Hi Micha,

Ed in the gideon pro UCI/WB did not the use hashtable from 1993 he use the hash from prodeo. Maybe if he puts the hash from that time we will have a 100% conversion or very close one.

new old hash tables regards
Ricardo

Hi Ricardo,

I still don't think so. You cannot make a judgement based on one Test Position.

You work for a living right? If so then every week you have produced something in your 40 hours of work.

A professional programmer is the same they are always working on improvements, hence your versions 2.1, 2.1a etc etc.. When converting Gideon Pro to PC the same thing happened, a hash change here, maybe more endgame there, a revised opening book etc So many things could have changed.

I am not even 100% sure that R30 Gideon 3.1 is 100% the same as Chessmachine 3.1. If you look on the R30 it says version 2.1 of the Gideon 3.1. This means ED had to make several changes to get to version 2.1 to make it work best on an R30. It could be Hardware compatibility adjustments, it could be piece recognition, it could be a thousand little changes.

So for Gideon Pro to be 100% identical even without the new Hash would almost be impossible since they did not come out at exactly the same time. Even if they did come out at the same time they would still not be 100% the same.

Best regards,

Nick

Hi Nick,

When writing a software for an engine there are 3 main tasks that give its strength and style of play:

1) Search
2) Evaluation
3) hash table

of course there are others. I do agree that gideon pro is not 100% the same but could be maybe 90% that is why i also say close but this has to be test to see how close it could be. Sadly i can not do that because i only have the gideon pro UCI. Nevertheless I had include in my to do list test the strenght of this gideon pro UCI.

Best Regards
Ricardo

Right, but I still don't understand what you are asking to be confirmed? Your posts keep looking for confirmation that Gideon 3.1, Gideon Pro and Gideon Pro UCI are the same after you change the Hash on Gideon Pro UCI version back to the original Gideon Pro Hash. All you will have then is still a modified UCI version of the Original Gideon Pro but not Gideon 3.1.

Ed could have confirmed Gideon Pro and Gideon 3.1 are identical, but he did not, he just pointed out that he made a big mistake in the Hash when he converted Gideon Pro to Gideon Pro UCI. There is no mention of Gideon 3.1. He also indicated that other minor changes were made to Gideon Pro UCI in the conversion. Therefore even with a Hash correction Gideon Pro UCI would not be the same as Gideon Pro DOS.

Under evaluation for example if two positions are calculated the same, one is a Fork and one is not. How would a program choose? Would it always choose a Fork because the programer prioritized this as the way to go or no?. Or is it Random choice. Or even always the first evaluation? There is a million subtle small changes in Evaluation alone that make the programs personality or ability different to a previous same version. Don't you think that programmers whenever they made a conversion, picked their last version of their software. Therefore it's impossible to pick Gideon 3.1 version 1 for example 3 months later when in front of the programer lies improvement version number 105, which in this case he would name it Gideon Pro instead of Gideon 3.1 for DOS.

In R30 you have 5 Play styles for Gideon 3.1 to choose. The Chessmachine has 3 play styles. Which of the 5 play styles compares to the 3 styles available to the Chesmachine? Howabout Gideon Pro how many play styles do you have. The standard Gideon Pro play style compares to which Gideon 3.1 if the assumption is that they are the same?

Kittinger's versions A, B and C are probably 99% the same but its some small changes that give people the feeling that they are different.

Chessmachine Gideon 3.1 has been thoroughly played and there are plenty of game examples that could be used for testing. I am not sure if Micha has any BT Tests for Chessmachine Gideon 3.1 but they might be useful for you to do some more comparisons.

As for R30 Gideon 3.1, you can be sure that this program will be played, tested and reported on extensively.

All the best,

Nick

Hi Ricardo,

Ed wrote the following in the Rybka forum:


I estimate it's 98-99% equal. It wasn't a time of improving the engine as Hegener & Glaser went into bankruptcy. It was either return to a normal livelihood or join the new future, the PC. So Gideon Pro was a hasty translation job which took me 8-9 months or so.

So he confirms that they are similar but I don't think that it is possible for it to be the same. 8-9 months translation from platform to another would make it almost impossible for it to be identical. The programs were made 20 years ago.

