Blogs
Man vs Machine - Part I, Initial Days of the Computer Bots
CPU used in the TRS-80 PC-1; Photo licensed through CC BY-SA 4.0, By Robert Baruch. at https://commons.wikimedia.org/w/index.php?curid=78034808

Man vs Machine - Part I, Initial Days of the Computer Bots

Avatar of pawn314159
pawn314159
Show your flair!
| 8

This is an Update that includes info and data on new options for the Practice Computer Levels. New Info in Yellow highlight.

The Bots Arrive ---    

Hints were given by staff in the summer of 2019 that new computer programs were in development, each having their own playing styles. This was exciting news, as all that was available up to that time was Stockfish and the various Komodo levels, with many of the levels playing the same basic sequence of moves in reply to particular openings. 

The new computer bots then began appearing in live play on Nov 1, 2019. One of my forum posts has details of the first days of bot activity. Dozens of bots showed up over a three day period on Nov. 1, 2, and 4, with members counting 52 separate bots. Another 8 bots came on-line Nov. 13, bringing the total to 60 new bots. Staff commented that would be the extent of development for now, but there would probably even more bots at some future date.

With so many bots being released, I thought it useful to play a bot a day, emulating the 30 Day Challenge of RandomJeff, In such a way, data and playing information can be accumulated on each bot.

As I began playing the bots, a number of things were immediately apparent. They were designed to be more human-like than other computer opponents. Each one has its own name, cartoon-like depiction, humorous sayings, nationality, and more importantly, individualized playing style and skill level. Some are very aggressive in the opening while others are more analytical and positional. A few are schooled on specific openings. Many are rank beginners, several others are good, solid players, and still others are very tough to beat, with playing strengths at CM and Master levels. A few were even assigned initial ratings at the GM level. 

After going off and on-line during their start-up, the bots as well as the Komodo levels are now being regularly offered for rated and non-rated play to members.  In March, 2020, options were also added to the practice computer levels to provide for various types of assistance. The Challenge option allows for non-rated play with up to 10 levels of difficulty, but with no assistance of any kind. The Friendly option allows take back in all 10 practice levels. The Assisted option provides for take back, hints, suggested moves, threats to the position, and possible lines, all with amazing visuals. In all three options, the game can be timed. The Shortlist (below) contains new info and data on levels 1-5 of the practice computers.

Completed Games with Lower Rated Bots ---

The goal was to play every day, if I could. If the bots weren't available, then I would fall back to the Komodo levels. Games with real opponents were also feasible, both live and finishing a daily, as I wanted to compare the bots with actual people. There were so many bots hitting the live play area that I set out to work through the easier ones first, then battle up to my playing level and possibly beyond. A few games of bots rated <1000 demonstrate the playing ability of the lower-rated programs.

Milica-BOT. The rapid game with this bot shows how positional play can take full advantage of the lower-rated bots. Milica made senseless moves with its pawns, knight, and bishop. I zeroed in on her king, then ended the game in 11 moves with a smothered mate.

Hans-BOT is one of the better bots among the programs having a rating of < 1000. At time of my rapid game, his rapid rating was 904. This bot played a standard opening, the closed Sicilian, but then varied by move 5. Within short order, he lost several pieces, and mate followed on move 26. 

Game Statistics ---

I have now played all computer bots and Komodo levels with rapid ratings of < 1000. In the first 21 days of November, I have played 33 rated games. So I am keeping up with the 30 Day Challenge concept. I have won all the games so far, but every opponent has been far under my rapid and blitz ratings. The wins have therefore been close to automatic against the beginner levels of bots and Komodo. The stats are as follows ---

  • 33 rated games in 21 days ---
    • 31 computer games, 2 human games ---
    • 17 rapid games – 15 bots, 1 Komodo 10 game; 1 open challenge w a person
    • 15 blitz games - 12 bots, 3 Komodo games (K 1-3).
    • 1 finish of a daily w a person

The Shortlist ---

While designed to be a convenient and abbreviated rendition of computer skills, with 60 bots now on-line, even this shortlist has grown in size. This reflects the immense scope of the computer bot project undertaken by the Chess.com staff.

The following is information collected on the lower-rated computer programs, both the bots and Komodo. I have played all the following programs at least once in rapid or blitz, sometime both. The list is sorted by recent playing strength for rapid play. On the bots I played in blitz, both recent blitz and rapid ratings are included. While the break-points are somewhat arbitrary, there are so many bots now active on the low end of the spectrum, that I am giving a break-point of <1000 points as the lower-rated categories. On the practice computer levels, sorting is by accuracy and best moves %, as play is non-rated.

