(Wasn't V2's Computer level 10 at Elo 2000?) It's amazing to me to see Computer's positional score progressively show -5, -6, -7, -8, etc. What an awesome opponent Chess must be on its highest level. I'm still really surprised at how quickly and convincingly Computer beats Chess Titans, a program I thought was reasonably good. I expect Chess Titans won't be able to win or draw until Computer's level is down to about 5. Computer won every game, just as I expected. Today I played Computer at level 9, then level 8, against Chess Titans level 10. The practical value to others is that if Computer's rating can be determined, then Chess Titans' rating can be fairly well estimated. The practical value for me is that once I find at which level the two programs are roughly equal in strength, I can start playing Computer at that level for practice, which will presumably gradually force me to increase my skill as I gradually increase the level of Computer. I decided to keep matching Chess Titans level 10 against Computer, while decreasing Computer's level until Chess Titans level 10 could beat it. If you study the scatter plot, I think you'll agree that on a macro scale, there is no difference between USCF and FIDE ratings below 2200 (and even much less than 200 rating points difference above 2200.) Of course, on a micro scale, it's important to know which data point you are. I'll also show the link to the scatter plot, but I wanted you to also have the link to Glickman's page. On Glickman's page below, go to the paragraph that begins with "I devised a method to update.", and look for the link to the scatter plot. I'm used to USCF ratings, and Chess Titans is said to have a rating of 1800 USCF (which is what I thought), or 1600 FIDE.ĭon't worry about differences between USCF and FIDE ratings, at least under 2200. I think what is throwing me off is the difference between USCF versus FIDE ratings. I found that stronger programs motivated me to get beyond that level, though I don't have enough rating points yet to prove I've advanced since then. I believe I got lazy, though, since I reached a comfortable level of proficiency looking ahead only 3-4 moves against computers for years, so I was insufficiently motivated to search deeper, which meant I got stuck at a plateau. And of course tactics, but computers will keep you on your toes for tactics about 3-4 moves deep. Another thing that helped was use of miniplans: making a knight outpost, doubling pawns, seizing an open file, opening the center whenever I got the two bishops, etc. Another thing that helped me was learning some positional/opening tricks that I never found in any book, like playing P-QB3 or P-KB3 if you have a bishop on your 3rd rank (to protect against your opponent's N-B5 to attack that bishop). Also, openings give you the best unit placement and suggest the best plans (assuming you know the plans associated with each opening). Unless you know the openings, you're going to be playing suboptimal moves, and with each suboptimal move your position declines several percent, so over time your position declines irrecoverably. In both cases the thing that made the most difference for me was knowing the openings. I'm having to learn all over with 's program, though, it seems, which must be about 1900 rating despite supposedly being 1600-definitely better than Chess Titans. I beat Chess Titans on level 10 about 85% of the time now. I think anybody can get to 1700 and beyond. Otherwise, I believe that's the fastest win I've ever seen against Chess Titans on level 10, so maybe you're looking ahead farther than I am. This is where the king belongs after queenside casting, to prevent the queen from invading at a1. Nd5 to gang up on Black's pinned KN earlier than you did.ġ5. In the above game, some things to have considered were.ġ4. I can't find an earlier thread at the moment, but although somebody posted here years ago that they estimated Chess Titans to be rated about 1650-1700, a more recent estimate when I mentioned that in an unrelated thread was that Chess Titans was rated maybe 1800, which sounds closer to me.