Deep Fritz 11 eval

[hyatt]

Where do you see that? I only see Fritz 12 rated at 2920, Fritz 11 rated at 2898 and deep fritz 10 rated at 2866. No deep fritz 11 there that I see...

[Damir Desevac]

Again, you are not reading what I write. I only mentioned single Fritz 11 beeing ahead of MP engines. Nowhere did I mention DF 11...

[BB+]

I think these are the PSTs of DF11. I don't know what the scaling is. Note that f8 is -8, for instance. These are accessed with a1=0, b1=1, etc.

Code: Select all

00 00 00 00 00 00 00 00 # 10129290
e8 f0 f8 f8 f8 08 f8 e8 (sic, asymmetrical)
f8 f8 08 10 10 00 f8 f0 (a3 =-8, h3=-16)
e8 00 08 30 30 08 00 e8
f8 08 10 38 38 10 08 f8
08 20 30 40 40 30 20 08
60 60 60 60 60 60 60 60 # 7th rank, WP
00 00 00 00 00 00 00 00

Code: Select all

d0 00 08 08 08 08 f8 d0 # 10129310, WN back rank
e0 18 20 20 20 20 18 e0
00 20 28 30 30 30 20 00 (note asym c3=40, f3=48)
f0 20 30 40 40 30 20 f0
f8 20 50 50 50 50 20 f8
00 20 40 60 60 40 20 00
f8 18 30 30 30 30 18 f8
e0 08 18 18 18 18 08 e0

Code: Select all

08 00 08 00 00 00 00 10 # 10129390, WB back rank
08 18 08 08 08 08 20 08 (again asym b2=24, g2=32)
08 10 10 10 10 10 10 08
08 08 10 10 10 08 08 08
08 08 08 08 08 08 08 08
08 08 08 10 10 08 08 08
e0 08 08 08 08 08 08 e0
00 00 00 08 08 00 00 00

Code: Select all

00 00 00 08 08 00 00 00 # 10129410, WR back rank
00 00 00 08 08 00 00 00
00 00 00 08 08 00 00 00
00 00 00 08 08 00 00 00
00 00 00 08 08 00 00 00
08 08 08 10 10 08 08 08
10 10 10 18 18 10 10 10
18 18 18 20 20 18 18 18

Code: Select all

00 00 00 10 10 00 00 00 # 10129490 WQ back rank
00 10 18 18 18 18 10 00
00 18 20 20 20 20 18 00
08 10 10 10 10 10 10 08
08 10 18 18 18 18 10 08
00 10 20 28 28 20 10 00
00 10 18 20 20 18 10 00
00 00 10 18 18 10 00 00

Code: Select all

08 09 07 00 00 02 09 07 # 10129510 (f1=2, a1=8)
07 07 02 fb fa 00 08 07
fc fa f8 f0 f0 f8 fa fc
f0 f0 e0 d0 d0 e0 f0 f0
e0 d8 d0 c0 c0 d0 d8 e0
c0 c0 b0 a0 a0 b0 c0 c0
b0 b0 a0 a0 a0 a0 b0 b0
a0 a0 a0 a0 a0 a0 a0 a0

I don't see any obvious Fruit or Rybka/Strelka influence. I have yet to come across any opening/endgame mix either.

[hyatt]

Again, you are not reading what I write. I only mentioned single Fritz 11 beeing ahead of MP engines. Nowhere did I mention DF 11...

Then WHAT does that have to do with the discussion where Mark is looking at DF 11? Your post says Fritz 11 is better than DF 10. What else is new? Shouldn't the newer version be better than the older version? If you have a point here, it is certainly escaping me...

[Damir Desevac]

You definetely should get reading lessons. Fritz 11 is better than DF 10 ? ? Where did you read I wrote something like that in my quote ?
From Mr Watkins post regarding DF11 code: I think these are the PSTs of DF11. I don't know what the scaling is.

Hmmm Bob, it appears your friend Watkins doesn't know what he is looking at. Way to go dude. Do you have any reasonable explanation for that ?
Is Fritz code written in alien language, since it can not be translated to human language ? If he can not understand the Fritz code, how than can he understand Rybka code, who is even more complicated ??

[hyatt]

You definetely should get reading lessons. Fritz 11 is better than DF 10 ? ? Where did you read I wrote something like that in my quote ?
From Mr Watkins post regarding DF11 code: I think these are the PSTs of DF11. I don't know what the scaling is.

