RS 2150 and 2150L significant difference ?

[Queegmeister]

Hi

I notice that the 2150 has many more book moves than the other - 100k total ?
same program (just huge book for 2150 ?)

curious if someone has played them against each other and the results.

I have a mint 2150L and was wondering if it would be worth getting the 2150 as well, if the difference is significant enough.


( I checked a website and it seems the ratings are similar ) It seems weird considering the incredible difference in the Opening Book size.

Queeg

[LWSteve]

Hi

I notice that the 2150 has many more book moves than the other - 100k total ?
same program (just huge book for 2150 ?)

curious if someone has played them against each other and the results.

I have a mint 2150L and was wondering if it would be worth getting the 2150 as well, if the difference is significant enough.


( I checked a website and it seems the ratings are similar ) It seems weird considering the incredible difference in the Opening Book size.

Queeg

The 2150 is noticeably stronger than the 2150L

In my computer tournament lately the 2150 drew with Mephisto Explorer Pro (elo 2060)
And it defeated Mephisto Talking Chess Academy. (elo 1965)

I remember I played it a game about a year ago and I had it beat and it
made a miraculous comeback on me with some great knight moves and it beat me at the end. But I'm only a class 'B' player so...

Ya the 2150 has some punch if you play it on tournament mode 3 min. average response (square b2)

LWSteve

[Larry]

This one is a bit strange. I never gave these two machines much thought.
I do remember assuming the "L" in the 2150L stood for "library",
meaning it had a bigger library. It has a smaller library by quite a margin.
The 2150L was manufactured 4 years later, runs on the later H8
processor, and is about 30elo weaker.
Julio Kaplan was a junior master, unlike most chess programmers,
who are mediocre players if they are interested in playing chess at all.
He is a strong player, but not a strong programmer, IMO.
L

[Reinfeld]

I've always been intrigued by the difference in opening books.

The question invites another test for Nick to consider:

1) How much do opening books matter?

I looked at this a while back, here -

http://www.hiarcs.net/forums/viewtopic.php?t=4754

One point struck me at the time - the Montreux and the RISC 2500 are virtually the same machine, by most accounts. The difference lies in the opening book. The Montreux almost quadruples the RISC. I don't have both (just the RISC), so I can't test it independently.

Yet the ratings between the two machines are negligible - my own little average of averages suggests they stand three points apart.

A bigger book doesn't mean a better book, obviously. You can add reams of crap variations without increasing quality of response. A reliable test would have to examine multiple variations at their limits- for instance, taking two machines into a 10-move variation of the closed Ruy Lopez known by both, and running the test when they cease to respond instantly.

- R.

[Theo]

I have a mint 2150L and was wondering if it would be worth getting the 2150 as well, if the difference is significant enough.

Hi Queeg,

these are two different programs, although from the same programmer. Instead of the 2150 I would suggest you get a Turboking or a Corona since they have more Mhz and probably also a newer 6502 program and thus play considerably stronger.

I ran some tests with the 2150L earlier this year and ended up selling it. Its playing style is weird with sometimes strong tactics, but bad strategy and king safety.

The results are:

vs Roma 68020 1 / 2 (2085)
vs Roma 68000 0 / 2 (2004)
vs Mach III 1 / 4 (2020)
vs Par Excellence 0 / 4 (1852)
vs Revelation Glasgow 2,5 / 4 (1910)
vs GK 2000 0 / 2 (1973)
vs Modena 0 / 2 (1963)
vs Rebell 5.0 2 / 2 (1857)
6,50 / 22 = 29,55 %; av. opponents = 1949,636364

The opponents have been too strong but I just do not own any weaker devices.

Based on this results the 2150L earned a .info ELO of 1777, 9 points below Saitek Prisma/Blitz which run at 10Mhz instead of 8.

Some of the games are published here:

http://www.schachcomputer.info/forum/f1 ... -4444.html

http://www.schachcomputer.info/forum/f1 ... -4442.html


2150 Regards,
Theo

[spacious_mind]

I've always been intrigued by the difference in opening books.

