|
Post by Rod on Feb 2, 2021 21:26:30 GMT
I just tested the export function in j-board and it seems to be working ok with Chess v2.3 I tried a white game and a black game and it loaded with the correct state of play.
Because you can only load or save on the players turn, Chess v2.3 loads any mid game file and gives the player control, not the AI. The only exception is loading a new game named "NewBlack.gam", where the AI is white and gets first move.
I also set up j-board with the final players move still to make, So whatever the last move was, set it back to where it came from then export and make the move manually in Chess v2.3
|
|
|
Post by B+ on Feb 3, 2021 17:12:34 GMT
I just tested the export function in j-board and it seems to be working ok with Chess v2.3 I tried a white game and a black game and it loaded with the correct state of play. Because you can only load or save on the players turn, Chess v2.3 loads any mid game file and gives the player control, not the AI. The only exception is loading a new game named "NewBlack.gam", where the AI is white and gets first move. I also set up j-board with the final players move still to make, So whatever the last move was, set it back to where it came from then export and make the move manually in Chess v2.3 Looks like a job for reverse, just add button with code you have. What's the name of this thread? ;-))
|
|
|
Post by Rod on Feb 3, 2021 19:07:57 GMT
Ok, this link is v2.4 with the reverse button implemented. Not sure it is as easy as swapping the board, who goes next, undo etc etc But see if it does what you expect. gamebin.webs.com/Chess%20v2.4.zip
|
|
|
Post by B+ on Feb 3, 2021 19:09:19 GMT
Thanks Rod I'll give it a test run as soon as I am back from errands.
|
|
|
Post by tsh73 on Feb 3, 2021 20:35:50 GMT
It looks like pressing Reverse we can make computer play itself.
|
|
|
Post by Rod on Feb 3, 2021 20:50:40 GMT
Yes, inevitable, the set at the top of the board is AI. Since you can only make the reverse choice while it is the players turn (lower set) when it reverses it is that sets turn to play. As it is at the top of the board now it is the AI that will make the play. Please don’t ask for anything different!
|
|
|
Post by B+ on Feb 3, 2021 21:46:53 GMT
It looks like pressing Reverse we can make computer play itself. You have to keep pressing reverse for that effect, OK! Rod: Whew! Rod you make it sound like a bad thing. I thought if I pressed Reverse the computer would get stuck playing itself (now that I say that, why would that happen? ) That's a bonus with reverse. I just tested first game, start with reverse and computer plays what it always does QBP out 2, why that? it's seems terrible to me, I can't lose. Maybe something to randomly select an equally good move? Anyway a couple of moves and I Reverse again and everything worked fine, I took that lousy start and won as white. So far, I like! But who can resist being able to beat a computer AI (with a little help from my good friend Mr Undo), a great way to clear cobwebs if you haven't played for awhile.
|
|
|
Post by Rod on Feb 3, 2021 22:52:09 GMT
The original code favours central moving forward play. It adds a little bit to the value the nearer the centre and the further forwards it can play. So that is creating the first move. It applies this to every move but it is just a tiny nudge compared to an expected capture.
|
|
|
Post by tsh73 on Feb 4, 2021 4:04:36 GMT
Speaking of piece estimation. When I did export I found that black king worth 7500, while white only 5000. So black AI values defending his king more then attacking white one?
Now then we reverse sides, these numbers stay. So white and black "think" differently?
|
|
|
Post by Rod on Feb 4, 2021 8:12:20 GMT
Yes, I am not really sure why. I think one of the consequences is that it can tell when black is in checkmate and should give up but I don’t think it can tell white is in check and should give up. The way it was written never anticipated a white AI. I had no hope of testing such nuances. But now we have J-Board I can try a few end games to see.
I thought I could have white in checkmate and it still let play continue. Now does reversing the game and the value just swap the problem or do we need a new way to assess check?
Just tested a few black and white end games, seems to recognise either color is checkmate. So I will leave well alone!
|
|
|
Post by B+ on Feb 4, 2021 18:27:03 GMT
Speaking of piece estimation. When I did export I found that black king worth 7500, while white only 5000. So black AI values defending his king more then attacking white one? Now then we reverse sides, these numbers stay. So white and black "think" differently? I think the asymmetrical number values for Kings might be an artifact from the code built from assumption human plays White and AI plays Black in original code. Of course the AI is going to value it's stuff more than other's. Ha! every human finds that out when they have a garage sale ;-)) It's a rule of the living, you have to value your stuff more, it is your responsibility for care of it. Nobody else wants to wash your car ;-)) Ha! just realized connection between Pinocchio and Pygmalion and the author of the original code for Chess here. Also it might be smart to swap those numbers when we swap human, AI places. AI might play better when it's White.
|
|
|
Post by Rod on Feb 4, 2021 20:48:34 GMT
Not sure I understand why there is the difference. All other pieces are of equal value. Until we do understand it a little better I will leave it as is.
|
|
|
Post by B+ on Feb 4, 2021 21:58:52 GMT
Not sure I understand why there is the difference. All other pieces are of equal value. Until we do understand it a little better I will leave it as is. I thought I explained it well, you have to value your King way more than the opponent's.
|
|
|
Post by tsh73 on Feb 5, 2021 6:11:08 GMT
I tried to change king values in newBlack.gam - to check if I could play side by side and see a difference in a gameplay - but found that this change breaks castling. Because pieces for castling checked by constant weight.
May be I'll try a bit harder next time.
|
|
|
Post by Rod on Feb 5, 2021 8:26:11 GMT
Well yes black values it’s king more than white’s king. But since black values positive outcomes and white values negative outcomes there is already a massive variance between the two. On top of that the value is so much higher than other players that it can only make a difference when it’s king on king. Even here 5000 to -7500 is such a gap that 5000 to -5000 would seem to make little difference.
The valuation only matters in the evaluation scoring code. Elsewhere it is sometimes used to know it’s a king, castling for example but I understand all those code points and they are not changed by the difference in value. (But need tweaked if we do change the value)
Why would king on king need the difference? I recon It does not actually care and 5000 to -5000 would play exactly the same. I need to study the original code. I am thinking it might have been a way to know black king was in check. But has been made redundant.
|
|