Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

Genius' asymmetric-search by example: TRY yourself

139 views
Skip to first unread message

mclane

unread,
Dec 30, 1997, 3:00:00 AM12/30/97
to


Posted by Thorsten Czub on December 29, 1997 at 18:04:54:

Please, my analysis have been done on a k6/200 with giving Genius5
3968
KByte for hash.
For point 1 use pgn1, for point2 use pgn2:

[Event "PGN1"]
[Site "?"]
[Date "????.??.??"]
[Round "?"]
[White "Karpov"]
[Black "Topalov"]
[Result "1-0"]
[SetUp "1"]
[FEN "rqr3k1/3Qbp2/p1n1p1p1/1pp5/2P2P2/2N3P1/PP3PB1/R3R1K1 w - -"]

1. *


[Event "PGN2"]
[Site "?"]
[Date "????.??.??"]
[Round "?"]
[White "Karpov"]
[Black "Topalov"]
[Result "*"]
[SetUp "1"]
[FEN "rq3rk1/3Qbp2/p1n1p1p1/1pp5/2P2P2/2N3P1/PP3PB1/R3R1K1 b - -"]

1. *


1.
Set up the position PGN1 in Genius5/WIN (import it as PGN or setup
manually).
Put level on infinite.
Let Genius5 compute on the whites move.
It considers Bxc6 in the 8/20 search.
Changes to cxb5 in this search.
It runs through the 9/21 search without changes until (arround 51' !!)
it computes in the 34th
branch (that is Rxe6), but did not see the point and starts the next
iteration (10/22) with main-line
(since 18'58") cxb5 Ra7 Qd3 Nd4 bxa6 Qxb2 Rab1 +0.60.


2.
Now set up again the position but take the rook from c to f8 !
Or import PGN2, that is the old position but with the rook-move taken
back.
Let Genius5 compute about THIS position that is ONE ply before the
unique key-move Rxe6 with
following Rxg6.
After 1'23 in only search 8/20 Genius5 finds the key move Rxe6 , but
here of course as the 2nd
ply of the main-line !
The main line is:
Rc8 Rxe6 Ra7 Rxg6+ fxg6 Qxe6+ Kg7 Bxc6 with an evaluation of -1.15.
It will still play Rc8 in the 9/21 search with -1.15 evaluation.

Conclusions:
This is what I call the asymmetric-search. It is the reason Genius5 is
that strong and plays that
"Levy"-like ("do nothing, but do it well").
The reason it finds the key moves FASTER although you go back (!!) one
ply shows that it
computes more brute-force (less pruning) in the plies 2,4,6,8,... !
The reason it has not found Rxe6 in the position with the rook on c8
is,
that is had to find the key
move with the plies that prune much and have pruned it away !!
ALL Lang program have this search-technique.
From the first dedicated machine Mephisto Amsterdam to the latest
Genius5 engine.
Lang has changed the knowledge and the tactical abilities, but never
has
he changed the search-
strategy.

I will further (when I have some time) show you a position where
Genius5
decides to choose the
boring (also weaker) moves/main-lines with having 1,3,5,7...
plies-search, and if you take one
ply back, it plays for the same color better, more active moves that
would lead to a more
powerful game. It IS the search that makes genius play "boring" since
1985 Amsterdam-module.


Thanks.

P.S.: Take care about those effects when analysing your own games with
Genius5. Especially
your email-chess-games ! You cannot trust Genius' analysis even if you
let it compute 12 hours
per move. The effect seen in the position above will not disappear
whatever time you give it. All
your analysis time invested could be WASTED because the FIRST move in
the main-line is NOT
the best one in the position you analyse.
Never trust genius5's 1,3,5,7... iteration-moves from the main-line !
They could be completely
overseeing STRONG moves.
Use Genius as an instrument. But use it not blind trusting. Remember
the
asymmetric search
and take care of it.

I remember a very old "argument" with Thomas Mally or Andreas Mader
(both writers) from PC-
SCHACH/MODUL in the past. He said that he cannot understand my
sceptical
position
concerning using the dedicated mephisto-machines (with Lang program)
for
analysis. In fact it
was the Mephisto Lyon or Almeria or something like this.
He told me that he even bought a second machine for analysis !!
I was very shocked. I still remember this episode, although it is
years
ago.
Sitting with Genius5 about the above position, having much faster
hardware than the old
Motorola 68000 or 68020 CPU, I am still convinced that my decision NOT
to trust those machines
due to their lack of "accuracy" that results from the highly contrast
between the 2 pruning-levels
was reasonable.
If a program is not able to find such an important move although it
runs
on such a fast machine,
how much would it would have overseen on a (how many times?) slower
Motorola 68020 ??!!


For more information and lots of other interesting INSIGHTS in
computerchess, also about those "secrets", share the
Computer-Chess-Club with us at http://www.icdchess.com under CCC.

Thanks ! All the best this year and best wishes for next year.

Komputer Korner

unread,
Dec 30, 1997, 3:00:00 AM12/30/97
to

And I was once castigated very thoroughly on this very newsgroup 18 months
ago for decrying the devil of ASYMMETRY!!!!!!!!!!!
Now it turns out that Bob Hyatt himself is getting rid of the King safety
asymmetry in Crafty and calling it the Komputer Korner version!!!!!!!!!!!.
--
- -
Komputer Korner

The inkompetent komputer

If you see a 1 in my email address, take it out before replying.
Please do not email both me and the r.g.c.c. at the same time. I read all
the postings on r.g.c.c.
Also every statement of mine should be taken with a grain of salt. Read at
your own risk and
assume that it is only this humble komputer's opinion.

Elisabeth van Aaring <Use-Author-Address-Header@[127.1]> wrote in article
<1997123010154...@nym.alias.net>...
> Hi Thorsten, thanks for this example.
> >
>
> Interesting (but I suppose not new) observations!
>
> Your observation may also explain, why Genius seems to have problems with

> weaker programs. Take a 2000 Elo program, it will be terribly destroyed
by
> a 2200 Elo program. Then let it play Genius. Suddenly, you think your
2000
> Elo program is much stronger, still alive at move 50! What is happening?
> Playing the 2200 Elo program it never survived move 40!!!
>
>
> In case of a "surprisingly low" value at the root Genius relaxes his
> pruning, takes some time, and, voila, the score goes up again. Then the
> pruning is switched on again, checking for coming threats by the
opponent.
>
> In conclusion it is possible to win the championship just by increasing
the
> evaluation by "only" +0.10 each move, that is a knight at move 30...
> Nothing for TV, I agree, but highly efficient. And BTW, things are not
that
> easy...
>
> For example: how does Genius know, when an evaluation is "suprisingly
low"?
>
> However, I will try this approach with my own program, hoping to increase

> the search depth by 4 ply... :-)
>
> Elisabeth
>
>
>

Lonnie

unread,
Dec 30, 1997, 3:00:00 AM12/30/97
to

On 30 Dec 1997 12:35:48 EST, "Komputer Korner" <kor...@netcom.ca> wrote:

|And I was once castigated very thoroughly on this very newsgroup 18 months
|ago for decrying the devil of ASYMMETRY!!!!!!!!!!!
|Now it turns out that Bob Hyatt himself is getting rid of the King safety
|asymmetry in Crafty and calling it the Komputer Korner version!!!!!!!!!!!.

| I remember that 1,heehee. The minute I saw the title of the thread and a response
to it i didn't even have to look,I'd knew it'd b KK :>)
같같같
Lonnie

This is like deja-vu, all over again!

**Put a "c" before the "ook" to respond to proper email address**

Robert Hyatt

unread,
Dec 30, 1997, 3:00:00 AM12/30/97
to

Komputer Korner <kor...@netcom.ca> wrote:
: And I was once castigated very thoroughly on this very newsgroup 18 months
: ago for decrying the devil of ASYMMETRY!!!!!!!!!!!
: Now it turns out that Bob Hyatt himself is getting rid of the King safety
: asymmetry in Crafty and calling it the Komputer Korner version!!!!!!!!!!!.

yep. But I have *never* done an asymmetric search. Don't plan on doing
so either. And eventually plan on no asymmetry at all, once I figure out
how to recognize when one side has an advantage in a blocked-pawn position.

:)

: --
: - -
: Komputer Korner


mclane

unread,
Dec 31, 1997, 3:00:00 AM12/31/97
to

Elisabeth van Aaring <Use-Author-Address-Header@[127.1]> wrote:

>Hi Thorsten, thanks for this example.

A pleasure...


>Interesting (but I suppose not new) observations!

As old as a mephisto Amsterdam (=1985).

>Your observation may also explain, why Genius seems to have problems with
>weaker programs. Take a 2000 Elo program, it will be terribly destroyed by
>a 2200 Elo program. Then let it play Genius. Suddenly, you think your 2000
>Elo program is much stronger, still alive at move 50! What is happening?
>Playing the 2200 Elo program it never survived move 40!!!

Because Genius is not doing much !!

>To me it seems Genius has more problems with himself than with his
>opponent.

A very good sentence !!
Each strategy has its advantages and its disadvantages !
The two point of views fight against each other and have a major
effect on the playing style.

>He realizes all the problems at ply 2,4,6,8,10... because of
>pruning away many "solutions" at 1,3,5,7...
>The result is a "minimal" approach. Genius examines some of his moves but
>all of the opponents' moves. Therefore the evaluation at the root is a
>pessimistic value!

Exactly !

