Atlanta/Magellan ...Redux

[Steve B]

Hi Nick
with all of these different test threads ..i thought i would set up a separate thread just for the Atlanta/Magellan discussion
although mildly interested in the other discussions my real interest is in the Atlanta/Magellan as it relates to these new testing methods of playing through games ,or the BT suites or whatever else you are using

please ..when ready...
post the results of your comparison between the Atlanta and Magellan here...
please just post the % of moves that each played exactly the same
and the the TOTAL number of moves played
in the 5-6 games you played
as an example..
the exact matches were .XX% out of a total of xxx moves played

we can then discuss the conclusions of your test

Best Regards
Steve

[spacious_mind]

Hi Nick
with all of these different test threads ..i thought i would set up a separate thread just for the Atlanta/Magellan discussion
although mildly interested in the other discussions my real interest is in the Atlanta/Magellan as it relates to these new testing methods of playing through games ,or the BT suites or whatever else you are using

please ..when ready...
post the results of your comparison between the Atlanta and Magellan here...
please just post the % of moves that each played exactly the same
and the the TOTAL number of moves played
in the 5-6 games you played
as an example..
the exact matches were .XX% out of a total of xxx moves played

we can then discuss the conclusions of your test

Best Regards
Steve

Hi Steve,
Thanks, I will start the games tomorrow night Central time when I get home from work. I will be doing a rating test, in other words adding both Atlanta and Magellan to the 5 Test games that I have posted, set at their best settings since they are being rated. No Ponder Off, brute force or anything like that.

In other words Selective and Ponder On just how they were made to play. 30 Seconds per move just like the rest of the tested computers.

Best regards,

Nick

[Larry]

Someone on another forum, I think it was schachcomputer.info,
mentioned that the early Magellan modules were 16mhz. The later ones
were 20mhz, same as Atlanta. This could explain the persistent rating
difference between the two.
Not being a Magellan owner, I can't check personally. Maybe a good
idea is for Magellan owners to slip the casing open on theirs and check
the reading on the crystal oscillator. Assuming it is a 20mhz unit, then
you can start comparison testing, otherwise you are comparing apples
with oranges.
nice day all...
Larry

[Steve B]

Someone on another forum, I think it was schachcomputer.info,
mentioned that the early Magellan modules were 16mhz. The later ones
were 20mhz, same as Atlanta. This could explain the persistent rating
difference between the two.
Not being a Magellan owner, I can't check personally. Maybe a good
idea is for Magellan owners to slip the casing open on theirs and check
the reading on the crystal oscillator. Assuming it is a 20mhz unit, then
you can start comparison testing, otherwise you are comparing apples
with oranges.
nice day all...
Larry

I think you are right and the difference in the Wiki ratings although very small is probably due to a mixture of 16Mhz computers in the rating pool..

the problem is the separate Wiki listing for these computers show the same exact Specs..20 Mhz
if there was a separate listing showing for example:
Magellan/Atlanta 20 Mhz
Magellan 16 Mhz
Then I would have no issue here
the decision to list both 20 Mhz computers separately was based on the notion that both 20 Mhz computers play a statistically significant different number of moves in the same position..hence they are not exactly the same program although with the same exact hardware
I have challenged this notion from day 1
I have compared the two computers with pondering OFF and found not one variation in moves
not to mention all of the literature,and advertising of the day ..Saitek/Mephisto Brochures ..etc etc all calling them the same

lets see the results of Nicks Games

Anticipatory Regards
Steve

[spacious_mind]

Here are the results of Test Game 5:

• ------ATLANTA------ ------MAGELLAN------
  13. … c5 2.0 13. … c5 2.0
  14. … Nc6 2.5 14. … Nb7 0.0 X
  15. … Nc6 3.0 15. … Bc6 1.5 X
  16. … Nxd5 3.0 16. … Nxd5 3.0
  17. … Nd4 2.6 17. … Nd4 2.6
  18. … Bh3 2.2 18. … Bh3 2.2
  19. … Bxg2 3.0 19. … Bxg2 3.0
  20. … Nxc4 3.0 20. … Nxc4 3.0
  21. … Bxb2 3.0 21. … Bxb2 3.0
  22. … Rxb2 3.0 22. … Rxb2 3.0
  23. … Qg4 3.0 23. … Qg4 3.0
  24. … Rb8 3.0 24. … Rb8 3.0
  25. … h6 1.2 25. … g5 0.0 X
  26. … Qh5 2.1 26. … Qd7 2.6 X
  27. … Qe5 3.0 27. … Qe5 3.0
  28. … Rxb1 2.0 28. … Rxb1 2.0
  29. … Qc3 3.0 29. … Qc3 3.0
  30. … a3 2.0 30. … a3 2.0
  31. … Qb2 3.0 31. … f5 0.0 X
  32. … h5 3.0 32. … h5 3.0
  33. … h4 2.0 33. … h4 2.0
  34. … gxh4 2.0 34. … gxh4 2.0
  35. … Qa1+ 2.6 35. … Kg7 3.0 X
  36. … Qh2+ 3.0 36. … Qh2+ 3.0
  37. … Qg5+ 1.1 37. … Qg5+ 1.1
  38. … Qd4+ 1.1 38. … Kf7 2.6 X
  39. … Qh2 0.0 39. … Qh2 0.0
  40. … Qg7 1.0 40. … Qg7 1.0

7 out of 28 variations = 75% match and 25 % mismatch.

Of course all the rest of the test games will be played as well. Both were turned on checked for Selective ON, set 30 seconds per move and play. I just have difficulty in matching my Magellan and Atlanta in Selective Mode.

@ Steve this was the cause of me spending all that time back then debating this! It is still the same today.

Best regards,

Nick

[Reinfeld]

So how does ponder on vs off work with these machines? I don't own either one (wish I did) and I don't understand the use of this term.

In the 2250XL Kaplan/Morsch thread, Steve said the only way to turn ponder OFF on the 2250XL is to turn off the machine after the opponent makes its move. Is that the way it works here?

If so, does "ponder" refer to thinking on the opponent's time? That would make sense to me. I know it's possible to shut that feature down on many machines.

If all this is right, then I would agree with Steve that the *purest* measure (for clone testing) would be ponder OFF. Otherwise, all sorts of interruptions could disrupt the test.

- R.

[IvenGO]

For disabling ponder on RS2250XL set the option EASY = ON

And by the way my proposal for clone-test:

A) You need to play 2 games between 2 computers with the same opening line for both sides (white/black) each computer;
B) You need to play yourself a simul on 2 boards with the same opening line at both.


Settings for A and B tests:
1) disable pondering for both computers;
2) set casual level (X-min per move);
3) set Selective = ON and Random = OFF;
4) select the line with same length at both computers, if one has smaller book put the moves on it manually - so they must start to play themselves from the same position at both A and B tests.

This way real clones will have very close to 100% (in theory) same moves; others could be claimed as modified software...

[Steve B]

For disabling ponder on RS2250XL set the option EASY = ON

this would be right if there is this option in the 2250XL
I didn't see this option in my 2250XL manual so I thought there was no way to turn pondering OFF
strangley when I look at the online manual is does show this option
I will have to check my manual and computer again
Thanks Regards
Steve

[Steve B]

@ Steve this was the cause of me spending all that time back then debating this! It is still the same today.

Waiting for all 5 games
if the mismatch rate stays at around 25% for all 5 games then this cannot be due to pondering being left ON and I will concede the point
pondering ON I think could account for a mismatch rate of about 10-15% not 1/4 ...unless of course you walk away from the chess board every other move


Work In Progress Regards
Steve

[Steve B]

If so, does "ponder" refer to thinking on the opponent's time? That would make sense to me. I know it's possible to shut that feature down on many machines..

Yes that's what "ponder" means

If all this is right, then I would agree with Steve that the *purest* measure (for clone testing) would be ponder OFF. Otherwise, all sorts of interruptions could disrupt the test.

Exactly Regards
Steve

[spacious_mind]

For disabling ponder on RS2250XL set the option EASY = ON

And by the way my proposal for clone-test:

A) You need to play 2 games between 2 computers with the same opening line for both sides (white/black) each computer;
B) You need to play yourself a simul on 2 boards with the same opening line at both.