 Juan-Bot ---

  • In the game I played, Juan had an off-beat opening, then quick loss of pieces.
  • His accuracy rate in my game was 8.1% and best move was at 18.8%; 2 mistakes, 2 blunders.
  • 557 rapid games played (as of 11-12-19). Win rate is 6.9%; recent rapid rating, 490.

 Komodo1 ---

  • Previously played two rated, rapid games, w different colors, winning both.
  • In first game 3-1-19, I took quick positional advantage, following by material gain, then mate in 31 moves. Its accuracy was 6.1%, best move 13.3%; 2 mistakes.
  • In 2nd game, steady positional advantages lead to mate on move 32. Komodo’s accuracy was low, at 8.4%, best move 18.8%; 5 mistakes, 1 blunder.
  • 230,898 rapid games (as of 11-13-19); 29.1% win rate; recent rapid rating, 494.
  • Recently played blitz game; accuracy 3.0%, best move 0.0%; 0 mistakes; 3 blunders.
  • In blitz, 39.1% win rate; recent blitz rating, 581;

Practice Computer, Level 1 ---

  • 2 games played, its accuracy was 7.92%, to 13.5%; best moves, 19.2% to 27.6%; 4 mistakes, 2 blunders in one game; 1 mistake, 4 blunders in other game.

 Fabian-Bot ---

  • In the game I played, Fabian had an off-beat opening, then immediately lost pieces.
  • His accuracy rate in my game was 8.1%; Best move was 22.2%; 0 mistakes, 3 blunders.
  • 330 rapid games played (as of 11-13-19). Win rate of 10.0%; recent rapid rating, 509.

 Mina-Bot ---

  • In the game I played, Mina lost its Queen and many other pieces before being mated quickly.
  • Her accuracy rate in my game was 5.1% and best moves at 11.1%. 2 mistakes, 2 blunders.
  • Only 43 rapid games (as of 11-13-19). Win rate is only 9.3%; recent rapid rating of 529.

 Elani-Bot ---

  • In my game, weak moves, followed by loss of pieces made the game easy to win.
  • Her accuracy was only 6.8%, best move 10.3%; 1 mistake; 1 blunder
  • 264 rapid games (as of 11-17-19); 11.7% win rate; recent rapid rating, 530.

 Komodo2 ---

  • Previously played 2 rapid games with different colors;
  • in game 1, mated in 26 moves; accuracy was at 10.9%, best moves 24.0%; 4 mistakes, 3 blunders;
  • In game 2, had advantage throughout; K’s accuracy went up to 33.0%, best move 37.0%; 2 mistakes, 6 blunders.
  • 163,055 rapid games (as of 11-14-19); 37.6% win rate; recent rapid rating, 544.
  • Recently played blitz game; accuracy 19.0%, best move 29.3%; 2 mistakes; 5 blunders.
  • In blitz, 43.0% win rate; recent blitz rating, 566.

 Filip-Bot ---

  • In the blitz game I played, this bot made senseless moves, then quickly began losing pieces. Went up 19+ points, with mate following.
  • His accuracy was low, at 9.1%; best moves 22.5%; 1 mistake, 4 blunders.
  • 723 blitz game (as of Nov 15, 2019); 18.9% win rate; recent blitz rating, 570.
  • 668 rapid games (11-15-19); 18.7% win rate; recent rapid rating, 553.

 Martin-BOT ---

  • Played this bot in a blitz game. He played the Bird’s, then blundered away the Queen and other pieces. Mated in 24 moves.
  • His accuracy in my game was very low, 8.8%; best move 16.7%; 2 mistakes, 2 blunders.
  • 60 rapid games (as of 11-13-19); 6.6% win rate; Recent rapid rating of 547;
  • 250 blitz games, 15.6% win rate; recent blitz rating of 593.

