Rating of Radio Shack 1650
G...@psuvm.psu.edu
of an older model RS 1650 (the new one is a 1650L) chess computer?
Thanks.
Gil Gall (g...@psuvm.psu.edu)
James H. Coombs
>As I am new to this group, does anyone know what the rating would be
>of an older model RS 1650 (the new one is a 1650L) chess computer?
The 2150L is about 1950. The 1650L is not worth rating. It might be
good for a child. You should get something at least 200 points
stronger than you so that you can improve and still use it. You will
also probably want to play it at faster speeds while you take full time
(don't want to wait for its moves). In this case, you subtract 100
points for every 50% reduction in time. You didn't say whether you own
one or are considering one. If you are only considering, you would
probably do better with the Fidelity 2000, which costs only about
$100 (and is rated about 1951).
--Jim
Ali Engin Aydar
That Radio Shack 1650 is no where near 1650. My USCF rating is
1372 and I can beat it every time......I would give it about a
1200 rating....but I am not sure. Perhaps I get lucky. Don't
know...and I am in now way putting this computer down. It can
get a lot of novices to an intermideate level.
--
Ali Aydar
bw...@cleveland.freenet.edu
Peter Marko
j...@m2lll.uucp writes:
>The 2150L is about 1950. The 1650L is not worth rating. It might be
>good for a child. You should get something at least 200 points
>stronger than you so that you can improve and still use it. You will
>also probably want to play it at faster speeds while you take full time
>(don't want to wait for its moves). In this case, you subtract 100
>points for every 50% reduction in time. You didn't say whether you own
>one or are considering one. If you are only considering, you would
>probably do better with the Fidelity 2000, which costs only about
>$100 (and is rated about 1951).
>
>--Jim
Jim,
Although I am not up on computer chess, your comment about "subtract[ing]
100 points for every 50% reduction in time" intrigued me. How did you
arrive at this rule of thumb?
According to Mr. Elo, one class (200 points) difference amounts to 3.17
times playing strength, so 100 points would equal the square root of
that, or 1.78. This is not far from a factor of two. About 120 points
of difference gives twice (or half) the power.
In another article, you give Zarkov's tournament (40 moves in 2 hours)
performance at 2300, and its blitz (60 moves in 5 minutes) performance at
about 1800 (on a certain machine at a certain speed). (This is from
memory, so please excuse me if the numbers are incorrect.) Given 36
times more time (180 minutes vs. 5 minutes for 60 moves), the difference
in theory is 620 points, so your estimate (500) is pretty good.
The other question is, of course, whether giving a computer twice the
speed would improve its play by a factor of two.
...Peter
-------------------------------------------------------
Peter Marko ..... (613)763-5344 ..... pma...@bnr.ca
Ottawa, Ontario, Canada
Derek Nelson, xt-dL
claim to have beaten the other.
Eric S. Perlman
Which version of Sargon? Sargon II was about 1300, Sargon III around
1800, not sure about Sargon IV.
> Of the PC based chess games which is the best? They all
>claim to have beaten the other.
According to Computer Chess Reports, which is the best source for
answering such a question, the strongest PC program (IBM compatibles) is
M-chess, with Zarkov 2.6 #2 but about 50-80 points lower-rated. M-chess
plays at about 2400 strength, and for those who own 386 or 486
computers, a new version, called M-chess professional will be coming out
in a month or two, which is supposed to about 80 points stronger than
M-chess.
Thanks for asking.
--
"How sad to see/A model of decorum and tranquillity/become like any other sport
A battleground for rival ideologies to slug it out with glee." -Tim Rice,"Chess"
Eric S. Perlman <per...@qso.colorado.edu>
Center for Astrophysics and Space Astronomy, University of Colorado, Boulder
James H. Coombs
>What is Sargon rated at? Of the PC based chess games which is the
>best? They all claim to have beaten the other.
I see a Sargon ARB 2.0 (MGSII) at 1645. I don't see anything else on
Sargon.
"MChess consistently dominates every program except Zarkov." After
MChess and Zarkov, it is Rexchess, KnightStalker. and Psion.