>If this pessimistic value is "enough" (e.g. increasing
>the evaluation by say +0.32), Genius can be quite sure that the opponent
>(doing a symmetric search) cannot outsearch him at the next ply. Genius
>already has seen all the opponents' moves because of his deeper searching
>by heavy pruning.

Right.


>In case of a "surprisingly low" value at the root Genius relaxes his
>pruning, takes some time, and, voila, the score goes up again. Then the
>pruning is switched on again, checking for coming threats by the opponent.

>In conclusion it is possible to win the championship just by increasing the
>evaluation by "only" +0.10 each move, that is a knight at move 30...
>Nothing for TV, I agree, but highly efficient. And BTW, things are not that
>easy...

And genius is a very good defender and plays these "positional"
knight-back-to-the-1rst-rank-moves.


>For example: how does Genius know, when an evaluation is "suprisingly low"?

>However, I will try this approach with my own program, hoping to increase
>the search depth by 4 ply... :-)

I am looking forward to your results, especially HOW this effects the
playing style concerning boring and active play.
Will your program change the style althoug you have NOT changed the
evaluations ??


>Elisabeth


Tell me when you have results.


mclane

unread,
Dec 31, 1997, 3:00:00 AM12/31/97
to

Elisabeth van Aaring <Use-Author-Address-Header@[127.1]> wrote:
please email me your results or different game-scores !!
My email is
mcl...@prima.ruhr.de. Thanks.


Jim Caccamise

unread,
Jan 1, 1998, 3:00:00 AM1/1/98
to


Robert Hyatt <hy...@cis.uab.edu> wrote in article
<68bvol$89q$1...@juniper.cis.uab.edu>...

I thought you claimed asymmetric search because your move ordering
differs from White and Black sides. I still think the best way to
eliminate search asymmetry is to always evaluate from the same side by
flipping the board when analyzing for Black. This also permits sharing
of cached position information in symmetrical positions. (Although
separate repetition tags need to be maintained for White and Black on
move.)
--
Jim Caccamise
(Remove X from e-mail address, anti-spam)

Robert Hyatt

unread,
Jan 1, 1998, 3:00:00 AM1/1/98
to

Jim Caccamise <Xc...@magicnet.net> wrote:

: I thought you claimed asymmetric search because your move ordering


: differs from White and Black sides. I still think the best way to
: eliminate search asymmetry is to always evaluate from the same side by
: flipping the board when analyzing for Black. This also permits sharing
: of cached position information in symmetrical positions. (Although
: separate repetition tags need to be maintained for White and Black on
: move.)

No... I have an asymmetric evaluation. But if you look at searchr.c,
search.c and quiesce.c, you'll notice *no* special treatment for plies
one, three, ..., N (N odd) vs plies two, four, ..., M (M even). The
only special thing I do is that for ply = 1, I have a special search
function (searchr.c) which does not try null moves (it is illegal to
"pass" at the root since it *must* choose a legal chess move) and does
not extend if in check, since this would extend *every* move and not
accomplish anything...

But the evaluation is a different story, where there is an asymmetric
penalty for blocked pawns, although this is now the *only* asymmetry
that is left, and it will go away eventually (I hope)...

--
Robert Hyatt Computer and Information Sciences
hy...@cis.uab.edu University of Alabama at Birmingham
(205) 934-2213 115A Campbell Hall, UAB Station
(205) 934-5473 FAX Birmingham, AL 35294-1170

Tord Kallqvist Romstad

unread,
Jan 2, 1998, 3:00:00 AM1/2/98
to

In article <68e8aj$cb7$1...@steve.prima.ruhr.de>, mclane wrote:
>I am looking forward to your results, especially HOW this effects the
>playing style concerning boring and active play.
>Will your program change the style althoug you have NOT changed the
>evaluations ??

I have experimented a lot with asymmetric search in my program Marvin. In
fact Marvin has two different search functions, one for odd plies and one
for even plies. The search functions call each other using mutual recursion.
The point of this is to facitlitate experimentation with asymmetric search.

My experience is that the playing style _does_ change, even when the
evaluation function is the same. Pruning lots of moves at even plies and few
moves at odd plies gives a passive, Genius-like style of play. Unfortunately
I do not get Genius-like playing strength. :-(

The opposite approach resulted in a much more interesting style of play,
but also in too many tactical oversights. At the moment, my search and eval
is almost completely symmetric.

Tord

Komputer Korner

unread,
Jan 2, 1998, 3:00:00 AM1/2/98
to

What you are referring to is the non mirroring that results because of
using bitboards. Crafty is one of the few that uses bitboards and it is
impossible to get mirrored ( black and white evaluations the same) results
using these. Bob can further explain it to you.
--
- -
Komputer Korner

The inkompetent komputer

If you see a 1 in my email address, take it out before replying.
Please do not email both me and the r.g.c.c. at the same time. I read all
the postings on r.g.c.c.
Also every statement of mine should be taken with a grain of salt. Read at
your own risk and
assume that it is only this humble komputer's opinion.

Jim Caccamise <Xc...@magicnet.net> wrote in article
<01bd16da$0d4a59e0$8aca06d0@jimc>...


>
> I thought you claimed asymmetric search because your move ordering
> differs from White and Black sides. I still think the best way to
> eliminate search asymmetry is to always evaluate from the same side by
> flipping the board when analyzing for Black. This also permits sharing
> of cached position information in symmetrical positions. (Although
> separate repetition tags need to be maintained for White and Black on
> move.)

Robert Hyatt

unread,
Jan 2, 1998, 3:00:00 AM1/2/98
to

Komputer Korner <kor...@netcom.ca> wrote:
: What you are referring to is the non mirroring that results because of

: using bitboards. Crafty is one of the few that uses bitboards and it is
: impossible to get mirrored ( black and white evaluations the same) results
: using these. Bob can further explain it to you.

It's not a question of mirrored evaluations, it is a question of
move ordering. IE if you search the same position twice, but have it
set up once with wtm, and then mirror it with btm, the scores will be
(usually) identical, but the node counts will vary somewhat because of
the move ordering differences...


Jim Caccamise

unread,
Jan 2, 1998, 3:00:00 AM1/2/98
to

Robert Hyatt <hy...@cis.uab.edu> wrote in article
<68jojv$it0$1...@juniper.cis.uab.edu>...

This is exactly what I was referring to, "asymmetric search because
your move ordering differs from White and Black sides." ( Quoted from
first sentence of my previous post.) You previously claimed that move
ordering differences from White and Black sides result in different cache
hits and tree evaluations. I was confirming that you still have some
asymmetry (perhaps this is differentiated with the term non-mirroring).
To eliminate all asymmetry you need to duplicate White and Black side
move ordering. I was suggesting the "simple" method of eliminating all
White and Black side asymmetry by evaluating a flipped board for Black. I
realize there is some overhead in doing this, but there are also some
advantages and possibly some clever tricks to implement this efficiently.

Robert Hyatt

unread,
Jan 3, 1998, 3:00:00 AM1/3/98
to

Jim Caccamise <Xc...@magicnet.net> wrote:

: This is exactly what I was referring to, "asymmetric search because


: your move ordering differs from White and Black sides." ( Quoted from
: first sentence of my previous post.) You previously claimed that move
: ordering differences from White and Black sides result in different cache
: hits and tree evaluations. I was confirming that you still have some
: asymmetry (perhaps this is differentiated with the term non-mirroring).
: To eliminate all asymmetry you need to duplicate White and Black side
: move ordering. I was suggesting the "simple" method of eliminating all
: White and Black side asymmetry by evaluating a flipped board for Black. I
: realize there is some overhead in doing this, but there are also some
: advantages and possibly some clever tricks to implement this efficiently.
:
: --
: Jim Caccamise
: (Remove X from e-mail address, anti-spam)

In that case, you are right. I was thinking of "asymmetry" in terms of
doing different things at different plies, *intentionally*, as in genius.
The different move ordering would be extremely difficult to overcome in
bitmaps...

Jim Caccamise

unread,
Jan 3, 1998, 3:00:00 AM1/3/98
to

Robert Hyatt <hy...@cis.uab.edu> wrote in article

<68k9af$oan$1...@juniper.cis.uab.edu>...

The way you use bitboards makes evaluating a flipped board "costly",
because many flipped bitboards are needed. But, perhaps you can fix move
ordering asymmetry without much "cost". For the following, I assume move
lists are ordered by destination squares. Simply processing the
generated moves backward for Black corrects rank ordering, but doesn't
compensate for the initial Black and White positions being "mirrored"
instead of "rotated". (Still this does improve move ordering similarity.)
"Mirrored" move ordering would also require processing moves backwards
within the ranks while processing the list backwards. A simpler way is
to generate 8 move sub-lists, one for each file, instead of one combined
list. For "mirrored" symmetry the rank sub-lists can be processed in 1-8
order for White, and in 8-1 order for Black. (You could experiment with
processing the rank sub-lists in different orders, as long as the order
is reversed for White and Black to maintain "mirrored" symmetry. Perhaps
processing moves in the center ranks first to take advantage of the
principal of controlling the center, and/or processing moves to distant
ranks first to emphasize aggressive play. This rank ordering could be
adjusted dynamically depending on position characteristics.)
I'm not sure how you generate and maintain move lists, but it seems
that move ordering symmetry is attainable without much "cost".

Robert Hyatt

unread,
Jan 3, 1998, 3:00:00 AM1/3/98
to

Jim Caccamise <Xc...@magicnet.net> wrote:


: The way you use bitboards makes evaluating a flipped board "costly",


: because many flipped bitboards are needed.