 Komodo3 ---

  • Previously played 2 games; I blundered in 1st game, letting it stalemate; it lost game 2; 1st computer level which drew me.
  • On the draw, accuracy was 43.3%, 50.8% best move; 2 mistakes, 5 blunders.
  • On the 2nd game, its accuracy was 44.0%, best moves 40.8%; 5 mistakes, 2 blunders.
  • 192,869 rapid games (as of 11-17-19); win rate 48.0%; recent rapid ratings, 570.
  • Recently played blitz game; accuracy 12.0%, best move 2.2%; 5 mistakes; 1 blunders.
  • In blitz, 52.6.1% win rate; recent blitz rating, 582;

 Janjay-Bot ---

  • Played blitz game; Bio says this bot likes to go after the K and is willing to lose pieces to do it; I only saw that it blundered a minor piece by move 7 and a rook by 20. Mate on 34.
  • Her accuracy was 10.5%, best move 21.2%; 12 mistake, 3 blunders.
  • 191 blitz games; 23.0% win rate; recent blitz rating, 636.
  • 103 rapid games; 11.6%win rate; recent rapid rating, 572.

 Milica-Bot ---

  • Played a rapid game against this bot. It made senseless moves with its pawns, N, and B. Ended the game in 11 moves with a smothered mate.
  • Her accuracy rate was very low, 3.8%, best moves 0.0%; 2 mistakes, 2 blunders in 11 moves!
  • 69 rapid games (as of 11-13-19); win rate of 11.5%; recent rapid rating, 575.

Practice Computer, Level 2 ---

  • 2 practice games; accuracy was 11.7% to 15.9%, best moves, 21.4% to 27.1%; 4 mistakes, 2 blunders in one game; 1 mistake, 1 blunder in the other game. 

 Practice Computer, Level 3 ---

  • 2 practice games; accuracy was 9.4%, to 13.8%; best moves 20.6% to 21.6%; 3 mistakes, 1 blunder in one game; 1 mistake, 2 blunders in other game.
  • Putting level 3 here, as it is probably better than level 2, even with lower accuracy % in both games. 

Santiago-Bot ---

  • Played OK in opening until losing its Queen in a discovered attack, then it was easy pickings.
  • Accuracy rate was 28.3%, best move 28.6%; 2 mistakes, 3 blunders.
  • 27 rapid games (as of 11-16-19); win rate, 14.8%. Recent rapid rating 579.

 Oliver-Bot ---

  • In the game I played, the bot moved it K into the open on move 5 for no reason. 14 checks later and the loss of most of its pieces, I mated.
  • His accuracy rate was 28.0%, best move 37.1%; 1 mistake, 5 blunders.
  • 19 rapid games played, as of 11-14-19; 21.0% win rate; recent rapid rate, 582

 Noel-Bot ---

  • Played rapid game; bizarre opening, quick loss of pieces; game was certain by move 14.
  • His accuracy rating was 17.4%, 33.3% best move; 2 mistakes, 2 blunders.
  • 93 rapid games (11-17-19); 64.5% win rate; recent rapid rating, 590.

 Aron-Bot ---

  • Played blitz game; he blundered away two pieces in the first 10 moves; after that, I mated with over 4 minutes left on time.
  • His accuracy rate was 9.2%, best move 22.2%; 1 mistake, 4 blunders.
  • 2313 blitz games (11-16-19); 22.0% win rate; recent blitz rating 614.
  • 657 Rapid Games (as of 11-16-19); 14.1% win rate; recent rapid rating 607.

Practice Computer, Level 4 ---

  • 2 games; accuracy was 14.6% to 25.4%; best moves 25.0% to 30.4%; 2 mistakes, 2 blunders in one game; 3 mistakes, 2 blunders in other game; New data.

 Komodo4 ---

  • Previously played 2 rapid games with different colors; both were easy wins.
  • It’s game 1 accuracy was 20.2%, best move 27.1%; 1 mistakes, 9 blunders
  • Game 2 accuracy 33.9%, best move 33.3%; 3 mistakes, 3 blunders
  • 180,125 rapid games (as of 11-17-19); 47.7% win rate.
  • Rapid rating at time of play was 1130 and 1178; recent rapid ratings are much lower, 640.

 Komodo5 ---

  • Previously played 2 rapid games with alternating games; won them both.
  • In 1st game, its accuracy was 31.0%, best move 33.3%; 4 mistakes, 4 blunders.
  • 2nd game, its accuracy was 29.0%, best move 30.0%; 2 mistakes, 3 blunders.
  • 224,202 rapid games (as of 11-17-19); 50.8% win rate.
  • Rapid rating at time of play were much higher, at 1279 and 1249; Recent rapid rating, 619

Practice Computer, Level 5 ---

  • 2 practice games; accuracy was 23.5% to 41.0%, averaging 32%; best moves 33.3% to 34.2%; 1 mistakes, 0 blunders in one game; 5 mistakes, 0 blunders in other game.

