Chess

LuisRodg · 2008-03-29 04:30:43

Could you please add different levels of difficulty?

Look at this game in pgn-viewer:

1: e4 Nc6
2: f4 Nf6
3: Nc3 d6
4: Nf3 Be6
5: d4 h5
6: d5 Nxd5
7: exd5 Bxd5
8: Qxd5 Nb4
9: Qb5+ c6
10: Qxb4 c5
11: Bb5+ Qd7
12: Bxd7+ Kxd7
13: Qxb7+ Ke8
14: Nb5 Rd8
15: Qc6+ Rd7
16: Qc8+ Rd8
17: Nc7++

in move 5-6 it let me fork one of his pieces but instead of losing just that one, it ate with the knight which ended up losing 2 pieces. It went and gave me another piece later on.

Also, when a move gives checkmate, such as in my example "17: Nc7+" it is written with 2 plus signs. like "17: Nc7++"

LuisRodg · 2008-03-29 04:38:26

Also. I was just playing this game:

1: f4 Nf6
2: Nc3 Nc6
3: e4 Nd4
4: Nf3 Nxf3+
5: Qxf3 c5
6: Bc4 d6

and I was trying to castle for the 7th move and it didnt let me. I think its because when the knight ate on f3 it was checking me. But since I ate with the queen and my king never moved I could still castle under chess rules. Might want to check on that.

JaneFairfax · 2008-03-29 05:44:09

Er, I just made another en passant move but only my pawn was moved and the computers pawn that was supposed to be captured was not captured.

Also Luis was perfectly entitled to castle from his position. The fact that his king was temporarily in check a few moves earlier doesnt mean he cant castle for the rest of the game.

MathsIsFun · 2008-03-29 09:37:46

The moves help me, as I can actually play the game and see what the computer does.

Luis: for some reason it thinks the King had been moved ... I will investigate.

JaneFairfax · 2008-03-29 10:22:53

The king had not been moved in Luiss game. My suspicion is that the king was in check earlier and the program thought Luis couldnt castle any more. This is not correct. While you cant castle when your king is in check, you can when the check is removed (provided you havent moved your king or the rook towards which you are castling).

Zach · 2008-03-29 10:24:22

JaneFairfax wrote:

I also just found out that the computer doesnt understand the rule of en passant.

It understands the rules, it allows you to do it.

It just doesn't allow you to do it IN FULL and take the piece. Also, I'd like to choose what my prawn becomes.

MathsIsFun · 2008-03-29 11:49:46

It was the routine that looks for Checkmate. It tests all possible moves to remove Check, and did not correctly reset the status after doing this.

I will now check the en-passant thing. If anyone can post the moves that lead to the fail, that will help.

MathsIsFun · 2008-03-29 12:06:24

Posted a new version (v0.74) which hopefully fixes both the castling and en-passant naughties.

I hope to come up with a *slightly* more difficult version, but I am already hitting time limits.

JaneFairfax · 2008-03-30 04:50:26

The computers current level is just fine for beginners. It has already beaten me twice (and I consider myself high-intermediate).

MathsIsFun · 2008-03-30 08:13:44

OK, then!

mathsyperson · 2008-03-30 10:28:13

I managed to beat it in around 10 moves just now. I don't have the game notation because halfway through I clicked the "Test" button to see what it did and lots of weird stuff overwrote it.
It also flashed up the "Well Done" message as if I'd won.

It was a slight variation of the one where you put your queen diagonally next to the enemy king and a bishop stopping the king from taking her. A knight helped out too. The computer won't fall for that anymore though.

JaneFairfax · 2008-03-30 12:58:43

mathsyperson wrote:

I managed to beat it in around 10 moves just now. I don't have the game notation because halfway through I clicked the "Test" button to see what it did and lots of weird stuff overwrote it.
It also flashed up the "Well Done" message as if I'd won.
It was a slight variation of the one where you put your queen diagonally next to the enemy king and a bishop stopping the king from taking her. A knight helped out too. The computer won't fall for that anymore though.

This is the classic game known as Fools Mate. I managed to beat the computer with it in only six moves:

1: e3 Nc6
2: Bc4 Nf6
3: Qf3 d6
4: d4 Nxd4
5: exd4 Nd7
6: Qxf7

Really, Fools Mate is so stupid I hardly thought any self-respecting chess-playing software would even try to be beaten by it.