I'm not sure what you mean here. Evaluation isn't a problem at all, because
the "order" of static things doesn't matter at all. IE whether I evaluate the
h-pawn first or last doesn't effect the evaluation outcome at all.

: But, perhaps you can fix move


: ordering asymmetry without much "cost". For the following, I assume move
: lists are ordered by destination squares. Simply processing the
: generated moves backward for Black corrects rank ordering, but doesn't
: compensate for the initial Black and White positions being "mirrored"
: instead of "rotated". (Still this does improve move ordering similarity.)

that would be most difficult.. because I use either "FirstOne()" or
"LastOne()" to pick off a bit, and these are not symmetric for both
colors. Since I use this to generate moves, the ordering is going to
be different. I *could* (for black) pick off bits and shove 'em onto
a stack, and then pop to generate, but that would be expensive, and
wouldn't affect things at all, other than to slow move generation down
a significant amount.

: "Mirrored" move ordering would also require processing moves backwards


: within the ranks while processing the list backwards. A simpler way is
: to generate 8 move sub-lists, one for each file, instead of one combined
: list. For "mirrored" symmetry the rank sub-lists can be processed in 1-8
: order for White, and in 8-1 order for Black. (You could experiment with
: processing the rank sub-lists in different orders, as long as the order
: is reversed for White and Black to maintain "mirrored" symmetry. Perhaps
: processing moves in the center ranks first to take advantage of the
: principal of controlling the center, and/or processing moves to distant
: ranks first to emphasize aggressive play. This rank ordering could be
: adjusted dynamically depending on position characteristics.)
: I'm not sure how you generate and maintain move lists, but it seems
: that move ordering symmetry is attainable without much "cost".
: --

the term "rotated" really doesn't apply here... those bitmaps are used
only to detect the status of a rank/file/diagonal for sliding pieces, and
doesn't affect move generation at all. But "mirroring along the 4th-5th
ranks does change things a little, but only in move ordering for the search,
which only affects the node counts very slightly, except for the rare case
where the transposition table lets this ordering variance affect the final
score somehow...

But it would cost... maybe 10% if not more, for really no gain anywhere
else...

: Jim Caccamise
: (Remove X from e-mail address, anti-spam)

mclane

unread,
Jan 3, 1998, 3:00:00 AM1/3/98
to

rom...@filoktetes.uio.no (Tord Kallqvist Romstad) wrote:

>>Will your program change the style althoug you have NOT changed the
>>evaluations ??

>I have experimented a lot with asymmetric search in my program Marvin. In
>fact Marvin has two different search functions, one for odd plies and one
>for even plies. The search functions call each other using mutual recursion.
>The point of this is to facitlitate experimentation with asymmetric search.

>My experience is that the playing style _does_ change, even when the
>evaluation function is the same. Pruning lots of moves at even plies and few
>moves at odd plies gives a passive, Genius-like style of play. Unfortunately
>I do not get Genius-like playing strength. :-(

I think the asymmetric search is ONE thing.
I have not discussed about the extensions. I think genius extends
other way than other programs do.
As Don (Daily) said in CCC, Genius is normally a good tactician. This
is absolutely right. But we should not forget that Genius was not
always that good in tactics. In the old dedicated-computer times, the
mephisto machines had a hardware advantage, a deeper search AND this
resulted in a better strength. But genius was NOT a tactics-killer.
Over the years Mephisto programs had to compete with Fritz and other
programs. Weiner/Lang was forced to give Genius some extra-extensions
to stand fritz in blitz and to solve all the test-suites as good as
the others.

>The opposite approach resulted in a much more interesting style of play,
>but also in too many tactical oversights. At the moment, my search and eval
>is almost completely symmetric.

Thanks for your personal experience Tord.


>Tord

mclane

unread,
Jan 5, 1998, 3:00:00 AM1/5/98
to

>Your answers to my comments tell me that we have the same theoretical point
>of view. The first thing I will try to do is a simple forward pruning.

We have (had) a discussion about this in CCC ( the Computer Chess
Club) and Don Daily said, he believes Genius has only asymetric
extensions but a full-width-base.
This could also be an explanation.
In fact the difference between a full-asymetric search and only a
asymetric extensions is not much.
The difference would appear after a while when the full-width-search
would have reached the ply the extensions pruned the branch away and
FINDS the move !
Since the selectiveness in genius5 is fixed to 12 in default, it could
be a long time to wait if you HAVE to wait until full-width reaches
it.
I still have problems to believe it, because I did my test with
Genius5 with selective sero and STILL it found Rxe6 faster in ONE PLY
BACK.
In the end it could be all 3 possibilities or a mix of all:
1.complete asymetric search
2.full-width + asymetric search
3.asymetric evaluations.

Whatever it is, it is worth a try !! Therefore I am interested in
seeing a game in both styles or playing myself with the 2
engines/programs (program A without asymmetrie, program B with it).


>On ply (1?), 3, 5, 7 etc. I will allow only some moves to be played.
>Perhaps move 1 to 4 according to the movelist. Or I will apply a more
>dynamic approach: ply 1 no pruning, ply 3 50%, ply 5 75%, ply 7 90% of all
>legal moves.

I understand. It is very risky to do it linear instead of tapered.

>This technique seems to be very easy to implement. However,
>testing takes a long time to get significant or meaningful differences.

I would help you with my k6/200 Mhz autoplayer pair. I always wanted
to get 2 genius versions (asymmetrie Genius and symmetrie Genius) and
was never able to prove my ideas. Therefore I would like to help YOU
with testing. Maybe this helps me also to find out about genius.

please contact me on email if you are ready for testing.
I will send you all the games and comment them.


>Is there anyone who already tried this? Results? Modifications?

>Elisabeth

I hope you also take a look into CCC.


Elvis PCD

unread,
Jan 5, 1998, 3:00:00 AM1/5/98
to

Im Artikel <689avq$r82$1...@steve.prima.ruhr.de>, mcl...@prima.ruhr.de (mclane)
schreibt:

>Thema: Genius' asymmetric-search by example: TRY yourself
>Von: mcl...@prima.ruhr.de (mclane)
>Datum: Tue, 30 Dec 1997 00:19:08 GMT

>Posted by Thorsten Czub on December 29, 1997 at 18:04:54:
>
>Please, my analysis have been done on a k6/200 with giving Genius5


....and much more......

I have been watching this discussion, you set up. in CCC + now in RGCC.
Gladly it's fully within our hobby.......
what made me sneak into this - somehow - academic topic, is, that obviously
there is an important fact missing in the whole discussion :

>powerful game. It IS the search that makes genius play "boring" since
>1985 Amsterdam-module.

So, name the gameplay boring, call the search asymertic, symetric, blach +
white
cow - however you perfere - but :
when did you look into the SSDF list last time ?
And if you did, why couldn't your postings reflect - at least one single time,
that this
asymetric, boring, ply missing + whatsoever program is with all it's 3 last
engines is
among the 1st 13.....at present - 3 versions tested only on 90 MHZ.
And still (!) overplay brandnew programs - with, let's call them..." newer...
books....".
How come ??

>P.S.: Take care about those effects when analysing your own games with
>Genius5. Especially
>your email-chess-games ! You cannot trust Genius' analysis even if you
>let it compute 12 hours
>per move. The effect seen in the position above will not disappear
>whatever time you give it. All
>your analysis time invested could be WASTED because the FIRST move in
>the main-line is NOT
>the best one in the position you analyse.
>Never trust genius5's 1,3,5,7... iteration-moves from the main-line !
>They could be completely
>overseeing STRONG moves.
>Use Genius as an instrument. But use it not blind trusting. Remember
>the
>asymmetric search
>and take care of it.

In complete opposition to this opinion - I can state with the experience of
nearly 20
years of constant playing Corr.Chess :
Firstly - if one thinks, he can use the " tool " as playing instead of him - he
would
never even survive the so called " Hauptturnierklasse " - which could be named
main tournament class.
Not to mention the International Master Class, IM-title tournaments, WM
semi-finals
or such at all.
Within there you could use a P II 300 - 48 hours analysis per move - and would
be
crushed to bits + pieces - that's reality !
Secondly :
is the conclusion, you offered - to use it as a TOOL - within and along with
your own
chess possibilities -
but .
I can't understand, really in no way, that someone with your knowledge seems to
miss the functions, this product offers ! ( See below )

Without being a big fan of Genius at all - I can testify, that there is - up to
today -
no available program, that gives you a so correct sight of the board situation,
like
G 3-5, IF you use it RIGHT !
Which simply means - that you can start with the normal analysis - let's say
for
12 hours; write down the analysis....and then start anew - but with : ZERO
selective search - which is near to brute force.

Annother 12 hours - write down, and then look at your own written analysis +
the
2, you got by my example.
Then you can start to build this into your own strategy...if it fits.....and
because of
the - still - extremely well overall balance of this product, the benefit then
is obvious.

With this long explanation I just tried to give you an example out of reality -
because
we never spoke before about the league, in which you personally play chess.


>
>I remember a very old "argument" with Thomas Mally or Andreas Mader
>(both writers) from PC-
>SCHACH/MODUL in the past. He said that he cannot understand my
>sceptical
>position
>concerning using the dedicated mephisto-machines (with Lang program)
>for
>analysis. In fact it
>was the Mephisto Lyon or Almeria or something like this.
>He told me that he even bought a second machine for analysis !!
>I was very shocked. I still remember this episode, although it is

I still have this MODUL - solution : see above....

