+Andrew Davie Posted April 20, 2020 Author Share Posted April 20, 2020 (edited) Minor update - this version allows you to press the SELECT switch to force the computer to play a move immediately. Mind you, that's cheating... but it's there, so have at it. Edit: whoops I seem to have put it on RESET accidentally. Have replaced with a fixed version. chess_20200420_CHEATER!_fixed.bin Edited April 20, 2020 by Andrew Davie 2 1 Quote Link to comment Share on other sites More sharing options...
Voxel Posted April 20, 2020 Share Posted April 20, 2020 3 hours ago, Andrew Davie said: but don't forget this is running on a very underpowered Atari 2600. If you don't ask you don't get and I had to ask. Thank you for the consideration. It's just how I play, see a line, play it, then pick up where I left of, if I can. 4 hours ago, Andrew Davie said: Here is my first candidate version Thank you for the preview. Quote Link to comment Share on other sites More sharing options...
Voxel Posted April 20, 2020 Share Posted April 20, 2020 2 hours ago, Andrew Davie said: Minor update - this version allows you to press the SELECT switch to force the computer to play Could you confirm this is force best move found and not the current move being analysed? Quote Link to comment Share on other sites More sharing options...
+Andrew Davie Posted April 20, 2020 Author Share Posted April 20, 2020 Just now, Voxel said: Could you confirm this is force best move found and not the current move being analysed? Confirmed. That is how it should be behaving. 1 Quote Link to comment Share on other sites More sharing options...
Voxel Posted April 20, 2020 Share Posted April 20, 2020 Thanks, I'd given up on seeing this one. Quote Link to comment Share on other sites More sharing options...
+Andrew Davie Posted April 20, 2020 Author Share Posted April 20, 2020 Some notes: This latest version should be reasonable at opening and early gameplay. It will start to fall down at late midgame, because it is still "tuned" for openings. And it is completely clueless at endgames. If it mates you, then it's fortune not design. Fortunately, I have a plan. The positioning is mostly table driven, and it will be fairly simple to have "opening tables", "midgame tables", "late midgame tables" etc. And I will also be dynamically updating the search and quiescence maximum depth to try and throttle the move speeds. So, more depth for endgames. That's the plan. 2 Quote Link to comment Share on other sites More sharing options...
+Andrew Davie Posted April 20, 2020 Author Share Posted April 20, 2020 1 minute ago, Voxel said: Thanks, I'd given up on seeing this one. I do keep things in mind and try to get stuff in there 1 Quote Link to comment Share on other sites More sharing options...
+Andrew Davie Posted April 20, 2020 Author Share Posted April 20, 2020 6 minutes ago, Voxel said: Thanks, I'd given up on seeing this one. When I say "best move found" you still have the issue that the moves are pretty much unsorted, and it goes through them in order. The only "sorting" happening at the moment is to put captures first. Now, later down the track, when I do a preliminary 2 ply search to get a rough idea of the "value" of each move (probably at the top ply only), then I can sort the moves according to the scores of each, and that will in turn make huge differences to the alpha-beta pruning and the search will run faster AND the "better moves" are more likely to be searched first. And that means that using SELECT will give you a better chance of good moves when you abort. Right now it's a bit of luck of the draw - so when I say "best move found so far" that's true, but it's "best move out so far out of effectively randomly choosing moves". If that makes sense. 1 Quote Link to comment Share on other sites More sharing options...
+Andrew Davie Posted April 20, 2020 Author Share Posted April 20, 2020 Note: I just had one crash/reset. I don't for sure know this was code I'm fiddling with so please report any issues. It was close to a checkmate situation (computer winning). 1 Quote Link to comment Share on other sites More sharing options...