Best is to read CCR from ICD at 800-645-4710. --Jim
James H. Coombs
>Although I am not up on computer chess, your comment about "subtract[ing]
>100 points for every 50% reduction in time" intrigued me. How did you
>arrive at this rule of thumb?
I got it from ICD's catalog. I don't recall seeing it in CCR, but it
may be in one of those issues as well. I don't whether their
derivation is mathemetical or experiential.
>According to Mr. Elo, one class (200 points) difference amounts to 3.17
>times playing strength, so 100 points would equal the square root of
>that, or 1.78. This is not far from a factor of two. About 120 points
>of difference gives twice (or half) the power.
I'm not mathematically sophisticated. Would you explain (or cite the
para. in Elo) why you take the square root of 3.17 to arrive at the
relative playing strength when you cut the rating difference in 1/2?
Also, are you suggesting that the proper figure would be 120 and not
100?
>In another article, you give Zarkov's tournament (40 moves in 2 hours)
>performance at 2300, and its blitz (60 moves in 5 minutes) performance at
>about 1800 (on a certain machine at a certain speed). (This is from
>memory, so please excuse me if the numbers are incorrect.) Given 36
>times more time (180 minutes vs. 5 minutes for 60 moves), the difference
>in theory is 620 points, so your estimate (500) is pretty good.
I just used my fingers and went in the reverse direction:
Secs per move Rating
180 2300
90 2200
45 2100
22.5 2000
11.25 1900
5.625 1800
>The other question is, of course, whether giving a computer twice the
>speed would improve its play by a factor of two.
ICD says increasing the time per move above 180 seconds does not
significantly increase the performance of the program. Just from
watching programs on machines of very different power, I can see that
the they reach a state of computational explosion. Basically, you find
that 1/2 an hour later, it is still working on the move that it reached
after 5 minutes or so. (And I don't mean "same ply"; I mean same
candidate move.)
--Jim
Richard Nash
2 min/move, I beat it regularly while watching TV. I am rated 1248 USCF.
To add chalange and practice to my games, I force myself to move _imediately_
after the computer moves. It forces me to calculate all of my responses to
whatever the computer does on its time. Even while doing this I beat it more
often than I lose.
I doubt the computer is rated 1650.
Rich
--
Richard V. Nash
na...@vi.ri.cmu.edu
Peter Marko
j...@m2lll.uucp writes:
>I'm not mathematically sophisticated. Would you explain (or cite the
>para. in Elo) why you take the square root of 3.17 to arrive at the
>relative playing strength when you cut the rating difference in 1/2?
>
>Also, are you suggesting that the proper figure would be 120 and not
>100?
>
>--Jim
Jim,
The Elo rating system uses a so-called INTERVAL SCALE which means that a
certain RATIO in playing strength (like 2) translates to a certain
DIFFERENCE on this scale (120 points in this case). Please see paragraph
8.71 in Elo's book.
And yes, I am suggesting that twice the strength is equivalent to 120
points. Ten times strength translates to two class differences, or 400
points.
James H. Coombs
>To add chalange and practice to my games, I force myself to move _imediately_
>after the computer moves. It forces me to calculate all of my responses to
>whatever the computer does on its time. Even while doing this I beat it more
>often than I lose.
Perhaps you should try giving it odds. --Jim
James H. Coombs
>In article <1992Sep15.2...@cs.brown.edu> James H. Coombs,
>j...@m2lll.uucp writes:
>>Also, are you suggesting that the proper figure would be 120 and not
>>100?
>And yes, I am suggesting that twice the strength is equivalent to 120
>points. Ten times strength translates to two class differences, or 400
>points.
I can only offer that CCR/ICD must have concluded that reducing the
time by 50% does not reduce the strength by 50%. I don't have any
details. --Jim
CUFFELL TIMOTHY MICH
>100 points for every 50% reduction in time" intrigued me. How did you
>arrive at this rule of thumb?
>
>According to Mr. Elo, one class (200 points) difference amounts to 3.17
>times playing strength, so 100 points would equal the square root of
>that, or 1.78. This is not far from a factor of two. About 120 points
>of difference gives twice (or half) the power.
