Deep Analysis Chess GUIs 2022

A comparison of the best chess GUIs for deep analysis. August 2022.

Aquarium 2022

Banksia

Nibbler

HIARCS Chess Explorer 2

Fritz 18
Chessbase 16

Chess Openings Wizard Professional

Chess Position Trainer

Chessify

Shredder Chess

17 Comments

  1. This is so valuable. I'm sure it has saved me hours and hours of research and I'm also sure I would never have the breadth of knowledge to delve into so many competing products. Now if only I could stop dropping pieces… (on the floor during street chess 🙂 )

  2. Something is broken with the picturequality! I get only 360p.

  3. Hello Carl, Thanks for the video but the subtitles cannot be activated to see them in Spanish.

  4. I am working on an Aquarium 2020 project and it does not show me the depth statistics, I do not know if it was some configuration that changed keys, can you help me please

  5. Regarding Chessify (the 25-100 MN/s version): I tried to reach out to them regarding the subscription (web-only) version shutting down after a few hours, regardless of whether or not I am at the machine. Thus, there is little benefit in e.g. running analysis overnight or at work. It have not received a reply after nearly a week, so it might be that they have no solution to propose at present (although it would have helped just to say that). Since your experience was the total opposite, I will not rule out that I made a (technical or communicative) mistake in the process.

    You of course also articulated the very positive aspects of the speed of the software, which I have also experienced (and most often one might just be too greedy when wanting deeper analysis than 50-100 GN for a single move).

    There are also more minor things: Such a powerful setup should have full 7 men Syzygy (tablebases are only available with the 300 MN/s version or better) and at least a monthly update of the latest Stockfish development version (assuming this works in terms of the cluster settings).

    The average speed also seems to be less than 25 MN/s, but I can live with that; it might be the issue is that newer Stockfish versions are slower due to bigger nets.

    The most clear-cut thing (which cannot be intentional on their part) is a bug in the variation control when choosing to promote a move, which breaks the interface, but one can of course just remember to avoid that, and keep all important lines locally.

  6. Banksia feature + Chessify power is the best, using the chessify plug in… Banksia have a interesting database sql, may be will be the future of chess database. U can create books.
    Sometimes with Nibbler (+ chessify plug in), but u need to do the tree manual. Focus the lines u want, thean Go and Lock, check all the line, return to lock position.. check again and nibbler will save the variation. The great thin with Nibbler is that u can use the API of lichess and chessdbn.

  7. I love that when Nibbler was showed, Carl asked the question will lc0 finds the Nxf7 move? the funny is that it is a top policy move recommendation meaning, that even only 1 node analysis lc0 would go for that move! the AB types of engines relies on search to find it! As GM Matthew Saddler mentiones in his recent book it playing vs Lc0 only 1 node evaluation is a great chess, practise, lc0 going to play the top policy recommendation directly from the neural network which is around the elo of 2400 level, so it is another good reason to download Nibbler!

  8. Another MOST excellent production from you, Charles! I deeply appreciate the effort you put into this important chess topic, as well as your willingness to freely share the fruits of your labors.

  9. THANK you very much for all your fantastic videos!!🙏🏻👍🏻 – They are all very helpful !! 👍🏻🍀✨

  10. you mentioned that you have multiple computers running on the same idea project. How do you set that up.

  11. Banksia certainly seems to have the potential to be a great deep analysis tool. However, without some comprehensive tutorials to guide the user through the maze, it is useless. I tried to follow the video within your video showing how to create a project, but there was no explanation of the various fields in the creation window. Pity, I would have given it a go if I could get started! I wouldn't mind paying for Aquarium if that were updated and wasn't so buggy.

  12. I am using SCID and Caissabase as my chessbase. Can you do review with these two applications sir?

  13. Hello Carl, I just saw this video. And there is a big issue when running chess engines on many cores that people do not know about. Chess engines are not good hardware scalers. Meaning the more cores you give a chess engine. The more NPS you will show, but the search becomes very inefficient. You also own TR CPU. So you can do this test yourself. I will use your video as my example. In the position at time stamp 47:38. It took Stockfish 15 28 seconds to reach depth 39 in the position running at 224699 million nodes per second. From the time stamp in the video. From when you started the position and then let it run. That is about 20x more NPS then my desktop running Stockfish 15 in the same position.

    Now look at the time to depth for my Stockfish 15 vs the server running at 224699 Mnps. You are gaining very little because of the search inefficiencies. Here is my time to depth on my TR CPU at 20x less NPS. 32 seconds to reach depth 39 vs 28 seconds for the 300 MN/s server. And you can see exactly why this happens and the search inefficiencies. When you compare total nodes searched to reach a certain depth of analysis. Here in my example on the 300 MN/S server. The total nodes searched to reach depth 39 is 6833.77 MN. vs 384 MN on my TR CPU, or about 18x more nodes needed to reach the same depth of search in the position for the 300 MN/S server.

    Do not waste you money buying server time from Chessify, Chessbase or anyplace else. All you are buying is NPS, but no real usable gains in analysis over your desktop.

    Enter New Game Line 0.0

    r1bqkb1r/pppnpppp/8/3nN3/3P4/8/PPP2PPP/RNBQKB1R w KQkq – 0 1

    Analysis by Stockfish 15:

    6.Nxf7 Kxf7 7.Qh5+ Ke6 8.c4 N5f6 9.d5+ Kd6 10.Qf7 Nc5 11.b4 Bf5 12.bxc5+ Kd7 13.Nc3 Qe8 14.Qxe8+ Rxe8 15.Be3 e6 16.0-0-0 exd5 17.Bd4 Kc8 18.cxd5 Rd8 19.Bc4 Nd7 20.Na4 b6 21.c6 Nf6 22.Bxf6 gxf6 23.Kb2 h5 24.Nc3 h4 25.Kb3 Kb8 26.f3 Bg6 27.Rhe1 Bf7 28.a4 Bd6 29.Ne4 Be5 30.f4 Bxf4 31.Nxf6 Bxh2 32.Nd7+ Ka8

    White is clearly better.

    +/- (1.36 –) Depth: 39/52 00:00:32 384MN, tb=391

    (, 25.11.2022)

  14. Brilliant work, Carl. This was all crystal clear and level-headed. You've opened my eyes to the Banksia project– I will be following that with great interest.

Leave a Reply

Your email address will not be published. Required fields are marked *