>Sitting with Genius5 about the above position, having much faster
>hardware than the old
>Motorola 68000 or 68020 CPU, I am still convinced that my decision NOT
>to trust those machines
>due to their lack of "accuracy" that results from the highly contrast
>between the 2 pruning-levels
>was reasonable.
>If a program is not able to find such an important move although it
>runs
>on such a fast machine,
>how much would it would have overseen on a (how many times?) slower
>Motorola 68020 ??!!

Again - even if it might have fit to that time, because of the poorer adjusting
possibilities :
would you tell the readers of all this - WHO was leading the lists constantly
at that
time - and not just because of hardware.....?

>For more information and lots of other interesting INSIGHTS in
>computerchess, also about those "secrets", share the
>Computer-Chess-Club with us at http://www.icdchess.com under CCC.

Yes - and as a member of this group, as well - I take my right to express my
personal opinion concerning this, also :
This...and the other NG - are just, what the people make out of this.
Many of them just left because of....well....we all know....
BUT :
I've seen with my very eyes, that some - not all - of the so called elegibles,
who left
this NG - continue constantly throwing mud against eachother - as if the front
page with all it's well meant rules wasn't there at all.
Could you agree ? - Or should I give proof, here ?
Interesting people post here, and very " good " ones over there, as well.
Pityfully, " names " soacked up many readers towards CCC like a sponge -
without
a real reason !
So, I repeat myself, if you, we all + me.....set up threads like this....maybe
sometime RGCC will grow again to it's once meant purpose...elsewise there'll be
soon - I'm quite sure 'bout that - a 3rd group - only for....VERY elegibles.

Keep on rockin'

Elvi...@aol.com
elvi...@owl-online.de

Vincent Diepeveen

unread,
Jan 6, 1998, 3:00:00 AM1/6/98
to

> Thorsten explaining another idea about how genius should search

Ha ha ha.

This cannot be true of course. If you search only a few moves then what
score to return? What move to store in hashtable to be the best move?

Forward pruning in combination with extensions can however in a lot
of programs explain a lot of problems:

If you do forward pruning stuff gets stored wrongly in hashtable.
At n+1 ply with j = n+1 and j even depth, you give your opponent another
move now you backup for the forward pruning the score and map a score
for a depth which is odd to a depth that is even, still pruning on the
same condition.

This already gives programs so much trouble that most don't even
profit from bigger hashtables even when loading factor alfa < 0.1
Because when a < 0.1 we would expect at least the same performance.
Forget it. Forward pruning (based on changing things like alfa,beta,
or killermoves dependant) causes all this trouble.

Vincent

--
+----------------------------------------------------+
| Vincent Diepeveen email: vdie...@cs.ruu.nl |
| http://www.students.cs.ruu.nl/~vdiepeve/ |
+----------------------------------------------------+

mclane

unread,
Jan 6, 1998, 3:00:00 AM1/6/98
to

elvi...@aol.com (Elvis PCD) wrote:
>I have been watching this discussion, you set up. in CCC + now in RGCC.
>Gladly it's fully within our hobby.......
>what made me sneak into this - somehow - academic topic, is, that obviously
>there is an important fact missing in the whole discussion :

>>powerful game. It IS the search that makes genius play "boring" since
>>1985 Amsterdam-module.

>So, name the gameplay boring, call the search asymertic, symetric, blach +
>white
>cow - however you perfere - but :
>when did you look into the SSDF list last time ?

Today when the ICCA journal reached me.


>And if you did, why couldn't your postings reflect - at least one single time,
>that this
>asymetric, boring, ply missing + whatsoever program is with all it's 3 last
>engines is
>among the 1st 13.....at present - 3 versions tested only on 90 MHZ.

Dear Elvis, Genius or Richards programs were ALWAYS on the top.
SINCE 1985 this trend has not changed much.
But maybe you have overseen the point that "some few" programs
like

hiarcs
mchess
rebel
junior
virtual-chess
shredder
nimzo98
etc.
have or will have overtaken Genius for a couple of months, years by
now.

Maybe you "oversaw" this ... :-)

Also if you take a look in the ssdf list you see
GENIUS5 with 2431
GENIUS4 with 2410
GENIUS3 with 2408 ELO points.
(I am sure if Genius2 would have had more hash-tables, it would be
very near to genius3).

When I may you remind on the fact that Genius3 has no learning
meanwhile Genius4 and 5 have it, also Genius3 is completely outbooked
over the years by killer - lines of several programs it is obvious to
me that there is not much difference between Genius3-4-5.
In fact, the program has not changed in playing strength over the
years !

Please - I have nothing against Genius. It is a strong and good chess
program, it ever was. But - and I think the data collected by the
ssdf-guys over the years shows this - there is no doubt about that
Genius has not made any progress at all.

>And still (!) overplay brandnew programs - with, let's call them..." newer...
>books....".
>How come ??

This has something to do with ratings. If program A is 50 ELO better
than B, B will from time to time win too.
Take a pencil and make a graph of the programs' ELO in the ssdf list
and you will see that genius was not only overtaken in the years, but
also that Genius' graph is a line without any (meaningful) rise.


>>Use Genius as an instrument. But use it not blind trusting. Remember
>>the
>>asymmetric search
>>and take care of it.

>In complete opposition to this opinion - I can state with the experience of
>nearly 20
>years of constant playing Corr.Chess :

Yes - I listen ....

>Firstly - if one thinks, he can use the " tool " as playing instead of him - he
>would
>never even survive the so called " Hauptturnierklasse " - which could be named
>main tournament class.

I don't think so. I know several players playing ugly chess themselves
but are good equipped. They use PC's and programs. And they play in
the hauptturnierklasse/tournament class. Of course I will not out them
here :-)
On the other hand you are right. It is better to USE the programs
instead of misusing them. The programs shall not play for the human,
the human should USE the programs for analysing. But HE should play.
And not the computers.

>Not to mention the International Master Class, IM-title tournaments, WM
>semi-finals
>or such at all.

I have to disagree. I am in contact with people USING machines and are
fighting high. Again - I cannot show you evidence, this would out
them.

>Within there you could use a P II 300 - 48 hours analysis per move - and would
>be
>crushed to bits + pieces - that's reality !

I don't think so. If they "cheat" by using computers massively, they
could also "cheat" by asking some GM's or IM's in their neighbourhood
or in their chess-club...
And this happens more often than you would expect it.


>Secondly :
>is the conclusion, you offered - to use it as a TOOL - within and along with
>your own
>chess possibilities -
>but .
>I can't understand, really in no way, that someone with your knowledge seems to
>miss the functions, this product offers ! ( See below )

As I said, I do not use Genius for chess. I use Hiarcs, Mchess, Rebel
and other programs. Over the years I have made my experience. I had
the same opinion years ago with the Lang-dedicated machines. I never
used them for correspondance-chess due to the strange behaviour. A
program that is NOT willing to play it's own 48h hours computed
main-line is not a tool I want to use.
In fact I use genius for SHORT analysis... But long analysis with
genius does not make sense. It oversees too much.
The King for example sees more tactics, even Fritz does.


>Without being a big fan of Genius at all - I can testify, that there is - up to
>today -
>no available program, that gives you a so correct sight of the board situation,
>like
>G 3-5, IF you use it RIGHT !

Ok - i cannot refute this. It depends on your definition of RIGHT. If
you fill it different, it can mean different things. I cannot refute
words that have no real definitions.

>Which simply means - that you can start with the normal analysis - let's say
>for
>12 hours; write down the analysis....and then start anew - but with : ZERO
>selective search - which is near to brute force.

>Annother 12 hours - write down, and then look at your own written analysis +
>the
>2, you got by my example.
>Then you can start to build this into your own strategy...if it fits.....and
>because of
>the - still - extremely well overall balance of this product, the benefit then
>is obvious.

Nice usage. But I don't use genius with selectivity sero. Almost
anything takes too long.

>With this long explanation I just tried to give you an example out of reality -
>because
>we never spoke before about the league, in which you personally play chess.

And you don't know where and which games I play because my email-chess
games are not played under my name ! How should you ever find out ??

>Again - even if it might have fit to that time, because of the poorer adjusting
>possibilities :

Yes - I am listening...

>would you tell the readers of all this - WHO was leading the lists constantly
>at that
>time - and not just because of hardware.....?

NOT JUST BECAUSE ?
When richard appeared on the scene he had an advantage of maybe 100 to
150 ELO points to the others depending on their and his hardware.
His program was stronger. But Ed was not far away !
Of course Ed had to use 6502 hardware with less ram/rom size and 4 Mhz
!!

>Yes - and as a member of this group, as well - I take my right to express my
>personal opinion concerning this, also :

Yes - I am listening again....

>This...and the other NG - are just, what the people make out of this.
>Many of them just left because of....well....we all know....


I don't understand this sentence in the above context. Can you make it
clearer for me ? What do you want to say here ??
:-)


>BUT :
>I've seen with my very eyes, that some - not all - of the so called elegibles,
>who left
>this NG - continue constantly throwing mud against eachother - as if the front
>page with all it's well meant rules wasn't there at all.

And ? What has this to do with the position I mentioned that genius
solves faster when I take it back on ply ??


>Could you agree ? - Or should I give proof, here ?

