World Mail Championship Macc Banks Champion v Tommy Canning

[Lindus Edwards]

Hi Tommy,
I am not sure if the Welsh Open Postal Championship is in progress. I did however receive an invitation from Nigel Proffit to play in this event several months ago but declined to play.
Given your fine record in this extremely tough tournament I am surprised that you too did not receive an invitation.
Lindus.

[Dennis Pawlek]

Dear Tommy,

It was always announced in the EDJ. Maybe you should read the news page there sometimes.

[tommyc]

I was talking 2003 Denis before you was even heard of in checkers,and have some manners.

[Dennis Pawlek]

Dear Tommy,

Not my mistake you dont put years and im the current editor... Im in checkers a more long time then you think. Just I were not active.

[tommyc]

Well i thought it was a kinda funny the then current Welsh champion wasnt approached when the Ty resumed...........or was you around then?..

[Dennis Pawlek]

Well it was advertised in the EDJ and im not sure if Nigel did notify every player that there is another tournament coming up. So send your complain to him then. I did play in the 2004 and 2005 Welsh Mail Open.

[tommyc]

Get over it Denis..................and dont annoy me right now...im trying to draw Mac Banks and man short!! Get off the pot!

[Lindus Edwards]

Well done Tommy on making twenty-four draws against Mac. Congratulations to Mac on retaining his world mail championship title. The games were fantastic.

[mday]

Get over it Denis..................and dont annoy me right now...im trying to draw Mac Banks and man short!! Get off the pot!

Lets play some games online, tommyc.

[Martin Fierz]

Two years ago I entered the Welsh Postal Championship naively believing that I was certain to win it being in possession of several powerful computers loaded with the greatest programs in the world.
On the 10-14, 22-18, 12-16 against George Miller I used a program (still rated as the best in the world) and was soundly defeated.

Hi Lindus,

you make me very curious: you were using the program rated as the best in the world... which program is that? and who made the rating??

cheers
Martin

[tommyc]

Thanks for the offer Michael....but i dont do friendlies!! Sorry/.

[Lindus Edwards]

Hi Martin,
I was using Nemesis Pro. I should have used your program as I am sure I would not have lost as many games
Great mail players such as Mac Banks, George Miller, Bill Carter and Tommy Canning have ways of beating programs. Mac divulged a method to me but I will not reveal this here.
Best wishes, and keep up the wonderful work you are doing for checkers as it is so much appreciated.
Lindus.

[Lindus Edwards]

Martin, perhaps I was being presumptuous in saying that Nemesis Pro was the best rated program in the world at the time. I am not sure that this was this case now that you mention the subject.

[Martin Fierz]

Hi Lindus,

let me put this in perspective - my perspective, of course

In 2002, Nemesis won the computer world championship in Las Vegas. It won by a very narrow margin, finishing +1=47, Kingsrow finished +1-1=46, and Cake finished +1-2=45.
All programs were using the 8-piece endgame database, and big opening books, with Cake's book being by far the smallest. Murray had built the 8pc db some time ago himself, I had built it quite recently, and Ed was relying on the newly released Chinook 8pc db - which was provided without access code (you had to adapt the old code for the 6pc db). Kingsrow ended up losing a game against Cake because Ed hadn't adapted everything properly. Naturally he must have been extremely frustrated, but he never made any lame excuses after the championship. So let me say this in his place: I am convinced that without this bug, Kingsrow was just as good as Nemesis in 2002. Ed fixed the bug after the championship and even though he came second, he now had a program that was every bit as good as Nemesis, only somehow people didn't realize (or so it seems to me).
Cake was clearly inferior, losing two games more or less out of the opening, but also, in one of the games, not defending a difficult position properly. I was also frustrated, and made some very big improvements to Cake after Las Vegas (producing the "Sans Souci" version). After that, I believe that Cake was (and remained so until now - although Ed's latest Kingsrow version has nearly closed the gap) the best cross-board engine, but it still had by far the weakest book of the 3.
As a side note, I never re-examined the games of the Vegas championship to see whether anything was missed there - perhaps Ed could shed some light on this with his 10pc database.

So: if you had played with Cake in 2004, you would still have ended up losing games, because it had a small book with probably quite a few errors. On the other hand, Murray told me in Vegas that the commercial version of Nemesis didn't have his full book either. I don't think he ever changed that, but only he would know. If this is the case, then from late 2002 on, your best choice would have been Kingsrow, since Ed did publish his full opening book. This april, I finally released my full opening book, and now you can choose either Cake or Kingsrow - they are both nearly perfect. Both of them have huge improvements in the engine itself, making the much stronger crossboard than the versions which played in Vegas 2002, and both of them have a much larger book, generated and recomputed with the improved engines.

The bottom line is of course that things change quickly in computer checkers, and whatever was true a couple of years ago my not be true today - and what is true today may be obsolete in a year. But people still believe that Chinook is world champion, or that Nemesis is the best program, and that the earth is flat

As for mailplay, I have very little experience with this form of checkers. I played on the Banks ladder with Cake in 2001/2002, still with a 6-piece database and a really awful book, and I didn't do well. However, as far as I remember, Ed was doing very well on this ladder with Kingsrow, losing infrequently - and this was with a pre-Vegas version with only a 6-piece database and a book that was generated based on this old engine. To me, it seems very unlikely that a match between Cake or Kingsrow and the best mail players would end with something different than 24 draws. I do know of course that the programs are not invincible - even in my latest book I know of a losing move. It's not easy to find though! Perhaps Kingsrow-10pc is invincible, but of course that is not available to the public.

This post turned out longish, but I'm afraid that there is too much to say on this subject to just make a short post! Even with this long post, I only managed to tell a small part of the computer checkers story.

cheers
Martin

[Ed Gilbert]

Hi Martin,

I think your perspective on the state of checkers programs over the past few years is very similar to my own. I did want to comment on a couple of things:

Ed was relying on the newly released Chinook 8pc db - which was provided without access code (you had to adapt the old code for the 6pc db).

Actually Jonathan Schaeffer had provided 8pc access code with the db. There was a math overflow bug in the indexing code. When I identified and fixed the bug after the tournament, I sent my fixes to Jonathan, who eventually incorporated them into the access code he had posted on his web site. Of course it was my own fault for not testing this thoroughly enough. The bug was very difficult to find, and it only showed up in an extremely small number of positions. It had to be an 8pc position with all checkers (no kings), and the leading ranks of both the black and white men had to be one of the two highest ranks, the two rows just before a man becomes a king.

I never re-examined the games of the Vegas championship to see whether anything was missed there - perhaps Ed could shed some light on this with his 10pc database.

I have not done this. But I believe these games have been gone over by a number of people, including the 3 participants, and I'm not aware of any missed wins.

-- Ed