>
>In another article, you give Zarkov's tournament (40 moves in 2 hours)
>performance at 2300, and its blitz (60 moves in 5 minutes) performance at
>about 1800 (on a certain machine at a certain speed). (This is from
>memory, so please excuse me if the numbers are incorrect.) Given 36
>times more time (180 minutes vs. 5 minutes for 60 moves), the difference
>in theory is 620 points, so your estimate (500) is pretty good.
>
>The other question is, of course, whether giving a computer twice the
>speed would improve its play by a factor of two.
>
> ...Peter
>
and the Elo rating increases logrithmically with playing ability.
James H. Coombs
>>The other question is, of course, whether giving a computer twice the
>>speed would improve its play by a factor of two.
>It should. This computer's play increases logrithmically with with time/speed,
>and the Elo rating increases logrithmically with playing ability.
Well, we are varying only time here, not speed. That is my
interpretation from the context. Did I misunderstand? If not, then
what about the degeneration that occurs as search space grows,
culminating in computational explosion? --Jim
Peter Marko
j...@m2lll.uucp writes:
> In article <1992Sep16....@bnr.ca> pma...@bnr.ca (Peter Marko)
writes:
>> And yes, I am suggesting that twice the strength is equivalent to 120
>> points. Ten times strength translates to two class differences, or 400
>> points.
>
> I can only offer that CCR/ICD must have concluded that reducing the
> time by 50% does not reduce the strength by 50%. I don't have any
> details. --Jim
Jim,
According to rec.games.chess FAQ (which was just posted), a factor of two
in computer speed (or "Chess MIP") is equivalent to about 75 points (it
varies between 67 and 90):
> Processor Chess MIPs
>
> 8088 Speed in MHz divided by 19
> 80286, 1 wait state Speed in MHz divided by 8
> 80286, 0 wait states Speed in MHz divided by 6
> 80386, no Cache memory Speed in MHz divided by 6
> 80386 with Cache Speed in MHz divided by 4.7
> 80486 Speed in MHz divided by 2.3
>
> Now, if a program has a given rating on a "1 Chess MIP" machine, this
> is how to adjust the rating for other MIPs (interpolate between
points):
> MIP: 0.25 0.5 1 1.5 2 3 4 6 8 12 16 24 32 48 64
> Adj.: -180 -87 0 47 80 124 154 195 223 261 287 323 347 379 402
So you are right, reducing the speed by 50% does not reduce the strength
by 50%, but only by about 33%. In other words, a factor of 2 in computer
speed seems to correspond to a factor of 1.5 in playing strength (or 75
points).
James H. Coombs
>In article <1992Sep16.1...@cs.brown.edu> James H. Coombs,
>j...@m2lll.uucp writes:
>> In article <1992Sep16....@bnr.ca> pma...@bnr.ca (Peter Marko)
>>> And yes, I am suggesting that twice the strength is equivalent to 120
>>> points. Ten times strength translates to two class differences, or 400
>> I can only offer that CCR/ICD must have concluded that reducing the
>> time by 50% does not reduce the strength by 50%. I don't have any
>According to rec.games.chess FAQ (which was just posted), a factor of two
>in computer speed (or "Chess MIP") is equivalent to about 75 points (it
>varies between 67 and 90):
I am talking about varying time while holding the machine constant, and
you are talking about different machines. I have not paid much
attention to comparing machines and have only used CCR's figures with
my estimation where relevant. I know that they add 60 points to a
386/33 to get a 486/33 value. They have discussed machine speed
in various articles.
I just want to make it clear that the "rule" of deducting 100 points
for each 50% reduction in allotted time per move---that rule applies to
just that---cutting the program back from 180 seconds per move to 90
seconds, etc. Note that giving 360 seconds per move does not add 100
points to the playing strength (according to ICD), but I suppose that
using a computer that is twice as fast and giving it the same amount of
time might well add 100 points, or more, or less. FAQ says about 75?
Ok, by me. That's still different from reducing the time per move from
180 to 90....