I believe you. Maybe this has something to do with the fact that too
many people in this field like to be FAMOUS and behave like
primadonna's.
From my point of view this is idle narcistic egoistic stuff.
I shit on those things.
I prefer sitting (and in fact we did this in Paris) with some friends
in a small pub or restaurant and having nice chats instead of giving
interviews or reporting about IMPORTANT and IMPRESSIVE things like
Kasparov eating lobsters.
I prefer beeing anonymous. Not because I want to hide. But because I
think it is unimportant WHO somebody is, when his sayings are worthful
! I have seen many important people talking complete rubbish and
behave totally nasty.
I sometimes think that famous people often disbehave that nasty and
asocial because they have no friends telling them that what they have
done was complete bullshit. Always they swim in a group of fans and
other important people and they can never relax.
They don't live in reality. They live on stage. Always. And this is
the reason they get ill. And they misbehave. They are famous, so they
are feared to be wrong. They could lose their reputation. They could
be lose their face. So they misbehave and attack others to hide their
fears. I don't want to live in such a world of a 5 star-hotel lie !

>Interesting people post here, and very " good " ones over there, as well.
>Pityfully, " names " soacked up many readers towards CCC like a sponge -
>without
>a real reason !

I could name you a few reasons ! At least Rolf Tueschen is a very good
reason people went to CCC.
Anonymous is another person.

BTW: I miss Dong Fong very much. Is he still alive ? Could you post
here or in CCC to show me that you are still alive ?


>So, I repeat myself, if you, we all + me.....set up threads like this....maybe
>sometime RGCC will grow again to it's once meant purpose...elsewise there'll be
>soon - I'm quite sure 'bout that - a 3rd group - only for....VERY elegibles.

I understand your point. Maybe you are right.

In this context I want to apologize for having attacked you in CCC.
I was in bad mood and was very sarcastic. Normally I am not sarcastic.
Whenever i show sarcasm, irony or cynism, this is always a sign of
weakness. In those moments something in my private life was not "the
best" and I am not balanced. Later I don't like what I have said and
feel ashamed. I was unfair towards you and would like to apologize in
public, as I have done it in a private (group) email before.

I am glad you are here in CCC and rgcc and I am sure we will have more
interesting discussions like this.

We know each other for many years by now, since the days of the old
dedicated machines. I am sure we can exchange words and also sometimes
not very nice words without beeing suddenly enemies out of nothing.
Also I don't have to be the opposite opinion like you, just because
once we had an argument. We are not primadonna's. Hopefully :-)


>Keep on rockin'

>Elvi...@aol.com
>elvi...@owl-online.de

Right. Elvis is alive.

Tord Kallqvist Romstad

unread,
Jan 7, 1998, 3:00:00 AM1/7/98
to

In article <68ua7t$kul$3...@steve.prima.ruhr.de>, mclane wrote:
>Also if you take a look in the ssdf list you see
>GENIUS5 with 2431
>GENIUS4 with 2410
>GENIUS3 with 2408 ELO points.
>(I am sure if Genius2 would have had more hash-tables, it would be
>very near to genius3).
>
>When I may you remind on the fact that Genius3 has no learning
>meanwhile Genius4 and 5 have it, also Genius3 is completely outbooked
>over the years by killer - lines of several programs it is obvious to
>me that there is not much difference between Genius3-4-5.

I agree with most of your posting, but you seem to confuse the different
versions of Genius with each other. Genius 2 _had_ big hash tables, while
Genius 1 only supported 384 Kb. Genius 4 did not have any learning function.
Here is my opinions on the differences betweeen the Genius versions:

Genius 1->2: Bigger hash tables. Improved opening book (unsound openings like
the Latvian gambit removed). More active playing style.

Genius 2->3: These programs felt very similar to me. According to Peter
Gillgasch, Genius 2 was a leaf-node evaluator while Genius 3 was a
preprocessor. I find this hard to believe, because I think such a change
would change the programs style of play radically. I consider to buy the
Genius Gold Collection to do some experimentation about this.

Genius 3->4: Improved tactics. Especially, Genius' ability to find forced
mates was radically improved (of course, this does not necessarily improve
the playing strength). This seems to be the first Genius version which
extends some lines beyond the 12 plies of selective search. At the search
depth 5/17, for instance, Genius 3 never investigates any lines longer than
17 plies (I think). The reason I believe this is that Genius 1--3 never
found _any_ mates longer than the selective search depth (at least not
without hash tables). To find a mate in 10 moves, for instance, Genius 3
would need a search depth of at least 7/19. Genius 4 (and 5) often finds
these mates at much lower search depths. As a simple test, try the following
position:

+---+---+---+---+---+---+---+---+
8 | | *K| *B| *R| | | | *R|
+---+---+---+---+---+---+---+---+
7 | *P| *P| | | | | | |
+---+---+---+---+---+---+---+---+
6 | | | | | | | | |
+---+---+---+---+---+---+---+---+
5 | P | | *N| | | *P| *P| *Q|
+---+---+---+---+---+---+---+---+
4 | | | N | | | | *N| |
+---+---+---+---+---+---+---+---+
3 | R | | | | Q | | P | |
+---+---+---+---+---+---+---+---+
2 | | | | B | | P | | |
+---+---+---+---+---+---+---+---+
1 | | | R | | | B | K | |
+---+---+---+---+---+---+---+---+
a b c d e f g h


Genius 4 and 5 announces mate in 11 moves almost instantly (I don't remember
the search depth needed. If I remember right, G4 was fastest). Genius 3,
on the other hand, needs several minutes and a search depth of 9/21 to find
the mate (of course, the right move 1.Qf4+ if found much faster).

G4 also plays a more aggressive game than G3.

Genius 4->5: Learning function implemented. The learning function in G5
has no effect at fast time controls. Since I usually only play quick games
against my computer programs, I can not comment on how well the learning
works. G5 seems to be slightly slower than G4 in most positions, perhaps
because of increased knowledge. Apart from that, I don't notice much
difference between the chess engines in these two versions. The opening
book in G5 is much bigger and varied that in the previous versions, but
IMHO the G4 book was better. The opening books in G2-4 was very well tuned.
The Genius 5 book often puts the program in positions it does not understand
very well, and occasionally the program leaves the book in an obviously
lost position.

>In fact, the program has not changed in playing strength over the
>years !
>
>Please - I have nothing against Genius. It is a strong and good chess
>program, it ever was. But - and I think the data collected by the
>ssdf-guys over the years shows this - there is no doubt about that
>Genius has not made any progress at all.

I think there has been _some_ progress, but not much. Richard's programs
also seem to gain less than other programs as the hardware speed increases.

Tord

--
"The true Christian should be aware of mathematicians and those who make
empty prophecies. Chances are already there that mathematicians have
made a covenant with the Devil to confine man in the bounds of Hell"
- St. Augustin

mclane

unread,
Jan 7, 1998, 3:00:00 AM1/7/98
to

rom...@priamos.uio.no (Tord Kallqvist Romstad) wrote:
>I agree with most of your posting, but you seem to confuse the different
>versions of Genius with each other.

Right. It is a long time ago.


> Genius 2 _had_ big hash tables, while
>Genius 1 only supported 384 Kb. Genius 4 did not have any learning function.

Even better than ! Than I can now say: Genius1 with BIG HASH would be
... :-)
But - lets stop joking.
In fact my points need considerations.
It is obvious that Genius3 was outbooked in sweden due to other
programs. Right ?!
It is obvious that Genius4 is not much stronger than Genius3, or ?!
It is obvious that Genius5 is not much stronger than Genius4 but it
has - as you corrected me - a learning function !
So where is the playing-strength progress at all ??

>Here is my opinions on the differences betweeen the Genius versions:

>Genius 1->2: Bigger hash tables. Improved opening book (unsound openings like
>the Latvian gambit removed). More active playing style.

>Genius 2->3: These programs felt very similar to me. According to Peter
>Gillgasch, Genius 2 was a leaf-node evaluator while Genius 3 was a
>preprocessor. I find this hard to believe, because I think such a change
>would change the programs style of play radically. I consider to buy the
>Genius Gold Collection to do some experimentation about this.

I don't believe those preprocessing / not preprocessing changes in
Genius. Sorry Peter, but this would have resulted in a big playing
style change... I have not seen this... maybe I am mistaken, I don't
play much with genius, but - where is the evidence Peter (Gillgasch) ?


>Genius 3->4: Improved tactics. Especially, Genius' ability to find forced
>mates was radically improved (of course, this does not necessarily improve
>the playing strength). This seems to be the first Genius version which
>extends some lines beyond the 12 plies of selective search. At the search
>depth 5/17, for instance, Genius 3 never investigates any lines longer than
>17 plies (I think). The reason I believe this is that Genius 1--3 never
>found _any_ mates longer than the selective search depth (at least not
>without hash tables). To find a mate in 10 moves, for instance, Genius 3
>would need a search depth of at least 7/19. Genius 4 (and 5) often finds
>these mates at much lower search depths. As a simple test, try the following
>position:

Of course you are right ! But the increase of selective search has not
lead to a playing-strength increase.


>Genius 4 and 5 announces mate in 11 moves almost instantly (I don't remember
>the search depth needed. If I remember right, G4 was fastest). Genius 3,
>on the other hand, needs several minutes and a search depth of 9/21 to find
>the mate (of course, the right move 1.Qf4+ if found much faster).

>G4 also plays a more aggressive game than G3.

This is also true - but this leads also in less points in sweden !!
Therefore genius5 is more passive again.