The question invites another test for Nick to consider:

1) How much do opening books matter?

I looked at this a while back, here -

http://www.hiarcs.net/forums/viewtopic.php?t=4754

One point struck me at the time - the Montreux and the RISC 2500 are virtually the same machine, by most accounts. The difference lies in the opening book. The Montreux almost quadruples the RISC. I don't have both (just the RISC), so I can't test it independently.

Yet the ratings between the two machines are negligible - my own little average of averages suggests they stand three points apart.

A bigger book doesn't mean a better book, obviously. You can add reams of crap variations without increasing quality of response. A reliable test would have to examine multiple variations at their limits- for instance, taking two machines into a 10-move variation of the closed Ruy Lopez known by both, and running the test when they cease to respond instantly.

- R.

Hi Reinfeld,

I am just about completing game 6 of the Barden book and I had exactly the same thoughts last night as I was playing some of the computers through this game.

My thoughts are that opening book matters a lot and it is here that a human player has his distinct advantage against a chess computer because the human can steer the computer through repeating the same opening over and over again into a situation where he can beat it and claim that he is better. This is what happens all the time when a human plays his dedicated computer. But that is also natural because most people only know a very few openings well and therefore keep repeating them.

For dedicated against dedicated chess computer matches the opening knowledge from my experience has a lesser advantage unless a computer happened to have stumbled upon a Killer Variation that the other computer did not have. I have seen far too many games where a computer was out of the book after about 6 moves and another after about 12 moves where the 6 moves computer won anyway to make me think that the advantage is less.

A truly fair way to play human against computer would be just like we are doing in the Barden book is to start the game after the opening. In other words just pick a game at random don't even try to choose a game, just point your finger at a game in your database, then use Chessbase for example to go to the end of the book and then flip a coin of who plays black and who plays white, computer or human and then play and see who wins.

That I think would be a much truer strength test between human and computer. No anti-computer chess openings and such which all mislead.

Best regards,

[Larry]

The purpose of the opening book is to provide speed and accuracy,
and to arrive at a position that the evaluation function can hopefully
cope with. Notice that they tried to end most opening lines with an
obvious move, eg, a recapture?
I did think the opening book in the Fidelity SC9 was too deep and
not wide enough. Others maybe thought so too, because Fidelity
with their later Elegance and '12' level machines had shallower but
wider openings. This provides for more variety for us club hackers,
because most of us neither know nor care to learn opening sequences
15 ply deep, but will benefit from a wider book.
The memory resources saved with a modest opening book are better
utilized in providing knowledge for the all important endgame. Many
a time I have had a reasonable strength dedicated unit play a pretty
fierce middle game only to botch a fairly simple ending with a dumb
rook or king move, which turned a win into a draw or even a loss.
It does seem that chess "knowledge" is equivalent to underclocking.
An example of this is to be found with the Scisys Turbostar432 with
the openings "KSO" module, versus the same machine minus the
KSO module. Setting up random endgame mating positions and
timing them, I noticed that the non KSO is consistently slightly
faster than the KSO version. This suggested to me that the program
does not bypass the openings book after the opening library is exhausted,
but continues looking for an opening move before referring to the
evaluation function right through until the game ends.
Larry

[9RX]

Hey Queeg,

Yes there are some weird things about these two, the 2050 and 2050L.

Weird things that don't make sense to me?
Fist of all it's weird that the 2050 has double the ROM as the 2050L..
64KB verses 32KB for the 2050L. Maybe that's the larger library?

It's weird that the older board has a higher elo,
while having a processor going less than half as fast as the newer 2050L with a newer processor? Weird?

I've read that by doubling the clock speed or more of the 2050,
the increase of playing speed is only slightly faster, with maybe a 10% increase of elo? That's weird too?

A person, me, would think the computer could solve problems double as fast? Double the clock, double the speed?

This is kinda the same thing I see when thinking about dual processor boards..
take the Renaissance board with the Analyst module for example.
Two processors, two programs running together, but not double performance.
I suppose they supplement each other, or maybe one mostly carries the load?

I wanna read more 'bout this subject..

love this place,
9R