--Jim
CUFFELL TIMOTHY MICH
Time and speed are the same thing. Getting a computer that is twice fast is
the same as letting the slow one think for twice as long. As for computational
explosion, that is what exponenetial/logrithmic functions model. Let us
suppose a computer needs to evaulate 10 positions (because ten is round, no
accurate) per move. Assume the number of moves a given algorithm looks
ahead determines its play strength. One move ahead requires 10 evaulations,
two moves, 100, three, 1000, and so forth. Doubling the time doubles the
number of positions evaluated, but only increases the depth by log10(2) =
.32... times.
As for the Elo rating, I was mistaken. For some reason I thought both
logrithmic scales cancelled each other out. They don't. First of all, the
Elo scale is an arbitrary one that happens to have logrithmic properties.
This may render mathematical comparisons innacurate. Even so, the
logrithmic properties do not cancel, the add, i.e., play strength increase
logrithmically (read "very slowly") with computational power, and the Elo
rating grows logrithmically with play strength.
The moral of the story is that computers can play at Elo 2350, this is not
as close to the 2800 for humans as it might seem. Doubling computational
power should translate into a very small Elo jump. The FAQ same that my
20Mhz 386 plays 148 points better that an XT, even though various tests tell
me that it runs thirty times faster.
What's more, computational speeds are rapidly approaching various physical
barriers, like the speed of light, that prevent endless increases in speed.
Current technologies are running out, so improvements will have to come on
the software end.
I think the computer world champions will be carbon based well into the next
century. They can't even beat us in checker yet.
James H. Coombs
>In article <1992Sep16.1...@cs.brown.edu> j...@iris.brown.edu.UUCP (James H. Coombs) writes:
>Time and speed are the same thing. Getting a computer that is twice fast is
>the same as letting the slow one think for twice as long.
Is it really? Then why do "underpowered" machines start thrashing?
Perhaps that is not a problem for cpu-intensive applications on
single-user machines.
I think the reality of it is that one does not generally work with two
machines that differ in nothing but MHz. They differ also in the
amount of memory, the speed of the channels, the availability of
caches, etc. It would not tell us much about the real world to hold
all of these things constant or to say "twice as fast at running this
program". In my experience, at least, there is a threshold at which
giving lots of additional time yields an insignificant gain in work
completed.
>As for computational
>explosion, that is what exponenetial/logrithmic functions model. Let us
>suppose a computer needs to evaulate 10 positions (because ten is round, no
>accurate) per move. Assume the number of moves a given algorithm looks
>ahead determines its play strength. One move ahead requires 10 evaulations,
>two moves, 100, three, 1000, and so forth. Doubling the time doubles the
>number of positions evaluated, but only increases the depth by log10(2) =
>.32... times.
Well, doubling the time may double the number of positions evaluated up
to a point, but then you have to assume unlimited RAM for holding this
growing tree. Also, you need to take into account the overhead of tree
traversal. At some point, the program would do little more than
traversal. So, we prune. But then there goes that nice assumption
about holding constant the number of moves to be evaluated for each
position.
>The moral of the story is that computers can play at Elo 2350, this is not
>as close to the 2800 for humans as it might seem. Doubling computational
>power should translate into a very small Elo jump. The FAQ same that my
>20Mhz 386 plays 148 points better that an XT, even though various tests tell
>me that it runs thirty times faster.
I can only repeat CCR's claim that halving the time cuts 100 points.
Also, the claim that increasing the time beyond 3 minutes does not
significantly improve the playing strength. This type of analysis is
foreign to me, best to talk to them directly. A final thought from a
common-sense perspective, 3 minutes is "a lot of time", and these
programs may just start "thrashing" after that. I know that people run
programs for months, perhaps even years, that are going to tell us the
nature of the universe, but then I don't know the properties of their
algorithms either.
Ok, finally, finally. We are trying to analyze algorithms here without
knowing the algorithms. The hardware is only part of the issue, and
fast hardware does nothing but bring an algorithm to explosion more
quickly than slow hardware.