Regards,

Nick

Hi Nick,
Is clear for me now that the 3 engines are not the same (gideon 3.1, gideon pro and gideon pro UCI). The next question is how close they are? but that is another story. For this have to look for bt-tests in the different styles and probably have to come up with a similar test as you did for the gk-2100 clones and again with the different styles.

Best Regards
Ricardo

Hallo,

das Wort BT-Test ist gefallen, eine gute Gelegenheit diese zu präsentieren.

Wie erwartet, fallen die Ergebnisse etwas enttäuschend aus. Ich hatte es aber auch nicht anders erwartet, da der Risc 2 im BT-2630 mit 2203 vertreten ist.

Interessant ist in diesem Zusammenhang die Aufgabe 29 im BT-2630. Diese wird durch den Risc 2 nach 1,5 Minuten gelöst, während sich der R30 Gideon 3.1 auch nach 1 Stunde nicht zum Lösungszug durchringen kann. Wer nun denkt, die CM 32 MHz Gideon 3.1 würde diesen Zug ebenfalls nicht finden, der irrt. Sie findet ihn nach 39 Sekunden...

BT 2630	R30 Gideon 3.1	Risc 2	BT 2450	R30 Gideon 3.1
BT1	121	216	BT1	121
BT2	68	82	BT2	68
BT3	11	23	BT3	11
BT4	900	900	BT4	900
BT5	900	900	BT5	900
BT6	35	59	BT6	35
BT7	3	10	BT7	3
BT8	652	900	BT8	652
BT9	451	900	BT9	451
BT10	15	31	BT10	15
BT11	2	8	BT11	2
BT12	211	240	BT12	2
BT13	900	900	BT13	14
BT14	32	65	BT14	32
BT15	900	900	BT15	0
BT16	900	900	BT16	1
BT17	50	150	BT17	50
BT18	17	29	BT18	17
BT19	46	78	BT19	46
BT20	9	20	BT20	9
BT21	900	900	BT21	16
BT22	6	15	BT22	6
BT23	900	900	BT23	1
BT24	900	900	BT24	900
BT25	900	900	BT25	900
BT26	900	3	BT26	900
BT27	668	900	BT27	668
BT28	900	900	BT28	7
BT29	900	71	BT29	0
BT30	11	20	BT30	11
------	------	------	------	------
Gesamt	13208	12820	Gesamt	6947
BT-Wert	2190	2203	BT-Wert	2218

Gruß
Micha

Hallo Micha,

meine CM Gideon 3.1 findet den Schlüsselzug (0-0-0) in der Stellung 29 des BT-2630-Tests innerhalb von 15min nicht. Hab mal CM Gideon 3.1 und R30 Gideon parallel laufen lassen. Bester Zug (Lxf7), Bewertung und Hauptvariante sind identisch. Der Schlüsselzug erscheint bei beiden Programmen auch nie als bester Zug in der Hauptvariante. Unterschiede gibt es in der angezeigten Rechentiefe nach 15min ( CM: 4/9; R30: 3/8) und der Position des Schlüsselzuges in der Zugliste (CM: 6/51; R30: 5/51).

Gruß Wolfgang

Hallo Wolf,

das ist überraschend. Ich habe die Aufgabe soeben noch einmal getestet, meine CM 3.1 findet den Zug. OK, zweite Variante. Teste doch bitte mal die Aufgabe 5 (Ka6) mit deiner CM. Wie lange benötigt sie dafür?

Gut, die Rechentiefe wäre logisch, schließlich ist die CM etwas schneller getaktet.

Und zweite Frage. Welches Datum trägt deine CM Gideon 3.1 Version?

Gruß
Micha

Hallo Micha,

Die CM findet Ka6 nach 43 sek. Tiefe 7/10 Bewertung +3,69. Sie wechselt nach ca 3min auf Kc5 Tiefe 9/12 Bewertung +3,77. Nach 7 min 21 wieder Wechsel auf Ka6 Bewertung +3,76 Tiefe 10/13. Die Version ist vom 11.12.1992.

Gruß Wolfgang