>Genius 4->5: Learning function implemented. The learning function in G5
>has no effect at fast time controls.

In sweden they play 40/120 minutes. I guess there it affects the
ratings.

> Since I usually only play quick games
>against my computer programs, I can not comment on how well the learning
>works.

Very good - you can believe me ! Or if you don't believe me - ask
Chris. He has 5 autoplayer pairs, I have only one.

> G5 seems to be slightly slower than G4 in most positions, perhaps
>because of increased knowledge. Apart from that, I don't notice much
>difference between the chess engines in these two versions. The opening
>book in G5 is much bigger and varied that in the previous versions, but
>IMHO the G4 book was better. The opening books in G2-4 was very well tuned.
>The Genius 5 book often puts the program in positions it does not understand
>very well, and occasionally the program leaves the book in an obviously
>lost position.

Agreed.

>>Please - I have nothing against Genius. It is a strong and good chess
>>program, it ever was. But - and I think the data collected by the
>>ssdf-guys over the years shows this - there is no doubt about that
>>Genius has not made any progress at all.

>I think there has been _some_ progress, but not much.

If you substract the learning and the cooked book of genius3 (what
decreases genius3 ratings) genius5 would not be much stronger than
genius3.
Or ?

Elvis PCD

unread,
Jan 8, 1998, 3:00:00 AM1/8/98
to

Im Artikel <68ua7t$kul$3...@steve.prima.ruhr.de>, mcl...@prima.ruhr.de (mclane)
schreibt:

>Thema: Re: Genius' asymmetric-search by example: TRY yourself
>Von: mcl...@prima.ruhr.de (mclane)
>Datum: Tue, 06 Jan 1998 23:15:23 GMT

Hi there,

just to make it very short - wrote a long, dull answer yesterday night - must
have kinda lost it in my dumb brain - you'll know this by now - here's the
essence, only,
coram publico :

With my refections towards your thread here and within CCC I only tried to put
up
something like a counter balance towards Genius, just not to see that this prog
is academically torn to pieces.....

>Ok - i cannot refute this. It depends on your definition of RIGHT. If
you fill it different, it can mean different things. I cannot refute
words that have no real definitions.

At least I just didn't write something - I tried to explain how I use + why I
think this could be " right ".

>>With this long explanation I just tried to give you an example out of reality
-
>>because
>>we never spoke before about the league, in which you personally play chess.

>And you don't know where and which games I play because my email-chess
>games are not played under my name ! How should you ever find out ??

Nothing to be added at all - we just simply never spoke about this (...why
not...)

>NOT JUST BECAUSE ?
>When richard appeared on the scene he had an advantage of maybe 100 to
>150 ELO points to the others depending on their and his hardware.
>His program was stronger. But Ed was not far away !
>Of course Ed had to use 6502 hardware with less ram/rom size and 4 Mhz
>!!

I remember well, we ( Zens ) spent alot of money in this power booster ( what
was the name ? ) to catch up speed in MHZ against R.L. - dind't work out.....

>I don't understand this sentence in the above context. Can you make it
>clearer for me ? What do you want to say here ??
>:-)

First I didn't trust my eyes - then I noticed the..... :-)

>I could name you a few reasons ! At least Rolf Tueschen is a very good
>reason people went to CCC.
>Anonymous is another person.

Erh - now, ain't that a sacrilege.....I remember a...Pope....no - that must
have been at church.....

>In this context I want to apologize for having attacked you in CCC.

>I was unfair towards you and would like to apologize in
>public, as I have done it in a private (group) email before.

>We know each other for many years by now, since the days of the old
>dedicated machines.

Shoud have posted it, where you made a nut out of me......the last sentece is
the fact - we're all humans - forget it now. - Point taken ?

> We are not primadonna's. Hopefully :-)

Don't you under-estimate old ELVIS....:-)

Tord Kallqvist Romstad

unread,
Jan 8, 1998, 3:00:00 AM1/8/98
to

In article <690o0c$vee$1...@steve.prima.ruhr.de>, mclane wrote:
>rom...@priamos.uio.no (Tord Kallqvist Romstad) wrote:
>>I agree with most of your posting, but you seem to confuse the different
>>versions of Genius with each other.
>
>Right. It is a long time ago.
>
>
>> Genius 2 _had_ big hash tables, while
>>Genius 1 only supported 384 Kb. Genius 4 did not have any learning function.
>
>Even better than ! Than I can now say: Genius1 with BIG HASH would be
>... :-)
>But - lets stop joking.
>In fact my points need considerations.
>It is obvious that Genius3 was outbooked in sweden due to other
>programs. Right ?!

If you say so. I wasn't aware of that, but it doesn't surprise me. However,
I think that G2, G3 and G4 all had almost identical opening books.

>It is obvious that Genius4 is not much stronger than Genius3, or ?!
>It is obvious that Genius5 is not much stronger than Genius4 but it

Yes, in both cases it is obvious that the newer program is not _much_
stronger.

>has - as you corrected me - a learning function !
>So where is the playing-strength progress at all ??

>>Genius 2->3: These programs felt very similar to me. According to Peter


>>Gillgasch, Genius 2 was a leaf-node evaluator while Genius 3 was a
>>preprocessor. I find this hard to believe, because I think such a change
>>would change the programs style of play radically. I consider to buy the
>>Genius Gold Collection to do some experimentation about this.
>
>I don't believe those preprocessing / not preprocessing changes in
>Genius. Sorry Peter, but this would have resulted in a big playing
>style change... I have not seen this... maybe I am mistaken, I don't
>play much with genius, but - where is the evidence Peter (Gillgasch) ?

Then we agree about this. I often wonder if Genius is a pure preprocessor
or a hybrid leaf-node eval/preprocessing program. It seems obvious that
Genius does _some_ preprocessing, but it's positional style of play doesn't
resemble most other preprocessors. Does anybody know how many nps Genius
searches? The rumors I have heard vary wildly. Some people claims that the
program is as fast as Fritz, some people says it is more like MChess.

>>G4 also plays a more aggressive game than G3.
>
>This is also true - but this leads also in less points in sweden !!

I am not sure what you mean here. Genius 4 is higher rated than Genius 3 on
the SSDF list, this is true on the 486/66 as well as the P5/90. Admittedly,
the difference on the P5/90 is only two points. I also have a feeling that
the more active style of G4 makes the program a more difficult opponent for
humans, although it doesn't seem to help much against computers.

>Therefore genius5 is more passive again.

Yes. G5 is somewhere between G3 and G4 in this respect.

>>>Please - I have nothing against Genius. It is a strong and good chess
>>>program, it ever was. But - and I think the data collected by the
>>>ssdf-guys over the years shows this - there is no doubt about that
>>>Genius has not made any progress at all.
>

>>I think there has been _some_ progress, but not much.
>
>If you substract the learning and the cooked book of genius3 (what
>decreases genius3 ratings) genius5 would not be much stronger than
>genius3.
>Or ?

I am not convinced. I am not sure which handicap is worse, the cooked but
well-tuned book of G3 or the bad book of G5. I also doubt that the learning
function has much effect on the SSDF ratings. Remember that the games are
played by several different testers, who probably don't exchange their learn
files. The G5 book also has lots of variety, which makes it unlikely that
the same opening lines appear several times in one match. I also think it
is unfair to subtract the assumed effect of the learning function when you
compare the two different versions. To me, this is almost like saying that
"If you subtract the superior chess knowledge of Rebel 9, the program isn't
much stronger than Fritz 3". The interesting question is whether or not
G5 plays better than G3, not _why_ it plays better.

Anyway, I don't think we disagree very much about this. Your opinion seems
to be that Richard has not been able to improve his program the last few
years, while my opinion is that he has only made very small improvements.
If you look at the Swedish rating list, you will see that all Genius versions
appear in the "right" order (see below). I doubt that this is random.

Rating + - Games Won Oppo
------ --- --- ----- --- ----
5 Genius 5.0 Pentium MMX 200 MHz 2501 35 -33 438 65% 2392

10 Genius 5.0 Pentium 90 MHz 2431 23 -23 920 55% 2398
12 Genius 4.0 Pentium 90 MHz 2410 25 -25 809 64% 2311
14 Genius 3.0 Pentium 90 MHz 2408 27 -26 719 62% 2321

25 Genius 4.0 486/50-66 MHz 2371 25 -25 804 51% 2361
28 Genius 3.0 486/50-66 MHz 2357 24 -23 910 62% 2270
30 Mephisto Genius 2.0 486/50-66 MHz 2338 23 -23 917 58% 2278
44 Chess Genius 1.0 486/50-66 MHz 2284 23 -23 931 54% 2257
52 Chess Genius 1.0 486/33 MHz 2241 32 -32 463 50% 2241

mclane

unread,
Jan 8, 1998, 3:00:00 AM1/8/98
to

elvi...@aol.com (Elvis PCD) wrote:

>Im Artikel <68ua7t$kul$3...@steve.prima.ruhr.de>, mcl...@prima.ruhr.de (mclane)
>schreibt:

>>Thema: Re: Genius' asymmetric-search by example: TRY yourself
>>Von: mcl...@prima.ruhr.de (mclane)
>>Datum: Tue, 06 Jan 1998 23:15:23 GMT

>Hi there,

>just to make it very short - wrote a long, dull answer yesterday night - must
>have kinda lost it in my dumb brain - you'll know this by now - here's the
>essence, only,
>coram publico :

Thanks.