JaneFairfax · 2008-03-30 13:11:46

Wait. Mate in just four moves!

1: e4 Nf6
2: Qf3 Ng8
3: Bc4 Nc6
4: Qxf7

I hope the slightly more difficult version is going to take care of that.

MathsIsFun · 2008-03-30 13:26:21

Ouch

But I always get:

1: e4 Nf6
2: Qf3 Nxe4

Still ends in mate, though.

(BTW: there is a mild "ranomity" in the chess game, to keep it interesting ... you may have noticed it sometimes makes different moves)

Yes, hopefully I can get it past that.

AshJones555 · 2008-03-31 02:47:49

1: Nf3 Nf6
2: Ng5 Nc6
3: e3 d6
4: Bc4 Bf5
5: Nxf7 Bd3
6: Nxd8 e5
7: Nxc6 Be7
8: Nxe7 Bxc4
9: Nf5 O-O-O
10: b3 Bd3
11: cxd3 Rf8
12: Ne7 Kb8
13: Bb2 e4
14: Bxf6 Rxf6
15: Nd5 Rxf2
16: Kxf2 exd3
17: Qf3 Rf8
18: Nf4 d5
19: Nc3 d4
20: exd4 g5
21: Ne6 h5
22: Nxf8 g4
23: Qf5 h4
24: Ng6 c6
25: Re1 g3
26: hxg3 hxg3
27: Kxg3 Kc7
28: Re7 Kd6
29: Qc5

I'm a bit of a begginer but i managed to win, the computer seems to use a variety of decent moves and seemingly pointless moves and I agree that a + for check should be included in the coding because it's interesting to see how check could develop into forced mate etc.

Last edited by AshJones555 (2008-03-31 02:48:32)

AshJones555 · 2008-03-31 05:33:09

oh and:
1: e4 Nf6
2: Bc4 Nxe4
3: Qh5 Nc6
4: Qxf7

basically the same as Jane.

MathsIsFun · 2008-03-31 12:33:03

Posted version 0.76 (Chess) that now has difficulty levels. Refresh the page if you don't see "v0.76".

Can you still beat it easily at "Medium"?

I also have some code that will choose different moves (of the best available) to provide an interesting game, but it is switched off so that we can more easily duplicate errors if we need to.

MathsIsFun · 2008-03-31 21:08:35

Note: the "medium" level has the addition of an "attacked/protected" count. So if a piece is attacked but not protected a penalty of 30% of the pieces value is added.

To work out "attacked and protected" I look at each piece in turn and record on the board where it can move. This gives me a count for each square. Then I look at each piece's location to see if the enemy has more pieces that can move onto its square than friendly pieces can.

But this does not account for relative values of the pieces. For example if a pawn attacks the Queen.

Can anyone think of it quick and effective way of accounting for relative values of pieces?

mathsyperson · 2008-04-01 01:00:03

There's a fairly well-known system that gives labels each piece as being worth an amount of pawns:

Knight - 3 pawns
Bishop - 3 pawns
Rook - 5 pawns
Queen - 9 pawns

Of course, the actual value of pieces is very situational. Knights are worth far more than bishops in the early game, bishops are better than knights in endgame, and it's perfectly possible to have a pawn being more valuable than a queen in certain positions.
But this system is still much better than treating all pieces equally.

LuisRodg · 2008-04-01 02:22:34

Look at this game playing medium style:

1: e4 Nf6
2: Nc3 Nc6
3: f4 d6
4: Nf3 Bd7
5: Bb5 Nxe4
6: Nxe4 d5
7: Nc5 b6
8: Nxd7 a6
9: Bxc6 Rc8
10: Nf6

Why did it give me a free piece at move 5 with Nxe4?

Also when I checkmated it with a double check it didnt record the final move as Nf6++ or even Nf6+ indicating im checking him with knight and bishop.

LuisRodg · 2008-04-01 02:47:26

By the way, this topic made me make my own chess engine in C++. Its even weaker than this one though. The move generating and implementing the alpha-beta searching took quite a while for me so my evaluation function is as simple as you can get. Its basically what mathyperson suggested of the values of the pieces.

int Eval_function()
{
int value_piece[6] = {VALUE_PAWN, VALUE_KNIGHT, VALUE_BISHOP, VALUE_ROOK, VALUE_QUEEN, VALUE_KING};
int score = 0;
for (int i=0;i<64;i++) {
if (color[i] == WHITE)
score += value_piece[piece[i]];
else if (color[i] == BLACK)
score -= value_piece[piece[i]];
}
if (side == WHITE)
return score;
return -score;
}

This is as simple as you can get. If you want it to improve a little then you need to start implementing other deductions and bonus in scores. For example, as you can see in this code, the score consists ONLY of the pieces in the board. It doesnt take into account any positional knowledge etc.