Settings for A and B tests:
1) disable pondering for both computers;
2) set casual level (X-min per move);
3) set Selective = ON and Random = OFF;
4) select the line with same length at both computers, if one has smaller book put the moves on it manually - so they must start to play themselves from the same position at both A and B tests.

This way real clones will have very close to 100% (in theory) same moves; others could be claimed as modified software...

The easiest way I think is just to take any of the 3 games that are on my website and play them at 3 minutes per move or just take both computers and let them play against each other once with white and once with black and then replay the games. Then these games were played by a clone, so another clone would have the best chance repeating the same move. 3 minutes per move I think would be better for a clone test because it allows the programs just that little more time to settle into their calculations.
Also turning everything off and just playing Brute Force will most likely repeat more moves than a selective setting.

I think it is Selective setting that allows us to continue to debate clones, but then again it is selective which is the strongest setting so therefore debating a justification to see them split in a list is easier to argue because that is how we play them and a split can be justified.

Regards
Nick

[Steve B]

I think it is Selective setting that allows us to continue to debate clones, but then again it is selective which is the strongest setting so therefore debating a justification to see them split in a list is easier to argue because that is how we play them and a split can be justified.

Regards
Nick

I have no issues with Selective search Vs Brute Force
I also have no issues with a short 30 Sec. per move time control
when I performed my tests years ago I used
Selective Search
Random OFF
Pondering OFF

I got a 100% match

as I mentioned above if the mismatch rate stays at around 75% I will concede the separate listing on Wiki

there is no way your Magellan runs at 16 Mhz right?
I think we cleared that point up when we debated this back then?

Regards
Steve

[spacious_mind]

@ Steve this was the cause of me spending all that time back then debating this! It is still the same today.

Waiting for all 5 games
if the mismatch rate stays at around 25% for all 5 games then this cannot be due to pondering being left ON and I will concede the point
pondering ON I think could account for a mismatch rate of about 10-15% not 1/4 ...unless of course you walk away from the chess board every other move


Work In Progress Regards
Steve

Hi Steve,
Since these games are so short there is little need to go take a break . I had two ponder hits by Magellan and both were in situations where the move was obvious, ie where every other computer also played the same move and not just the clones. I think from what I have seen so far that ponder at 30 seconds per move might only have a very slight influence (if any at all) because of the many take back moves that are also occurring which always causes the programs to restart their calculations. At a longer time setting perhaps there may be a slightly larger effect but I am not sure.

Ps. Yes I think we checked all that back then too. But I will open mine again and recheck what MHz is on the Quartz. I am also going to add Senator to ensure that the MFR didn't perhaps swap a ROM.

Best regards,

[Steve B]

@ Steve this was the cause of me spending all that time back then debating this! It is still the same today.

Waiting for all 5 games
if the mismatch rate stays at around 25% for all 5 games then this cannot be due to pondering being left ON and I will concede the point
pondering ON I think could account for a mismatch rate of about 10-15% not 1/4 ...unless of course you walk away from the chess board every other move


Work In Progress Regards
Steve

Hi Steve,
Since these games are so short there is little need to go take a break . I had two ponder hits by Magellan and both were in situations where the move was obvious, ie where every other computer also played the same move and not just the clones. I think from what I have seen so far that ponder at 30 seconds per move might only have a very slight influence (if any at all) because of the many take back moves that are also occurring which always causes the programs to restart their calculations. At a longer time setting perhaps there may be a slightly larger effect but I am not sure.

Ps. Yes I think we checked all that back then too. But I will open mine again and recheck what MHz is on the Quartz.

Best regards,

Thanks Nick
again I have no issues with time controls OR Sel.Search Vs Brute Force
if the mismatch rate stays that high I cannot in good faith attribute that to Pondering ON

However if the mismatch rate drops to about 10-15% then in my mind we have clones

Pondering On Regards
Steve

[Steve B]

For disabling ponder on RS2250XL set the option EASY = ON

this would be right if there is this option in the 2250XL
I didn't see this option in my 2250XL manual so I thought there was no way to turn pondering OFF
strangley when I look at the online manual is does show this option
I will have to check my manual and computer again
Thanks Regards
Steve

OK
i re-checked the manual and the computer
you are right
there is an EASY ON option which will turn Pondering OFF

My Bad Regards
Steve