>With my refections towards your thread here and within CCC I only tried to put
>up
>something like a counter balance towards Genius, just not to see that this prog
>is academically torn to pieces.....

But this was not my idea. But it is time that Richard is doing
something to stop the misery. It is time that he changes Genius, or it
will sink and sink and sink or others overtake Genius ...
I don't know what exactly their crisis is, but from my point of view
it looks obvious that they are not able to CHANGE Genius. Whatever
they do is only cosmetic, make-up changes. I don't think this will
bring Genius back as a leader. If, if e.g. Richard transformes his
16-Bit assembler into 32-bit assembler, I am not sure if this is
enough if he is not changing other things.
I don't know what their problem is. But from my point of view it
looked that genius was tuned in the wrong direction.

>>Ok - i cannot refute this. It depends on your definition of RIGHT. If
>you fill it different, it can mean different things. I cannot refute
>words that have no real definitions.

>At least I just didn't write something - I tried to explain how I use + why I


>think this could be " right ".

True.
In the moment I have no correspondance game running to try it out.
Maybe in one month.

>>NOT JUST BECAUSE ?
>>When richard appeared on the scene he had an advantage of maybe 100 to
>>150 ELO points to the others depending on their and his hardware.
>>His program was stronger. But Ed was not far away !
>>Of course Ed had to use 6502 hardware with less ram/rom size and 4 Mhz
>>!!

>I remember well, we ( Zens ) spent alot of money in this power booster ( what


>was the name ? ) to catch up speed in MHZ against R.L. - dind't work out.....

Schaetzle und Bsteh's 16 or 20 Mhz Accelerator. It made a 6502-4 Mhz
into 20 Mhz !!
I think Rebel5 or MM4 had good , very good results with BOOSTER. So it
was obvious that a FASTER Schroeder program would be as top as
Richards programs.
Wasn't the Schroeder Booster a few months at the top until the
ssdf-guys decided to throw it out of the list (???!!!???) because it
was not commercial or whatever reasons.

This throwing it out was the end of this machine. The sense of the
machine was to transform slow 6502 programs into new categories by
making them 5 times faster !
Throwing them out of the list made it impossible for the public to see
the RESULTS of the boosting, and the company went into bancrupt.


>First I didn't trust my eyes - then I noticed the..... :-)

>>In this context I want to apologize for having attacked you in CCC.


>>I was unfair towards you and would like to apologize in
>>public, as I have done it in a private (group) email before.

>>We know each other for many years by now, since the days of the old
>>dedicated machines.

>Shoud have posted it, where you made a nut out of me......the last sentece is


>the fact - we're all humans - forget it now. - Point taken ?

Thanks. I was angry about another person we know both. Therefore I was
dick-headed and did not post anything towards you both for a while.
Sorry.


>Don't you under-estimate old ELVIS....:-)

The older I get, the less important are NAMES and FAME and glory and
BEEING RIGHT. The older i get the more important are friendship and
honest words.

Markus Koelblin

unread,
Jan 8, 1998, 3:00:00 AM1/8/98
to

Tord Kallqvist Romstad wrote:
>
> In article <690o0c$vee$1...@steve.prima.ruhr.de>, mclane wrote:
> >rom...@priamos.uio.no (Tord Kallqvist Romstad) wrote:

[snip]

> >>Genius 2->3: These programs felt very similar to me. According to Peter
> >>Gillgasch, Genius 2 was a leaf-node evaluator while Genius 3 was a
> >>preprocessor. I find this hard to believe, because I think such a change
> >>would change the programs style of play radically. I consider to buy the
> >>Genius Gold Collection to do some experimentation about this.
> >
> >I don't believe those preprocessing / not preprocessing changes in
> >Genius. Sorry Peter, but this would have resulted in a big playing
> >style change... I have not seen this... maybe I am mistaken, I don't
> >play much with genius, but - where is the evidence Peter (Gillgasch) ?
>
> Then we agree about this. I often wonder if Genius is a pure preprocessor
> or a hybrid leaf-node eval/preprocessing program. It seems obvious that
> Genius does _some_ preprocessing, but it's positional style of play doesn't
> resemble most other preprocessors. Does anybody know how many nps Genius
> searches? The rumors I have heard vary wildly. Some people claims that the
> program is as fast as Fritz, some people says it is more like MChess.

I am replying from the account of a friend. I saw this post by pure
accident...

According to Ossi Weiner it is in the 30-40 K ballpark on a P90 if my
memory
is correct. Regarding the root processor issue, I probably don't have
hard evidence
for that. I am seeing big evaluation changes from move to move, which
usually
points to the presence of a preprocessing routine. In Genius' case it
could be
that the asymetric search regime, which is driven by a odd combination
of parity
ply pruning, tapered forward pruning and detecting threats of the
opponent results
in those changes, but they are often much too big to be completely
explained by
that, especially when a piece class is removed from the game.

In any case I don't think that Richard does a very complex interior node
evaluation,
beside a tactical analysis to detect mating threats. The stone age
version of DarkThought
(pre-Hong Kong version) which was basically a pure preprocessor
(although a very primitive
one) had no major problems against Genius (P90 vs. Alpha 175 I think).
It was hard to
get a win, but not hard to get a draw, so I'd say the power of Genius
comes from it's
search regime.

The historical / commercial context of the Genius3 release is also quite
important.
Didn't Richard loose ground to Fritz at that time? The only thing
Richard could strip
down to increase speed was the interior node evaluation. If the 30 K
number from Ossi
is correct then it is obvious that there can't be a real complex
evaluation routine
inside that thing...

-- Peter

mclane

unread,
Jan 8, 1998, 3:00:00 AM1/8/98
to

Markus Koelblin <Markus....@hadiko.de> wrote:
>It was hard to
>get a win, but not hard to get a draw, so I'd say the power of Genius
>comes from it's
>search regime.

I completely agree here !
Thats the reason we have to find out about HOW the search works.

>The historical / commercial context of the Genius3 release is also quite
>important.
>Didn't Richard loose ground to Fritz at that time? The only thing
>Richard could strip
>down to increase speed was the interior node evaluation. If the 30 K
>number from Ossi
>is correct then it is obvious that there can't be a real complex
>evaluation routine
>inside that thing...


Again you have seen the point !
Weiner/Lang was forced to react on Fritz !!
And Fritz made big headlines with BLITZ results and tactical things !
Therefore Weiner forced Lang to compete against Fritz.
I think in the end this decision was completely wrong and only Fritz
won from it.

>-- Peter

Komputer Korner

unread,
Jan 8, 1998, 3:00:00 AM1/8/98
to

Richard Lang has finished the job of transforming Genius into 32 bit code.
The problem is that it doesn't play any better than the 16 bit version. He
is still trying to determine why this is so,but maybe Bob Hyatt can shed
some light on this.
--
- -
Komputer Korner

The inkompetent komputer

If you see a 1 in my email address, take it out before replying.
Please do not email both me and the r.g.c.c. at the same time. I read all
the postings on r.g.c.c.
Also every statement of mine should be taken with a grain of salt. Read at
your own risk and
assume that it is only this humble komputer's opinion.

mclane <mcl...@prima.ruhr.de> wrote in article
<69365n$8s6$2...@steve.prima.ruhr.de>...

Robert Hyatt

unread,
Jan 8, 1998, 3:00:00 AM1/8/98
to

Komputer Korner <kor...@netcom.ca> wrote:
: Richard Lang has finished the job of transforming Genius into 32 bit code.

: The problem is that it doesn't play any better than the 16 bit version. He
: is still trying to determine why this is so,but maybe Bob Hyatt can shed
: some light on this.

I don't know why it would play better, if all that was done was to
eliminate the 16 bit instructions. It should be faster, but I have
no idea how much faster, since it would depend on whether it was run
on a Pentium, Pentium Pro, Pentium II, K6, or whatever. But I wouldn't
expect significant playing improvement unless he got significantly
faster, which I have no idea about.

I've never tested 16 bit code on my P6 for example, so I have no
idea how that bogs things down...

: --
: - -
: Komputer Korner

: The inkompetent komputer

: If you see a 1 in my email address, take it out before replying.
: Please do not email both me and the r.g.c.c. at the same time. I read all
: the postings on r.g.c.c.
: Also every statement of mine should be taken with a grain of salt. Read at
: your own risk and
: assume that it is only this humble komputer's opinion.

: mclane <mcl...@prima.ruhr.de> wrote in article
: <69365n$8s6$2...@steve.prima.ruhr.de>...
: . I don't think this will
:> bring Genius back as a leader. If, if e.g. Richard transformes his
:> 16-Bit assembler into 32-bit assembler, I am not sure if this is
:> enough if he is not changing other things.
:> I don't know what their problem is. But from my point of view it
:> looked that genius was tuned in the wrong direction.

:>

Elvis PCD

unread,
Jan 9, 1998, 3:00:00 AM1/9/98
to

Im Artikel <69365n$8s6$2...@steve.prima.ruhr.de>, mcl...@prima.ruhr.de (mclane)
schreibt:

>Thema: Re: Genius' asymmetric-search by example: TRY yourself
>Von: mcl...@prima.ruhr.de (mclane)

>Datum: Thu, 08 Jan 1998 19:36:40 GMT

Hi there,

> If, if e.g. Richard transformes his
>16-Bit assembler into 32-bit assembler, I am not sure if this is
>enough if he is not changing other things.
>I don't know what their problem is. But from my point of view it
>looked that genius was tuned in the wrong direction.