A suggestion for you MathIsFun, which is something im going to implement in mine in the summer (since im almost in finals right now and getting to where i am basically took me all day yesterday and I had to study for Discrete ) Start tweaking your evaluation function such as this:

give bonus score if a side has a rook on an open file.
give bonus score if 3 pawns are covering the king.
give bonus score if theres a knight near the king.
deduct score if the king has an open file.
give bonus score if the bishop's diagonal are longer = more mobility.
give bonus score if your pawns control the center of the board.

Etc. I read this a site once about how to tweak your evaluation fucntion. It looks very simple but it actually serves of help in the strength of play.

Also, MathIsFun you asked about relative value of the pieces. I assume you have your knights valued as 3 (or 300, as I do so all my work is in integer). In a simple evaluation function where you only take into account the value of the piece, your basically saying that a Knight in its original position is worth the same as a knight placed on a key spot such as e4 and d4 controlling the center of the board. Same can be said about having control of the center. If any of you took chess classes when you were little, I can still remember my trainer "control the center of the board..." etc. Having control of the center is a key element. So as I said, a knight placed and protected in d4 is worth more than 3 but also a backwards knight in b1 which offers no help is worth less than 3. Its all relative to the position. Every piece has similar stuff, if you add this to every piece etc. The engine will increase its playing strength dramatically even if it sounds that your doing simple stuff.

Hope I helped.

LuisRodg · 2008-04-01 04:05:35

Meh.

I was just playing with my stupid engine and the game went on like this:

e4 d6
f4 Na6
Nf3 Rb8
Bxa6 Ra8

Even if my evaluation function is as simple as you can get, when its searching, guiding itself by my eval function, why did it find Ra8 to be "better" than bxa6? If I understand this alpha-beta searching, the search function is supposed to evaluate every single position, bxa6 is supposed to give a higher score than Ra8 since with any other move that isnt bxa6 the score goes down by 3...

Hmm.

MathsIsFun · 2008-04-01 08:21:46

Alpha-Beta stops searching down a branch when it finds it is getting a worse result than a previous branch. It is searching for the "best worst", so a "worse worse" does not need to be checked further. (Example: if moving the queen results in losing a pawn, and then you start testing a Knight's move and find you have lost a Bishop you don't need to see if there are any worse possibilities for that Knight's move)

I already have material values (100 for pawn, 900 for queen etc) and lots of positional code. I am looking for an intelligent "piece exchange" function. But it all has to be super-rapid (it is a "how smart can I be in a limited time" thing).

Math Is Fun Forum

#26 2008-03-29 04:30:43

Re: Chess

#27 2008-03-29 04:38:26

Re: Chess

#28 2008-03-29 05:44:09

Re: Chess

#29 2008-03-29 09:37:46

Re: Chess

#30 2008-03-29 10:22:53

Re: Chess

#31 2008-03-29 10:24:22

Re: Chess

#32 2008-03-29 11:49:46

Re: Chess

#33 2008-03-29 12:06:24

Re: Chess

#34 2008-03-30 04:50:26

Re: Chess

#35 2008-03-30 08:13:44

Re: Chess

#36 2008-03-30 10:28:13

Re: Chess

#37 2008-03-30 12:58:43

Re: Chess

#38 2008-03-30 13:11:46

Re: Chess

#39 2008-03-30 13:26:21

Re: Chess

#40 2008-03-31 02:47:49

Re: Chess

#41 2008-03-31 05:33:09

Re: Chess

#42 2008-03-31 12:33:03

Re: Chess

#43 2008-03-31 21:08:35

Re: Chess

#44 2008-04-01 01:00:03

Re: Chess

#45 2008-04-01 02:22:34

Re: Chess

#46 2008-04-01 02:47:26

Re: Chess

#47 2008-04-01 04:05:35

Re: Chess

#48 2008-04-01 08:21:46

Re: Chess

Board footer