lucasart wrote: With this TwinFish you not only prove my point,
But you never had a point
Engines that will show a 65+% similarity are derived, you don't need the source code, that still stands.
Until the source code is revealed, nothing proves that a closed source engine contains no foreign code.
Sure, never claimed otherwise.
The tool is useless to proof an engine original.
See the difference now?
Sorry, this cuts BOTH ways. It does not PROVE that engines are not original either. It just suggests that further analysis (code inspection) is required. This will NEVER "prove" anything.
Yes, and fingerprint was not considered a proof until 1900 or so.
This is NOT a "fingerprint". Two people walk down the street on different days. They stop at the same places. Coincidence? Do they know each other? Same person wearing two different disguises? Behavior just suggests something connects them. And we are talking behavior of a chess program, not fingerprints or DNA which comes from inside the source code.
What sense are making these analogies? It's about statistics, and no false positives in hundreds of studied engines.
And no false negatives until someone just produced one?
You can't conclude it works just because no exceptions have been found. Finding an exception proves a lot. Not finding one just leaves it "unknown".
This "I have not seen one, so one must not exist" REALLY is not very convincing to me. There are a LOT of things I have never seen, but I am convinced they exist.
Laskos wrote:What sense are making these analogies? It's about statistics, and no false positives in hundreds of studied engines.
Again talking BS as usual.
How do you know there are no false positives???
Do you have source code of all those hundreds of studied engines?
The only thing you do when you see high score after using BS similarity test is scream clone. You never perform any serious analysis, look at sources, disassemble engines in question.
Your claims are a joke.
Several closed source engines that have the highest similarity percentages have been looked at by Mark Watkins, though some in more detail than others. The only one in question may be Fritz 11. Otherwise, there has not been any signs of a false positive. If you have any evidence of a false positive, please share it.
The problem is this:
(1) citing that "hundreds of tests have been done with no false positives found."
(2) citing that "several closed source engines have been looked at."
"hundreds" and "several" leave me in a vague state. For example, you said that several with the highest similarity were tested and found to be not similar by mark.. Doesn't that actually show that there might be a weakness in the test if it says "they are similar" but inspection says "they are not?"
This is unsound reasoning. If you are suspicious of something, just saying "I have not seen an exception" doesn't really support the argument very well. If someone shows an exception, it will prove the process is flawed. Until one is shown, it is only a guesstimate of whether the process is flawed or not. How long did the search for a Higgs boson go on? Since no one had found one, did that mean such did not exist?
If you want to ask me "do you believe that the test is pretty accurate, statistically?" I would answer yes. "pretty accurate" however. NOT "perfect". Do I consider it proof that two programs are clones? No. I consider it a suggestion, one that requires code inspection to actually prove the clone status. Do I consider it proof that two programs are not related? No, I consider it pretty reasonable evidence they are not, but not proof. That STILL requires code comparison.
bob wrote:If you want to ask me "do you believe that the test is pretty accurate, statistically?" I would answer yes. "pretty accurate" however. NOT "perfect". Do I consider it proof that two programs are clones? No. I consider it a suggestion, one that requires code inspection to actually prove the clone status. Do I consider it proof that two programs are not related? No, I consider it pretty reasonable evidence they are not, but not proof. That STILL requires code comparison.
CW wrote:My position is that an accused person is innocent until proven guilty.
What's yours?
Mine is the same, but the evidence has become substantial. We have the gun that killed someone. We have fingerprints on the gun. We have gunshot residue on the suspect. We have established motive. We have established opportunity. The suspect was seen entering and leaving the building during the time the victim was killed. Gunshots were heard from inside the building while the suspect was there. The suspect had victim's blood on his clothes. All we lack is an eye-witness. But the case _still_ looks pretty bad and people have been convicted on far less.
lucasart wrote: With this TwinFish you not only prove my point,
But you never had a point
Engines that will show a 65+% similarity are derived, you don't need the source code, that still stands.
Until the source code is revealed, nothing proves that a closed source engine contains no foreign code.
Sure, never claimed otherwise.
The tool is useless to proof an engine original.
See the difference now?
Sorry, this cuts BOTH ways. It does not PROVE that engines are not original either. It just suggests that further analysis (code inspection) is required. This will NEVER "prove" anything.