As far as I know, this is, what he actually is on at present. 1st results
didn't show
the wished results - so they decided cleverly to wait.
The man from Munich is busy like a mouse to create masters, gold masters,
collective master, ultimate.....just to press the very last dime outta that
thing - which, btw., gives proof of a buisness man of extra-class.

After I once had the enlightning possibility of personal contact, my mind is
set for good on the rest, but that ain't nothing of public address.

Personally I " fear ", that the possibilities of this engine have been wrung
out to the last drop....perhaps Richard come back lion-hearted and challenge
them all again - let's wait.

>Throwing them out of the list made it impossible for the public to see
the RESULTS of the boosting, and the company went into bancrupt.

Correct - Schaetzle + Bsteh......a real pity that such an innovatice Co. went
down yonder - I'd wonder what they'd might invent with today's possibilities.

>Thanks. I was angry about another person we know both. Therefore I was
dick-headed and did not post anything towards you both for a while.
Sorry.

ELVIS knows...as I said - forget it now.

>The older i get the more important are friendship and
honest words.

The best thing to do is remember this....lot's of people rise up, coming out of
nowhere, create a lot of rumble in the jungle - use ( ! ) the NAMES in this
scene -
and " throw " them away as soon as they think, they reached their aim.

I noticed this twice, recently.
One thing - from...let's say, the far east of our country......
and the other thing....from the country, " where the fog was invented
"......:-)

Very likely I don't have to add clearer explanations.

BTW - ELVIS has a hidden function in a sub-menue :) : squirrel.....
clicking this one he starts to collect, collect.......
he found something really rare -
the 1st CC magazine of a gent from Lünen......ain't that a thing !?

Let's see. if we find a new thread here - maybe to help questioning
comrades...??

mclane

unread,
Jan 9, 1998, 3:00:00 AM1/9/98
to

"Komputer Korner" <kor...@netcom.ca> wrote:

>Richard Lang has finished the job of transforming Genius into 32 bit code.
>The problem is that it doesn't play any better than the 16 bit version. He
>is still trying to determine why this is so,but maybe Bob Hyatt can shed
>some light on this.

As I said before, it is NOT NEW to me that whatever they do, it is not
increasing their playing strength. I told a few times here that I
believe that there is not much difference between the versions.
I would like to work with Richard and try to change Genius into a
stronger program. But
i tried it 3 times and whenever i spoke about asymmetrie and asked for
a symmetric genius the communication stopped for an unknown reason.

I have spent many years watching Genius.
I think I could work out ways for this program to make it stronger.

mclane

unread,
Jan 9, 1998, 3:00:00 AM1/9/98
to

elvi...@aol.com (Elvis PCD) wrote:

>As far as I know, this is, what he actually is on at present. 1st results
>didn't show
>the wished results - so they decided cleverly to wait.
>The man from Munich is busy like a mouse to create masters, gold masters,
>collective master, ultimate.....just to press the very last dime outta that
>thing - which, btw., gives proof of a buisness man of extra-class.

>After I once had the enlightning possibility of personal contact, my mind is
>set for good on the rest, but that ain't nothing of public address.

I cannot subscribe it more accurate than you :-)


>Personally I " fear ", that the possibilities of this engine have been wrung
>out to the last drop....perhaps Richard come back lion-hearted and challenge
>them all again - let's wait.

I hope so too.


>Correct - Schaetzle + Bsteh......a real pity that such an innovatice Co. went
>down yonder - I'd wonder what they'd might invent with today's possibilities.

I think there are still some Turbo-Kit's in colletcors treasure-boxes.
I will ask Karsten, maybe he knows somebody with a Turbo-Kit 20
Mhz....

>>Thanks. I was angry about another person we know both. Therefore I was
>dick-headed and did not post anything towards you both for a while.
>Sorry.

>ELVIS knows...as I said - forget it now.

Yeap !

>>The older i get the more important are friendship and
>honest words.

>The best thing to do is remember this....lot's of people rise up, coming out of
>nowhere, create a lot of rumble in the jungle - use ( ! ) the NAMES in this
>scene -
>and " throw " them away as soon as they think, they reached their aim.

?!?

>I noticed this twice, recently.
>One thing - from...let's say, the far east of our country......

?!?!
Far east of germany ?
The only Genosse I know from far east called me some hours ago and
told me about his latest version to download, Joerg Burwitz. You don't
talk about him, I am sure. His version 4.10 makes 67.000 NPS on my
k6/200 and plays a nice game against Rebel9 so far.
No - you speak about another guy I guess. Hm.

>and the other thing....from the country, " where the fog was invented
>"......:-)

Britain ?

>Very likely I don't have to add clearer explanations.

We can do this in email... :-)

>BTW - ELVIS has a hidden function in a sub-menue :) : squirrel.....
>clicking this one he starts to collect, collect.......
>he found something really rare -
>the 1st CC magazine of a gent from Lünen......ain't that a thing !?

Oh god, no ! I did this together with Mr. Zens from NOVAG.
Uff. I stopped with edition 2 when Karsten suggested to censor me out
of my own magazine. This was too much for me.

>Let's see. if we find a new thread here - maybe to help questioning
>comrades...??

Ok. No problem.
But I am no Komputer. :-)
I will go sleep now.

Komputer Korner

unread,
Jan 9, 1998, 3:00:00 AM1/9/98
to

Bob, In the Intel environment, Does Crafty run fastest on a PII-300 or on a
Pentium Pro?
--
- -
Komputer Korner

The inkompetent komputer

If you see a 1 in my email address, take it out before replying.
Please do not email both me and the r.g.c.c. at the same time. I read all
the postings on r.g.c.c.
Also every statement of mine should be taken with a grain of salt. Read at
your own risk and
assume that it is only this humble komputer's opinion.

Robert Hyatt <hy...@cis.uab.edu> wrote in article
<693o5h$arc$1...@juniper.cis.uab.edu>...


> Komputer Korner <kor...@netcom.ca> wrote:
> : Richard Lang has finished the job of transforming Genius into 32 bit
code.
> : The problem is that it doesn't play any better than the 16 bit version.
He
> : is still trying to determine why this is so,but maybe Bob Hyatt can
shed
> : some light on this.
>

> I don't know why it would play better, if all that was done was to
> eliminate the 16 bit instructions. It should be faster, but I have
> no idea how much faster, since it would depend on whether it was run
> on a Pentium, Pentium Pro, Pentium II, K6, or whatever. But I wouldn't
> expect significant playing improvement unless he got significantly
> faster, which I have no idea about.
>
> I've never tested 16 bit code on my P6 for example, so I have no
> idea how that bogs things down...

> Robert Hyatt Computer and Information Sciences

Tord Kallqvist Romstad

unread,
Jan 9, 1998, 3:00:00 AM1/9/98
to

Thanks for your interesting and informative answer, Peter!

In article <34B52AFA...@hadiko.de>, Markus Koelblin wrote:
<Big snip...>

>The historical / commercial context of the Genius3 release is also quite
>important.
>Didn't Richard loose ground to Fritz at that time?

I can't understand why they considered Fritz to be the most serious
opponent. At the time of Genius 3, the main competitors of Genius was
the King, Rebel and MChess. Fritz was far behind.

> The only thing
>Richard could strip
>down to increase speed was the interior node evaluation. If the 30 K
>number from Ossi
>is correct then it is obvious that there can't be a real complex
>evaluation routine
>inside that thing...

Yes. However, I see no evidence that the interior node evaluation has been
reduced in Genius 3 compared to Genius 2. These programs seemed almost
identical to me.

Robert Hyatt

unread,
Jan 9, 1998, 3:00:00 AM1/9/98
to

Komputer Korner <kor...@netcom.ca> wrote:
: Bob, In the Intel environment, Does Crafty run fastest on a PII-300 or on a
: Pentium Pro?

PII/300... but not 50% faster than on the P6/200, because the PII's cache
is slower (L2 cache). But the PII is definitely faster..


mclane

unread,
Jan 10, 1998, 3:00:00 AM1/10/98
to

rom...@kassandra.uio.no (Tord Kallqvist Romstad) wrote:

>I can't understand why they considered Fritz to be the most serious
>opponent. At the time of Genius 3, the main competitors of Genius was
>the King, Rebel and MChess. Fritz was far behind.

Of course Fritz was far behind and not that strong like the programs
YOU mentioned. BUT Genius was mainly sold from germany via Ossi Weiner
and Fritz was sold via ChessBase, also a german company. Therefore
there was a big competition in germany concerning advertising and
getting market-parts. Genius had to compete with Fritz. Believe it or
not. The program that cries the loudest looks often like the
strongest. The marketing guys of ChessBase have shown us that a
program that was never one of the tops can be seen as one of the top
just by doing some advertising gags and a little help of some friends
:-)


Komputer Korner

unread,
Jan 11, 1998, 3:00:00 AM1/11/98
to

I am starting not to trust the analysis of Fritz 5, at least not in in 3 K
mode. Bur Fritz 5 does have the best features of any playing program I
know.
--
- -
Komputer Korner

The inkompetent komputer

If you see a 1 in my email address, take it out before replying.
Please do not email both me and the r.g.c.c. at the same time. I read all
the postings on r.g.c.c.
Also every statement of mine should be taken with a grain of salt. Read at
your own risk and
assume that it is only this humble komputer's opinion.

mclane <mcl...@prima.ruhr.de> wrote in article
<69917s$4vb$2...@steve.prima.ruhr.de>...

0 new messages