Voxel Posted April 20, 2020 Share Posted April 20, 2020 I've not yet had enough time to try a full game have so far had a few starts around 10 moves in and I'm pretty impressed. It's playing a very tight formation with the pieces to begin with which makes it hard to break its' defence lines. So I'll agree that the opening is very good. It's quite cunning. If I mirror it's style which I did in the second game we got to the point where we had a big face off and just like the more recent iterations it was preparing to exchange. It's still a little reluctant to exchange the queen (trying to bait her is becoming quite the challenge). As the game becomes more intense it does a lot of pondering, which means I need to make some time to have a complete game. I will add I've not used the force move, I'd like an indicator when it goes deep via AtariVox, or a sound when it's deep and has a satisfactory best line (by whatever internal score system it measures a good move). Really don't want to force move, when it's compromised with searching another node. 1 Quote Link to comment Share on other sites More sharing options...
AW127 Posted April 20, 2020 Share Posted April 20, 2020 This updated version of the game really looks good, "Andrew". One suggestion - maybe the two colors used for the two pieces-sets should be made a little bit more different? Now it looks like rosa and orange and it would be better if they would be more different. Either the orange ones could be made a little bit darker or the pink ones a little bit brighter. Quote Link to comment Share on other sites More sharing options...
+Andrew Davie Posted April 21, 2020 Author Share Posted April 21, 2020 3 hours ago, AW127 said: This updated version of the game really looks good, "Andrew". One suggestion - maybe the two colors used for the two pieces-sets should be made a little bit more different? Now it looks like rosa and orange and it would be better if they would be more different. Either the orange ones could be made a little bit darker or the pink ones a little bit brighter. Thank you. Changing the colours is complex/difficult. One set of pieces are made of colours (ABC), the other made of colours (AB), and the board squares made of C It's not a matter of just choosing a new colour for each side. You have to find colours that mix in the way describe above... that look different enough. And that's very difficult indeed. I've gone through many, many iterations. Quote Link to comment Share on other sites More sharing options...
RevEng Posted April 21, 2020 Share Posted April 21, 2020 Regarding colors (you're probably sick of hearing about colors) you may want to dial down the lumas from A to 8, and see if that makes a difference. The constant-chroma palette scheme tends to make high luma colors very pastel, and hard to distinguish from one another. This is less of a problem on 6-switchers with their chroma boosting resistor, and it would be a wild-card on units with video mods. Also, $F# tends to drift as the console warms up from yellow to orange. (and beyond in some consoles) $1# would be a similar chroma, but would be a lot more stable. (I made an illustration of this effect for the 7800, but the 2600, A8 computers, and the 7800 all use the same basic color generation scheme, and all exhibit the same warm-up palette drift.) Quote Link to comment Share on other sites More sharing options...
+Andrew Davie Posted April 21, 2020 Author Share Posted April 21, 2020 2 minutes ago, RevEng said: Regarding colors (you're probably sick of hearing about colors) you may want to dial down the lumas from A to 8, and see if that makes a difference. The constant-chroma palette scheme tends to make high luma colors very pastel, and hard to distinguish from one another. This is less of a problem on 6-switchers, which have a chroma boosting resistor, and a wild-card on units with video mods. Also, $F# tends to drift as the console warms up from yellow to orange. (and beyond in some consoles) $1# would be a similar chroma, but would be a lot more stable. (I made an illustration of this effect for the 7800, but the 2600, A8 computers, and the 7800 all use the same basic color generation scheme, and all exhibit the same warm-up palette drift.) Thanks for the analysis/suggestion. This binary uses $1X and reduced from A to 8 as suggested. Let me know what you think. chess20200421_palette_test.bin Quote Link to comment Share on other sites More sharing options...
+Andrew Davie Posted April 21, 2020 Author Share Posted April 21, 2020 Or this... $86/$48/$18 Changed from $9X to $8X per your shifting diagram, as it stays "bluer" over all conditions. Very interesting. With colour 8 I was able to increase luma from 4 to 6, thus making the contrast between white/black pieces a bit greater too. Quote Link to comment Share on other sites More sharing options...