Yes, and fingerprint was not considered a proof until 1900 or so.
This is NOT a "fingerprint". Two people walk down the street on different days. They stop at the same places. Coincidence? Do they know each other? Same person wearing two different disguises? Behavior just suggests something connects them. And we are talking behavior of a chess program, not fingerprints or DNA which comes from inside the source code.
What sense are making these analogies? It's about statistics, and no false positives in hundreds of studied engines.
And no false negatives until someone just produced one?
You can't conclude it works just because no exceptions have been found. Finding an exception proves a lot. Not finding one just leaves it "unknown".
This "I have not seen one, so one must not exist" REALLY is not very convincing to me. There are a LOT of things I have never seen, but I am convinced they exist.
1. Hundreds of studied engines mean tens of thousands of pairs to compare, and not a single false positive appeared. So, it's entirely plausible pink elephants exist, but highly unlikely.
2. It makes sense for someone to try to avoid detection and produce a false negative. The opposite, to try intentionally to make your engine more similar on Sim to another engine, would be silly. There is no incentive in doing that.
Laskos wrote:What sense are making these analogies? It's about statistics, and no false positives in hundreds of studied engines.
Again talking BS as usual.
How do you know there are no false positives???
Do you have source code of all those hundreds of studied engines?
The only thing you do when you see high score after using BS similarity test is scream clone. You never perform any serious analysis, look at sources, disassemble engines in question.
Your claims are a joke.
Milos, come to think of what similarity tester measures. It's a collection of 8238 careful chosen chess positions where multiple good moves are present. Volume (8238) plus the chance that 2 supposed original engines pick the same move with an interval of 2 out of 3 (66%) statistically is out of the question.
For more information I refer to my experiments with similarity tester.
lucasart wrote: With this TwinFish you not only prove my point,
But you never had a point
Engines that will show a 65+% similarity are derived, you don't need the source code, that still stands.
Until the source code is revealed, nothing proves that a closed source engine contains no foreign code.
Sure, never claimed otherwise.
The tool is useless to proof an engine original.
See the difference now?
Sorry, this cuts BOTH ways. It does not PROVE that engines are not original either. It just suggests that further analysis (code inspection) is required. This will NEVER "prove" anything.
Yes, and fingerprint was not considered a proof until 1900 or so.
This is NOT a "fingerprint". Two people walk down the street on different days. They stop at the same places. Coincidence? Do they know each other? Same person wearing two different disguises? Behavior just suggests something connects them. And we are talking behavior of a chess program, not fingerprints or DNA which comes from inside the source code.
What sense are making these analogies? It's about statistics, and no false positives in hundreds of studied engines.
And no false negatives until someone just produced one?
You can't conclude it works just because no exceptions have been found. Finding an exception proves a lot. Not finding one just leaves it "unknown".
This "I have not seen one, so one must not exist" REALLY is not very convincing to me. There are a LOT of things I have never seen, but I am convinced they exist.
1. Hundreds of studied engines mean tens of thousands of pairs to compare, and not a single false positive appeared. So, it's entirely plausible pink elephants exist, but highly unlikely.
2. It makes sense for someone to try to avoid detection and produce a false negative. The opposite, to try intentionally to make your engine more similar on Sim to another engine, would be silly. There is no incentive in doing that.
The incentive is to have a stronger engine and based on my memory the programmer of Naum already did it with Rybka based on his words(I think that he did it with Rybka2.3.2a but I am not sure about the exact version of Rybka).
I think that we can at least agree that big similarity is not something that can happen by accident and the engine is derived from the code or from the output of another engine.
Uri Blass wrote:
The incentive is to have a stronger engine and based on my memory the programmer of Naum already did it with Rybka based on his words(I think that he did it with Rybka2.3.2a but I am not sure about the exact version of Rybka).
I think that we can at least agree that big similarity is not something that can happen by accident and the engine is derived from the code or from the output of another engine.
