The 2nd Derek Oldbury Memorial 'Man v Machine' Open
"Annotated games from the tournament"

     This was a 7 round freestyle quick play tournament held at Bristol on December 9th 1995 organized by Richard Pask.  Fourteen players took part including the three top PC computer programs, Cornell Checkers (Gil Dodgen), Colossus (Martin Bryant) and Sage (Adrian Millett).

    The results were as follows, 1st Cornell Checkers 13/14 pts, 2nd Sage 11/14, 3rd Colossus 10/14, 4th Derek Oldbury  9/14. The computers dominated the event as was to be expected with unfortunately none of the top British players present.

    This year Cornell Checkers and Colossus ran on 133MHz Pentiums and Sage on a 90MHz Pentium. The tournament was sponsored by the BDF and the EDA.

    Presented are all three of the computer-computer games. Colossus drew with both Cornell and Sage, whilst Cornell beat Sage.


Round 2
Black:Colossus White:Sage

11-15 24-19 15x24 27x20!?(a) 9-14 22-18 5-9 28-24 8-11 25-22 10-15 23-19 14x23 19x10 7x14 26x19 4-8 22-17 6-10 17-13 11-15 13x6 2x9 31-26 15-18 32-28 1-6 30-25 18-23 26-22(b) 8-11 19-16 12x19 24x8 3x12 22-17 23-26 28-24 26-30 24-19 14-18 17-14 10x17 21x5 30x21 19-15 6-9 5-1 18-23 1-6 9-14 6-9 14-17 9-14 17-22 14-18 23-27 18x25 21x30 15-11 30-26 11-7 26-23 29-25 27-32 25-21 32-27 7-3 23-18 21-17 27-32 3-8 32-28 8-3 28-24 17-13 18-14 3-8 24-28 8-3 28-32 3-8 32-28 8-3 28-24 3-7 24-19 7-11 19-23 11-15 23-26 15-19 26-30 20-16 14-18 16-11 18-23(c) Draw

(a) I was suspicious the moment I saw this! Why should Adrian have Sage play a known inferior line?
(b) Adrian had copied this line straight out of Colossus' book into Sage's thus ensuring that a fairly safe
late midgame position was reached. Although I understand Adrian's motives for doing this it's just
not cricket! (Ed: I wonder what our US readers will make of that last comment!) The rest of the game
Colossus tried to make something from nothing but to no avail. I guess I'm going to have to put some
serious thought into how to avoid this scenario again against well prepared opponents.
(c) With Colossus having the 6-piece endgame databases (and Sage only the 5-piece) it was worth playing it out to see if Sage would falter.


Round 4
Black:Cornell Checkers White:Sage

9-14 22-17 11-15 23-19 7-11(a) 25-22 11-16 27-23?!(b) 8-11(c) 17-13 5-9 29-25 16-20 19-16 12x19!?(d) 23x7 20x27 31x24 2x11 32-27?(e) 4-8 24-20 3-7 26-23 1-5 22-17 14-18 23x14 9x18 27-24 18-23 24-19 15x24 28x19 23-27 25-22 27-31 22-18 31-27 30-26 8-12 17-14 10x17 21x14 27-31 26-23 6-10 13-9 10x17 19-15 5x14 15x8 31-27 18x9 27x18 9-6 18-15 6-2 7-11 20-16 11x20 Black Win

(a) Cornell chooses the Whilter opening.
(b) Weaker than the usual 26-23 as in Lafferty-Davis '84 game 5, Long-Banks '34 game 11 and Tinsley-Chinook '94 game 6.
(c) Lees' Guide and BDP give 5-9 first.
(d) Varying from 20x27 in Lees' and BDP and perfectly playable.
(e) Seems to lose. Instead 24-19 should survive.


Round 5
Black:Colossus White:Cornell Checkers

11-15 23-19 9-14 22-17 6-9 17-13 2-6(a) 25-22 8-11 29-25 4-8 24-20 15x24 28x19 11-15 27-24 14-17 21x14 9x18 26-23 18x27 32x23 10-14 19x10 6x15 13-9 14-17 22x13 5x14 25-22 1-6 23-19 7-10 30-25 14-17 25-21 17x26 31x22 8-11 21-17 3-7 17-14 10x26 19x3 26-30 3-8(b) 11-15 8-11 15-18 24-19 30-26 19-15 26-23 11-8 18-22 15-11 6-10 11-7 22-26 7-2 26-31 2-7 10-14 8-11 14-17 Draw

(a) The Souter opening which against human opposition can be quite profitable as it is not so well known and has several tricky positions. However against a strong well prepared opponent it should only ever be a draw. Unfortunately I had no spare time prior to the tournament to 'prepare' Colossus, i.e. prime it to play only 'fighting' openings. This led to Colossus' relatively poor showing by allowing several draws (even a painfully dull Glasgow against George Miller) whereas it would normally be set to scrap it out with anyone.
(b) After the tournament Richard Pask asked why Colossus preferred 3-8 here (which Cornell had also predicted) rather than the pp 3-7. Well either move draws and once a computer can see a draw it has no concept of a risky looking draw over a safe looking draw and as the resultant position was evaluated as slightly better than after 3-7 it chose the alternative 3-8.