 Wayne-Bot ---

  • This bot’s bio says that he love to use the Q.
  • In the blitz game I played. He moved the K into the open on move 3. After a series of checks, I had 4 pieces buzzing around the K, mating it in 22 moves.
  • Incredibly, his accuracy rate 48.5%; best move 45.5%; 1 mistake, 5 blunders.
  • 36 blitz games; win rate, 16.7%; recent blitz rating, 607.
  • Only 8 rapid games (as of 11-13-19); win rate 25%; recent rapid rating, 667.

 Karim-Bot ---

  • Playing blitz game; True to the bio, this bot pushed the f pawn on his 1st I forked hi K and R on 5, then skewered his K and Q on 16, with the K in the open.
  • His accuracy was 37.5%, best moves 40.4%; 1 mistake, 8 blunders.
  • 622 blitz games (11-18-19); 32.4% win rate; Recent blitz rating, 767;
  • 181 rapid games; 19.3% win rate; recent rapid rating, 748.

 Zara-Bot ---

  • Played blitz game; it went down a R for N exchange on 7, then lost its Q to prevent mate.
  • Her accuracy was 61.8%, best move 44.4%; 2 mistakes, 2 blunders.
  • 453 blitz games; 35.7% win rate; recent blitz rating, 621.
  • 159 rapid games (11-17-19); 27.0% win rate; recent rapid rating, 754.

 Emir-Bot ---

  • Played rapid game; the bot blundered several times; mated the K in the open on move 27 with the 4 minor pieces, never bringing out the Q or the Rs.
  • His accuracy was 11.9%, best move 11.5%; 1 mistake, 2 blunders.
  • 1,114 rapid games (as of 11-18-19); 33.3% win rate; recent rapid rating, 767.

 Komodo6 ---

  • Previously played 3 games when it was rated much higher; won all with alternating colors.
  • Games accuracy ranged from 42.0% - 42.9%; best moves 34.8% - 42.9%; 2 to 11 mistakes; 1 to 9 blunders.
  • 204,221 rapid games (as of 11-17-19); 55.6% win rate; recent rapid rating, 772.

 Maxim-Bot ---

  • Played rapid game; Bot at least played a typical opening, the Sicilian, but then made 4 weak moves in the first 10 moves; it then lost 2 minor pieces and 2 pawns over a series of moves.
  • Its accuracy was 64.2%, best moves, 41.7%; 2 mistakes, 4 blunders.
  • 139 rapid games (11-18-19); win rate 34.5%; Recent rating 867.

Hans-Bot ---

  • In rapid game I played, Hans played a closed Sicilian, then varied on move 5. Within short order, it lost minor and major pieces.
  • In my game, his accuracy rate was 3.8% and best move was 7.7%; 1 mistake, 3 blunders.
  • 3,061 rapid games (as of 11-13-19); Win rate of 37.2%; recent rapid rating, 873.
  • Also played a blitz game; His accuracy was 7.3%, best move 9.6%; 8 mistakes, 2 blunders; win rate 46.0%; recent blitz rating, 794.

 Sven-Bot ---

  • Bio says this bot like to play defensively; found the opposite; His Q came out early, threatening; he then quickly lost two minor pieces, then was mated with the K in the open.
  • His accuracy was 22.5%, best move 30.0%; 2 mistakes, 2 blunders.
  • 358 blitz games; 46.6% win rate; recent blitz rating, 1004;
  • 184 rapid games; 41.8% win rate; recent rapid rating, 977.

 Yabani-Bot ---

  • Played a rapid game; the bio said to expect off-beat openings; that was true, with 1. c3; even more bizarre was 2. Qc2; He then lost 3 pieces, followed by a K side attack and mate on 23.
  • His accuracy was only 5.78%, best move 8.7%; 4 mistakes which probably were all blunders.
  • 681 rapid games (11-19-19); win rate 35.3%; recent rapid rating, 987.

More to Come ---

The moderate and higher-rated computer bots and programs will be described in my next two posts. If I have the energy, I may then round out this series of posts with a summary of my thoughts on computer play in a plus and minus style of post.

Stay tuned for Part II of Man vs Machine!

Originally posted Nov. 21, 2019; 2nd Update March 16, 2020.