I don't believe in this Naum story. Try to optimize engine's strength via a test suite instead of pure Elo-wise tests. You will get a weaker engine. Trying to optimize to play these neutral positions from Sim similarly to a stronger engine, if your own engine has a very different eval, will only wreck the engine. There are hundreds of parameters to tune, it would be a miracle for a completely different eval to be tunable according to the same parameters, and to get a stronger engine.
bob wrote:If you want to ask me "do you believe that the test is pretty accurate, statistically?" I would answer yes. "pretty accurate" however. NOT "perfect". Do I consider it proof that two programs are clones? No. I consider it a suggestion, one that requires code inspection to actually prove the clone status. Do I consider it proof that two programs are not related? No, I consider it pretty reasonable evidence they are not, but not proof. That STILL requires code comparison.
CW wrote:My position is that an accused person is innocent until proven guilty.
What's yours?
Mine is the same, but the evidence has become substantial. We have the gun that killed someone. We have fingerprints on the gun. We have gunshot residue on the suspect. We have established motive. We have established opportunity. The suspect was seen entering and leaving the building during the time the victim was killed. Gunshots were heard from inside the building while the suspect was there. The suspect had victim's blood on his clothes. All we lack is an eye-witness. But the case _still_ looks pretty bad and people have been convicted on far less.
If you are going to be that obtuse, you are going to have to explain your point. In the post you liked to, I said "innocent until proven guilty". Not ONE comment in this thread has said anything contrary to that statement.
I said (a) the similarity tester can provide a suggestion that one program is a derivative of another, NOT "proof". (b) the similarity tester can provide a suggestion that two programs are completely different and one is not a derivative of the other, but not "proof" that they are not.
So please explain what your point is supposed to be, I have always been "innocent until proven guilty" in this clone nonsense.
lucasart wrote: With this TwinFish you not only prove my point,
But you never had a point
Engines that will show a 65+% similarity are derived, you don't need the source code, that still stands.
Until the source code is revealed, nothing proves that a closed source engine contains no foreign code.
Sure, never claimed otherwise.
The tool is useless to proof an engine original.
See the difference now?
Sorry, this cuts BOTH ways. It does not PROVE that engines are not original either. It just suggests that further analysis (code inspection) is required. This will NEVER "prove" anything.
Yes, and fingerprint was not considered a proof until 1900 or so.
This is NOT a "fingerprint". Two people walk down the street on different days. They stop at the same places. Coincidence? Do they know each other? Same person wearing two different disguises? Behavior just suggests something connects them. And we are talking behavior of a chess program, not fingerprints or DNA which comes from inside the source code.
What sense are making these analogies? It's about statistics, and no false positives in hundreds of studied engines.
And no false negatives until someone just produced one?
You can't conclude it works just because no exceptions have been found. Finding an exception proves a lot. Not finding one just leaves it "unknown".
This "I have not seen one, so one must not exist" REALLY is not very convincing to me. There are a LOT of things I have never seen, but I am convinced they exist.
1. Hundreds of studied engines mean tens of thousands of pairs to compare, and not a single false positive appeared. So, it's entirely plausible pink elephants exist, but highly unlikely.
2. It makes sense for someone to try to avoid detection and produce a false negative. The opposite, to try intentionally to make your engine more similar on Sim to another engine, would be silly. There is no incentive in doing that.
Nobody mentioned "incentive". Just "the possibility it could happen." And we have nothing to prove it can not. It just hasn't happened yet.
Laskos wrote:What sense are making these analogies? It's about statistics, and no false positives in hundreds of studied engines.
Again talking BS as usual.
How do you know there are no false positives???
Do you have source code of all those hundreds of studied engines?
The only thing you do when you see high score after using BS similarity test is scream clone. You never perform any serious analysis, look at sources, disassemble engines in question.
Your claims are a joke.
Milos, come to think of what similarity tester measures. It's a collection of 8238 careful chosen chess positions where multiple good moves are present. Volume (8238) plus the chance that 2 supposed original engines pick the same move with an interval of 2 out of 3 (66%) statistically is out of the question.
For more information I refer to my experiments with similarity tester.
I wonder what would happen if you took a human GM, and a younger GM that trained under/with him for years, and tested both? You don't think they would play similarly?
There is no such term as "statistically out of the question". It might have a low probability of happening, but low != 0.0... So there is absolutely room for a false positive just as there is room for a false negative as already shown.
