowdays anyway.

      There is a early free version of Zappa.
      You Right about Thinker being non UCI, sorry, my mistake.

      Delete
    3. "There is a early free version of Zappa"
      Where can I find ?

      Delete
    4. Hi,

      There is a free Zappa version at author's web:
      http://www.acoz.net/zappa/zappa_11_all.zip

      Have you tried Wb2Uci for Winboard engines?
      https://sourceforge.net/p/craftychessinterface/wiki/wb2uci%20with%20multiple%20engines/

      I have been able to use Thinker :-)


      Delete
  3. * Arasan: Balanced style. Well configurable, tunable, learning.
    * Atlas: Good king safety.
    * Booot: Strong king safety. Balanced style.
    * Equinox: Strong king safety.
    * Fire: Strong king safety. Very configurable. Extremly well tested. Also Firenzina, based on earlier version of Fire with unique touch.
    * Fizbo: Very dangerous in middlegame.
    * Hakkapeliitta: Very dangerous in middlegame.
    * Laser: Very balanced.
    * Nemo: Often dangerous in middlegame.
    * Nemorino: Strong in endgame.
    * NirvanaChess: Very balanced. Strong in endgame.
    * Pedone: Balanced style.
    * Protector: Strong king safety.
    * Quazar: Playing style generally unusual. High move average.
    * Spark: Very dangerous in middlegame. Strong king safety. Data-mined evalutaion.
    * Wasp: Balanced and human style.
    * Vajolet2: Very balanced. Highly selective search.
    * Tao: Noted for playing in a sharp attacking style.
    * AnMon: Tactically very strong MTD(f) searcher. Very strong in blitz.
    * Bright: Complex humanlike evaluation.
    * Cyclone: Very configurable. Sophisticated.
    * Giraffe: Most successful attempt thus far at using end-to-end machine learning to play chess. Probability-based search. Deep Neural Network-Based Evaluation.
    * Homer: Interesting playing style and analysis. Learning.
    * Naum: Only engine to have a reasonable grasp of fortress positions. Nice style. Enjoyable to watch.
    * ProDeo: One of the most tunable chess engines. Distinctive. Much chess-knowledge. Deep positional understanding.
    * RomiChess: Unique learning approach.
    * neuroStock: Stockfish 2.01 with back propagation neural network (machine learning) for evaluation.
    * Yace: Highly configurable, tunable, learning. Enormous endgame knowledge and ability of backward analysis.
    * Weak engines: Most weak engines are unsuitable for adult beginners. They are either too silly, giving away pieces for free, or too strong, requiring tactics & exact calculations to win. Most weak engines are also buggy. The following are some good weak engines, stronger than Acqua but weaker than bikjump 1 - NSVChess, Inuyasha. Inuyasha is especially interesting being instantly quick, doing almost no calculation, but still stronger than Acqua. The following are stronger than bikjump 1 but weaker than Irina - APILchess, Violet, Killer Queen, VirutorChess, Nanook, Fusch, T.rex, Safrad. Other engines in this category include Pwned, Ace, Suff, EGM (Electronic Grand Master), Hedwig, JFresh, Minnow, Piranha, Powder, Pulse, Quokka, Revati, SCE, Skiull, Superpawn, Ziggurat, CPW-Engine, Chessputer.

    (continued)

    ReplyDelete
    Replies
    1. I like weaker engines, less than 2000, they can help a lot of. I take note of your proposals.

      Delete
  4. 5. In "Fics-Elo", "Fide-Elo", and "Play like a grandmaster", when the user makes a mistake (his move is much worse than the original player's), the position could be saved automatically for later review in "Find best move" or "Learn tactics by repetition". Also, when the game ends with resignation, optionally continue playing on the winning side against an engine untill checkmate, training ability to win in won positions.

    6. In "Play against an engine of your choice", there could be a button to show what the engine would have done if it were possible for the side to move to do nothing (the engine had an extra move), thus learning to recognize threats. To make it more difficult and instructive, the user may have to indicate the threat before being shown. (Both Fischer and Carlsen studied by "playing both sides").

    7. The "Personal Opening Guide" would be better based on positions instead of moves, handling transpositions.

    I hope it's not too much of a nuisance getting so many suggestions at once, but of course you are free to take your time and review them one-by-one at leisure...

    Thank you so much, and keep up the good work!

    ReplyDelete
    Replies
    1. Thank you very much, philmath, I need time to study your proposals.

      Delete
    2. 5. Good idea, I´ll try something.

      6. There are a kibitzer = Threats

      7. Now "Personal Opening Guide" is based on positions.

      Thank you very much

      Delete
    3. Sorry I didn't know about kibitzer = Threats.

      Thank you very much!

      Delete
  5. Hi.
    In tournament between engines, can games generated as Round-robin table?
    Example:
    3- 6
    5- 1
    4- 2
    6- 5
    2- 3
    1- 4
    2- 6
    4- 5
    3- 1
    6- 4
    1- 2
    5- 3
    1- 6
    3- 4
    2- 5

    and not:

    1-2
    1-3
    1-4
    1-5
    1-6
    etc....

    Thank you....
    Giovanni




    ReplyDelete
    Replies
    1. Giovanni, what about random?: first program calc all pairs, then randomize the order.

      Delete
    2. No, in the next update may be.

      Delete
    3. Thank you.
      It could be interesting also to choose. with a checkbox, if the players must play also the "return".
      Example:

      1-2
      2-1
      1-3
      3-1
      2-3
      3-2

      or:

      1-2
      1-3
      2-3

      Ciao

      Delete
    4. Thank you.
      In every case, here is the alghoritm to schedule Robin Round pairings:

      https://en.wikipedia.org/wiki/Round-robin_tournament

      Bye
      Giovanni

      Delete
  6. Hello,
    sorry for my bad english...

    When I make an analysis with an engine, the response change each time on same position.
    There is a "random" function in that analysis?

    Tested with Stockfish 8, Komodo 8, McBrain.

    Thanks.

    Fernando

    ReplyDelete
    Replies
    1. I believe that when do the evaluation of a position it adds a little random value, but I don´t know, only I know that the GUI doesn´t change the evaluations.

      Delete
  7. I activated the sounds but there is no sounds ...

    ReplyDelete
  8. Replies
    1. It is necessary to create: Options-Sounds-Custom sounds
      Then config: Options-Configuration-Sounds.

      I have asking google for permission to use the voices in Google Translator, I am waiting a response. If this is positive I will include english-voices.

      Delete
  9. Is there a hope to include new personalities based on Rodent chess engine?

    ReplyDelete
    Replies
    1. Previous to read your message, I have downloaded Rodent III, and I am thinking about to add internal personalities.

      Delete
    2. Hi Lucas,
      I have tried to use Rodent III as kibitzer but no moves appears in the table. The process is running but the printout is not working. Any idea? Have you seen the same? Is there anyway to start LucasChess with some kind of log to see what it is happening?

      I have added Rodent II as external engine and I can use personalities without any problem.

      Thanks

      Delete
    3. Rodent works with Best move, but not others that need MultiPV option.

      Delete
  10. Hi
    1- thanks for good lucas chess program.
    2-Where is ' Options>Change board size ' submenu in Lucas 10.11 portable version ?
    3- Suggestion change in lucas chess in below jpg file :
    http://uupload.ir/files/rp0l_lucas.jpg.

    ReplyDelete
    Replies
    1. 2. Right button over the border of any board.

      3. Problem is when lauching scanner, there isn´t a good position to the vertical black pieces.

      Delete