I have released the CT800 software V1.10. The announcement is a bit delayed because the release overlapped with changing my web hosting provider in favour of one who embraces Let's Encrypt SSL certificates.
- More aggressive time management for game-in and tournament modes.
- New easy move detection.
- New horizon search PV deepening.
- Corrected king safety evaluation for the pawn shelter.
- Added some backports of the NG-Play update from v9.86 to v9.87 that speed up the search.
- Battery shutdown monitoring works at all times.
- "Always on" light mode works.
- Saving the display contrast works.
- Custom display characters make the dialogues look cleaner.
- Some code speedup and cleanup.
- Updated the ARM compiler from GCC 5.4 Q3/2016 to GCC 6.3 Q1/2017.
- Enhanced the opening book.
- The opening book uses combined CRC-40 hashing instead of CRC-32.
- Opening book allows implicit queening.
- The opening book compiler got a code refactoring for more clarity and robustness. It displays error messages in clear text. Besides plain ASCII input book files, it also accepts UTF-8 with and without BOM.
- The build batch file resp. shell script does a fully automated and complete build, including useful error messages if something goes wrong.
I have also re-designed the website to look somewhat more modern, but still simple, fast, responsive and accessible. The start page https://www.ct800.net is just 35.6 kB - 70 times smaller than the average web page size these days. Oh, and it is https only now.
Daniel Anulliero wrote:Great !
It looks like Compuchess or Conic chess from early days of dedicated chess machines !
Thanks. What I had in mind was the famous Mephisto "briquette", just with current technology. Oh, and with more strength, should be around ELO 2100, i.e. a challenge for hobby players. Especially since it is geared to play humans, not computers.
It can't rival with Stockfish, of course - but the power consumption of the CT800 in full thinking mode would barely suffice to power the mouse of a PC running Stockfish.
For your amusement, here is a game of the CT800 vs. Mephisto MM5 with HG550 opening module, both at 30 minutes per game. The MM5 is one of my favourite challengers because it plays very actively. My compliments to Ed Schroeder for this marvel on a tiny 5 MHz 6502 CPU.
The CT800 starts out in a calm way and soon realises a way to use its rooks - something that the baseline NG-Play usually does not, btw. The MM5 uses its own rooks to counter the pressure and likes to grab space with pawns, but it doesn't take long for the CT800 to turn the position into a confusing mess. The MM5, being a computer itself, keeps track of the position, more or less, and doesn't lose that much, but average hobby players would face a major headache. The endgame of KRP:KN still is a clear win, and the CT800 doesn't blow it.
When analysing the game, it turned out that the MM5 could have gained a probably winning advantage in the 26th move. However, the depth of these machines is not what you may be used to on PCs. The CT800 only reaches 8 plies depth (plus QS) in the middlegame at these time controls, and the MM5 is in the same range (or less).
But the point is, both show quite some fighting spirit at the level of advanced hobby players.
[pgn][White "CT800"]
[Black "Mephisto MM5"]
[ECO "C47"]
[Result "1-0"]
Daniel Anulliero wrote:Great !
It looks like Compuchess or Conic chess from early days of dedicated chess machines !
Thanks. What I had in mind was the famous Mephisto "briquette", just with current technology. Oh, and with more strength, should be around ELO 2100, i.e. a challenge for hobby players. Especially since it is geared to play humans, not computers.
It can't rival with Stockfish, of course - but the power consumption of the CT800 in full thinking mode would barely suffice to power the mouse of a PC running Stockfish.
Oh oh yes the legendary "briket" I have one of them !
2100 elo is good indeed
Good work
Ras wrote:
When analysing the game, it turned out that the MM5 could have gained a probably winning advantage in the 26th move.
[d]1r4k1/2qn2p1/pr5p/Qpp2b2/3P1p2/2P5/PR1BPPBP/2R3K1 b - - 0 26
Those tactics are nice! What a shame MM5 missed it. Nice position.
"Good decisions come from experience, and experience comes from bad decisions."
__________________________________________________________________
Ted Summers
The MM5 didn't have a chance to see that. White has to give some material after 26. ... f3, but only to avoid even more loss later. This "later" is way beyond the horizon of both machines. DeepShredder is at 26/49 plies on my PC already after 1 second, so that's a different league.
But that's normal when you analyse games of amateur players, which is what both machines are: you'll see an oscillating graph, and the last error loses. That's the difference to master games which have a much more steady graph.
On the upside, it's actually fun to play the machine because a human amateur player can win. My only problem here is that I'm better at C than at chess.