--Jim
j...@iris.brown.edu -- don't trust reply-to fields
na...@pa881a.inland.com
>
> Which version of Sargon? Sargon II was about 1300, Sargon III around
> 1800, not sure about Sargon IV.
>
The Sargon V (DOS) manual claims that its rating exceeds 2300 USCF! They give
no reference on what hardware this would be on or what level. I am assuming
tournament mode and a 486. I don't stand behind the claim and am just
providing the quote.
--
My opinions are my own and not those of my company.
Free advice is worth its cost.
--- John M. Nason - NA9U ---|--------------------------------------------|
Inland Steel Company | -- na...@pa881a.inland.com |
Process Automation MS 2-464 | -- Fax: 219-399-7856 |
3210 Watling Street | -- Voice: 219-399-2450 |
East Chicago IN 46312 |"Ya can't shoot em till they stop sparklin" |
----------------------------|--------------------------------------------|
Paul Hsieh
>In article <1992Sep15.2...@colorado.edu>, per...@qso.Colorado.EDU (Eric S. Perlman) writes:
>>
>> Which version of Sargon? Sargon II was about 1300, Sargon III around
>> 1800, not sure about Sargon IV.
>>
>
>The Sargon V (DOS) manual claims that its rating exceeds 2300 USCF! They give
>no reference on what hardware this would be on or what level. I am assuming
>tournament mode and a 486. I don't stand behind the claim and am just
>providing the quote.
>
2300?!?!?! They *MUST* be joking ... I am about 1800 USCF and have never lost
to it in tournament mode. My guess is more like 2000 or so (I attribute my
success to being good at the "David Levy" method for beating computers). BUt
seriously, don't let such outlandish claims mislead you would be Sargon
buyers out there. And if you want a hint as to how to crack it as white, play
1. d4 and it will likely play one of the Dutch which no-one should have too
much of a problem with or one of, the kings indian and the gruenfled both off
which it totally misplays.
Paul
pa...@ipact.com
>
> The Sargon V (DOS) manual claims that its rating exceeds 2300 USCF! They give
> no reference on what hardware this would be on or what level. I am assuming
> tournament mode and a 486. I don't stand behind the claim and am just
> providing the quote.
Incredible. Anytime you see excrement like this printed by a manufacturer
on their product, take it as a big red flag that says "Don't buy me!" Not
only are the level (time control) and especially the hardware critical to
the program's performance, but the conditions under which the program was
rated. Is it playing blitz against exhausted humans? Or a one-round per
day tournament against humans at 3 minutes a move?
As for Sargon's rating being > 2300 USCF - on a Cray, 5 minute games,
vs. sleep deprived humans.
--
Randy Pals |pa...@ipact.com
IPACT, Inc. |
na...@pa881a.inland.com
> In article <1992Sep18...@pa881a.inland.com>,
> na...@pa881a.inland.com (John Nason) writes:
>>
>> The Sargon V (DOS) manual claims that its rating exceeds 2300 USCF!
>
> Incredible. Anytime you see excrement like this printed by a manufacturer
> on their product, take it as a big red flag that says "Don't buy me!" Not
> only are the level (time control) and especially the hardware critical to
> the program's performance, but the conditions under which the program was
> rated. Is it playing blitz against exhausted humans? Or a one-round per
> day tournament against humans at 3 minutes a move?
>
> As for Sargon's rating being > 2300 USCF - on a Cray, 5 minute games,
> vs. sleep deprived humans.
Unfortunatley, they don't put that statement on the box. It's buried in the
manual on page 9. I had choice between Sargon V and Chessmaster 3000. Too bad
for me I picked Sargon V. For anyone following this thread, *DON'T* buy this
package. I made a previous post sighting other problems with it, ie:
must reboot after a stalemate!
1 d4 or 1... d5 translates to P-K4 in descriptive notation????
unpleasant deviations from standard openings, tough on novices trying to learn.
virtually non-existant documentation.
Will not import saved games if you saved them in ASCII. though it claims itwill
I am still trying to figure out how to put some of the FS_II games in, without
playing them move by move!
etc.......
Get the point!