Hmmm Bob, it appears your friend Watkins doesn't know what he is looking at. Way to go dude. Do you have any reasonable explanation for that ?
Is Fritz code written in alien language, since it can not be translated to human language ? If he can not understand the Fritz code, how than can he understand Rybka code, who is even more complicated ??

This is simple to me. Perhaps not so simple to you. You sent a link to a final result. Had these three entries, dealing with Fritz:

Fritz 12 2920
Fritz 11 2898
Deep Fritz 10 2CPU 2866

Those are found here:

http://www.husvankempen.de/nunn/40_120_ ... liste.html

Which is a link you posted. We were talking about DF 11 that Mark examined. You chimed in that Fritz 11 was stronger than DF 11. You gave that link to support your point. Quote: "Maybe this would put more light to the matter. Single Version of Fritz 11 ahead of MP engines. I think it is very common in my opinion." Which was posted with that link. So, what does the link, which does not reference Deep Fritz 11 have to do with Mark's looking at Deep Fritz 11? Others have provided links that show that DF 11 is significantly stronger than F11 when used on multiple cores. For some unknown and irrational reason, you want to suggest that F11 and DF11 are far different programs. Where's the evidence? DF 11 came out well after F11. You think Frans REMOVED code from F11 to make DF 11? What is the basis for that.

In short, your entire series of posts make absolutely no sense. It is not I that need to learn how to read, it is YOU that need to learn how to write. Just follow your link above and tell me what it has to do with the discussion about "Deep Fritz 11"

As far as your comment(s) about Mark, totally irrational. He said the PST values came from Deep Fritz 11. I take that statement at face value since he has not made a false post about disassembly as of yet. What, EXACTLY, is your point? Do you actually HAVE a point?

jeez...

[kingliveson]

I think Bob meant DF 11 instead of DF 10.

Edit: Oops, he did mean 10.

Since this all started from DD's similarity tool, here are some comments from him:

You should not use this test to draw conclusions at all. I can make a version that does 10,000 positions if you want and it will be statistically more accurate at the expense of taking a much longer time for each program but you still should not use it to draw conclusions - at least without other significant information to supplement it.

If the tool were designed to actually detect clones you would be right. But it is not suited for that.

[BB+]

it appears your friend Watkins doesn't know what he is looking at. [...] Do you have any reasonable explanation for that ?
Is Fritz code written in alien language, since it can not be translated to human language ? If he can not understand the Fritz code, how than can he understand Rybka code, who is even more complicated ??

I have spent only a few hours looking at Fritz. I spent much more time looking at Rybka. It is clear to me at this point that DF11 is not Fruit-derivative in its eval or PST. I am not too concerned about every last detail (such as whether the PST is scaled when folded in: it seems not to be -- look at the make-move routine of 0x1002c800 and the 0x10129f60 redirection(s) to the above data). I may look a little at the search, but at this point, it is largely a time-wasting exercise to me -- particularly as no matter what I say, or what I evidence I give, the claim will just shift.

I'd also say, from my standpoint the Fritz code is less understandable than Rybka's, due to the use of more obscure ASM/MMX instructions. I also am unable to actually run the engine, unlike with Rybka. At any rate, as can be seen from my comments in the disassembly posted at the head of this thread, it is possible to translate DF11 to "human language".

[Adam Hair]

I think Bob meant DF 11 instead of DF 10.

Edit: Oops, he did mean 10.

Since this all started from DD's similarity tool, here are some comments from him:

You should not use this test to draw conclusions at all. I can make a version that does 10,000 positions if you want and it will be statistically more accurate at the expense of taking a much longer time for each program but you still should not use it to draw conclusions - at least without other significant information to supplement it.

If the tool were designed to actually detect clones you would be right. But it is not suited for that.

This the most sensible position to take. However, it is capable of showing similarities in evaluation. The cases involving high similarity rates have been shown to involve extensive similarities in evaluation. Of course, the tool is more properly suited to be used as a test to indicate engines to check out instead of being used to incriminate.

[Jeremy Bernstein]

This the most sensible position to take. However, it is capable of showing similarities in evaluation. The cases involving high similarity rates have been shown to involve extensive similarities in evaluation. Of course, the tool is more properly suited to be used as a test to indicate engines to check out instead of being used to incriminate.

+1. Kai's contention that his dendogram proves that Fritz is derived from Strelka is over the top. Given the similarities, a motivated investigator should absolutely look into it using conventional tools. But this tool provides nothing but one view into the software it's evaluating, and can be very misleading.