Beginners Guide to TikTok for Search - Rachel Pearson - We are Tilt __ Bright...
THE IMPORTANCE OF PIECE DIFFERENCE IN CHECKERS AI
1. THE IMPORTANCE OF A PIECE DIFFERENCE
FEATURE TO BLONDIE24
Belal Al-Khateeb Graham Kendall
bxk@cs.nott.ac.uk gxk@cs.nott.ac.uk
School of Computer Science
(ASAP Group)
University of Nottingham
6. Samuel’s Checkers Program
6
- 1959, Arthur Samuel started to look at
Checkers
- The determination of weights through
self-play
- 39 Features
- Included look-ahead via mini-max (Alpha-
Beta)
- Defeated Robert Nealy
7. Blondie24
7
- Produced by Fogel in 1999-2000
- Neural network as an evaluation function.
- Values for input nodes
Red (Black) – positive
White – negative
Empty – zero
- Piece differential
- Subsections (sub-boards)
9. Blondie24
9
- Initial population of 30 neural networks
(players).
- Each neural network plays 5 games (as red)
against 5 randomly chosen players:-
+1 for a win
0 for a draw
-2 for a loss
-Best 15 players retained, the other 15 players
eliminated.
-Copy the best 15 players (replacing the worst
10. Blondie24
10
- Repeat the process for 840 generations and
the best player after these generations is
retained.
- Played 165 games at zone.com.
- Rating: 2045.85 at that time
- In top 500 of over 120,000 players on
zone.com at that time.
- Better than 99.61% of registered players on
zone.com
11. Blondie24
11
- Fogel received many comments about
Blondie24 design. One of them is concerned
with the piece difference feature and how it
affects the learning process of Blondie24.
Piece-count
Win Draw Lose
Blondie24 12 0 2
Table1: Results of Playing 14 Games between
Blondie24 and Piece-count Using Material
Advantage to Break Tie.
12. Blondie24
12
Piece-count
Win Draw Lose
Blondie24 10 3 1
Table2: Results of Playing 14 Games between Blondie24 and
Piece-count Using Blitz98 to Break Tie.
- It is clear that Blondie24 is significantly better than a
piece-count player, and by using a standard rating
formula, the results suggest that Blondie24 is about
311 to 400 points better than the piece-count player.
13. Brunette24
13
- Designed by Evan Hughes as a re-
implementation of Blondie24.
- Hughes used the same structure that is used
for Blondie24.
- Hughes used the same experiment as Fogel
to show the importance of a piece difference.
14. Brunette24
14
Piece-count
Win Draw Lose
Evolved Piece 680 300 20
Count
Table3: Results of Playing 1000 Games between the Evolved
Piece Count player and Piece-count player.
- By using a standard rating formula, the
results suggest that the evolved piece
difference player is about 528 points better
than piece difference player.
15. Brunette24
15
Xcheckers
Win Draw Lose
Evolved Piece 220 660 120
Count
Table4: Results of Playing 1000 Games between the
Evolved Piece Count player and xcheckers.
- By using a standard rating formula, the
results suggest that the evolved piece
difference player is about 80 points better
than xcheckers.
16. Experimental Setup
16
- Two implementations of Blondie24 were
done, one with a piece difference feature,
which is called Blondie24-RPD, while the
other is without a piece difference feature and
is called Blondie24-R.
- Our previous efforts to enhance Blondie24
introduced a round robin tournament. The
resultant player (Blondie24-RR) is used to
show the importance of the piece difference
feature. This is done by implementing a player
which is the same as Blondie24-RR, but,
17. Experimental Setup
17
- To measure the effect of a piece difference
feature in Blondie24, Blondie24-RPD was
played against Blondie24-R by using the idea
of a two-move ballot.
- The games were played until either one side
wins or a draw is declared after 100 moves
for each player.
- The same procedure was also used to play
Blondie24-RR against Blondie24-RRNPD
18. Results and Discussion
18
Opponent:Blondie24-R
Win Draw Lose
Blondie24- 59 14 13
RPD
Table 5: Results when Playing Blondie24-RPD against
Blondie24-R using the Two-Move Ballot
-By using a standard rating formula, the results
suggest that Blondie24-RPD is about 428
points better than Blondie24-R.
19. Results and Discussion
19
Opponent: Blondie24-RNPD
Win Draw Lose
Blondie24-RR 61 16 9
Table 6: Results when Playing Blondie24-RR against
Blondie24-RRNPD using the Two-Move Ballot
- By using a standard rating formula, the
results suggest that Blondie24-RR is
about 489 points better than Blondie24-
RRNPD.
20. Conclusions
20
- Piece difference feature is important to the
design of Blondie24.
- Neural network is also an important element
of the whole design but the results presented
here demonstrate a simple feature is able to
significantly improve the overall playing
strength.
21. References
21
1- Samuel, A. L., Some studies in machine learning using the game of checkers 1959,1967.
2- Fogel D. B., Blondie24 Playing at the Edge of AI, United States of America Academic Press, 2002.
3- Chellapilla K. and Fogel, D. B., Anaconda defeats hoyle 6-0: A case study competing an evolved
checkers program against commercially available software 2000.
4- Fogel D. B. and Chellapilla K., Verifying anaconda's expert rating by competing against Chinook:
experiments in co-evolving a neural checkers player.
5- Chellapilla K. and Fogel D.B., Evolution, Neural Networks, Games, and Intelligence,” 1999..
6- Chellapilla K. and Fogel D. B., Evolving an expert checkers playing program without using human
expertise.
7- Chellapilla K. and Fogel D. B., Evolving neural networks to play checkers without relying on
expert knowledge.1999.