Keatah Posted April 21, 2020 Share Posted April 21, 2020 Just wondering. I have two versions of Stella 3.9.3 and 6.1.1. I get different colors between each emulator. But only on this chess game. All other roms I tried look identical between the versions. Ideas? Or not a big deal..? Quote Link to comment Share on other sites More sharing options...
RevEng Posted April 21, 2020 Share Posted April 21, 2020 5 minutes ago, Andrew Davie said: Or this... $86/$48/$18 Changed from $9X to $8X per your shifting diagram, as it stays "bluer" over all conditions. Very interesting. With colour 8 I was able to increase luma from 4 to 6, thus making the contrast between white/black pieces a bit greater too. Yep, I think that's the one. They're subtle changes in stella, but I think it will make a bigger difference on those units with weaker chroma to begin with. I guess we'll see what other testers report. Unfortunately I don't have a way to run this format on real hardware, at the moment. Quote Link to comment Share on other sites More sharing options...
+Karl G Posted April 21, 2020 Share Posted April 21, 2020 Here's a side by side on my CRT of the previous binary and the posted palette test binary, FWIW. The left one is the original. It's hard to see any difference in my pictures, and it was very subtle. I'd say that color contrast between the two sides is fine in both, though. Edit: on an Atari Jr. with a composite mod. 1 Quote Link to comment Share on other sites More sharing options...
+Andrew Davie Posted April 21, 2020 Author Share Posted April 21, 2020 Just now, Karl G said: Here's a side by side on my CRT of the previous binary and the posted palette test binary, FWIW. The left one is the original. It's hard to see any difference in my pictures, and it was very subtle. I'd say that color contrast between the two sides is fine in both, though. Ha! They look identical to me. But my wife would say "of course you think they do!" Quote Link to comment Share on other sites More sharing options...
+Karl G Posted April 21, 2020 Share Posted April 21, 2020 Just now, Andrew Davie said: Ha! They look identical to me. But my wife would say "of course you think they do!" Yeah, the difference was hard for me to see, and even harder to discern from my pictures. I only know which is which because of the piece positions. Quote Link to comment Share on other sites More sharing options...
+Andrew Davie Posted April 21, 2020 Author Share Posted April 21, 2020 Here's a binary with the "approved" palette -- this will look a bit different on the CRT as it's not just an intensity change. chess20200421_palette_86_48_18.bin Quote Link to comment Share on other sites More sharing options...
+Andrew Davie Posted April 21, 2020 Author Share Posted April 21, 2020 Here's one I like slightly better. See, get me started on colours and it's a rabbit hole... chess_20200421_palette_86_4a_16.bin 1 Quote Link to comment Share on other sites More sharing options...
+mksmith Posted April 21, 2020 Share Posted April 21, 2020 38 minutes ago, RevEng said: Regarding colors (you're probably sick of hearing about colors) you may want to dial down the lumas from A to 8, and see if that makes a difference. The constant-chroma palette scheme tends to make high luma colors very pastel, and hard to distinguish from one another. This is less of a problem on 6-switchers with their chroma boosting resistor, and it would be a wild-card on units with video mods. Also, $F# tends to drift as the console warms up from yellow to orange. (and beyond in some consoles) $1# would be a similar chroma, but would be a lot more stable. (I made an illustration of this effect for the 7800, but the 2600, A8 computers, and the 7800 all use the same basic color generation scheme, and all exhibit the same warm-up palette drift.) Palette fun Mike ? We are having similar discussions about a project we've been working on for the 7800 1 Quote Link to comment Share on other sites More sharing options...
+Karl G Posted April 21, 2020 Share Posted April 21, 2020 I've added the third screenshot to the same post for comparison. The new one is the one with the white knight not in the starting position. It does indeed look different, but I'm not sure that I prefer it to the others. Edit: From the post that says "approved palette" Quote Link to comment Share on other sites More sharing options...
+Andrew Davie Posted April 21, 2020 Author Share Posted April 21, 2020 OK. Please make it stop! chess20200421_palette_84_4a_16.bin 1 Quote Link to comment Share on other sites More sharing options...
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.