Vas about IPP engine

[Anton]

I have been testing Ипполит 0.084 at long time-controls, it scored 3.0-4.0 (2:2:0) in 4 hour (2 hours per engine) games against Stockfish, and in a 48 hour game as black against Crafty, it secured an easy win (no books used) in an open game that began with the Italian four knights opening

Ипполит has a problem with time management, at one point it had a fail low at 24 ply deep, and took 3+ hours in deep thought before resurfacing with a move, by move 24, Ипполит had 7 hours remaining compared to 14 for Crafty. this is how the game stood after 32 moves. both engines considered themselves a pawn ahead at this point. Crafty resigned 28 moves later.
[d]4b3/ppp3k1/3p4/1P1P1p1P/2BP1p2/5PqP/P7/5Q1K w - - 1 33[/d]

I am also mystified by the claims that it is buggy and crashes often, I have yet to find a bug, have a crash, or a loss on time.

[Michael Sherwin]

can i ask a question, & sorry to ask for the clarification to be repeated, but why is this just blowing up now?

i found the ippolit site months ago; why did no one capable have the curiosity to compile this?? iirc the files were available, i just assumed id missed the explosion over it (had no idea itd be this big) and it had already gone the way of strelka.

wtf took so long??

Because, it was squashed like a bug before many programmers found out about it. I never heard of Ippolit before two days ago!

[Christopher Conkie]

can i ask a question, & sorry to ask for the clarification to be repeated, but why is this just blowing up now?

i found the ippolit site months ago; why did no one capable have the curiosity to compile this?? iirc the files were available, i just assumed id missed the explosion over it (had no idea itd be this big) and it had already gone the way of strelka.

wtf took so long??

Because, it was squashed like a bug before many programmers found out about it. I never heard of Ippolit before two days ago!

We found the page with the source for the first time on 6th May 2009.

Then we compiled it and showed it to various programmers asking them what they thought of it.

Christopher

[Alexander Schmidt]

• No tables in Ippolit
  Different Movegen
  Completely different UCI handling
  Different evaluation
  Different search
  Different Maximum Depth

One addition:

• Different time management

Show me something similar except the strength

I am still waiting...

[Anton]

Just correcting my earlier post, Crafty evaluated the above position as -0.84/21, not positively as I suggested

[M ANSARI]

I have been testing Ипполит 0.084 at long time-controls, it scored 3.0-4.0 (2:2:0) in 4 hour (2 hours per engine) games against Stockfish, and in a 48 hour game as black against Crafty, it secured an easy win (no books used) in an open game that began with the Italian four knights opening

Ипполит has a problem with time management, at one point it had a fail low at 24 ply deep, and took 3+ hours in deep thought before resurfacing with a move, by move 24, Ипполит had 7 hours remaining compared to 14 for Crafty. this is how the game stood after 32 moves. both engines considered themselves a pawn ahead at this point. Crafty resigned 28 moves later.
[d]4b3/ppp3k1/3p4/1P1P1p1P/2BP1p2/5PqP/P7/5Q1K w - - 1 33[/d]

I am also mystified by the claims that it is buggy and crashes often, I have yet to find a bug, have a crash, or a loss on time.

Testing at long time controls against a non Rybka engine will only show how strong Rybka 3 or a clone made of stolen code of Rybka 3 is. In my earlier testing of Rybka 3, a zero contempt R3 on a single core will squeak by a full bored 8 core Naum 4. That is saying quite a bit since Naum 4 at the time was the strongest non Rybka 3 engine. So even a single core R3 against a non Rybka 3 engine will show an advantage. Why not try playing the R3 cloned engine against another R3 engine with zero contempt at the same time limits or even 5 0 or 10 0. I think you will find results favor original R3. Something very similar to this happened with Rybka Winfinder, when many people used fast time control results to think that it was a stronger engine than the normal Rybka. In some cases Rybka Winfinder would find critical lines 10X or even 15X faster than the default Rybka, but in a long time control match it would be many ELO points weaker. I think here is a similar case, some essential parts of R3 are there, but evaluation tables and other stuff is missing. The result is an extremely efficient and fast searcher ... but slightly lobotomized.

[IWB]

Hello

...

We found the page with the source for the first time on 6th May 2009.

Then we compiled it and showed it to various programmers asking them what they thought of it.

Christopher

and ... what did the various programmers thought of it ...?

Bye
Ingo

[Uri Blass]

I have been testing Ипполит 0.084 at long time-controls, it scored 3.0-4.0 (2:2:0) in 4 hour (2 hours per engine) games against Stockfish, and in a 48 hour game as black against Crafty, it secured an easy win (no books used) in an open game that began with the Italian four knights opening

Ипполит has a problem with time management, at one point it had a fail low at 24 ply deep, and took 3+ hours in deep thought before resurfacing with a move, by move 24, Ипполит had 7 hours remaining compared to 14 for Crafty. this is how the game stood after 32 moves. both engines considered themselves a pawn ahead at this point. Crafty resigned 28 moves later.
[d]4b3/ppp3k1/3p4/1P1P1p1P/2BP1p2/5PqP/P7/5Q1K w - - 1 33[/d]

I am also mystified by the claims that it is buggy and crashes often, I have yet to find a bug, have a crash, or a loss on time.

Testing at long time controls against a non Rybka engine will only show how strong Rybka 3 or a clone made of stolen code of Rybka 3 is. In my earlier testing of Rybka 3, a zero contempt R3 on a single core will squeak by a full bored 8 core Naum 4. That is saying quite a bit since Naum 4 at the time was the strongest non Rybka 3 engine. So even a single core R3 against a non Rybka 3 engine will show an advantage. Why not try playing the R3 cloned engine against another R3 engine with zero contempt at the same time limits or even 5 0 or 10 0. I think you will find results favor original R3. Something very similar to this happened with Rybka Winfinder, when many people used fast time control results to think that it was a stronger engine than the normal Rybka. In some cases Rybka Winfinder would find critical lines 10X or even 15X faster than the default Rybka, but in a long time control match it would be many ELO points weaker. I think here is a similar case, some essential parts of R3 are there, but evaluation tables and other stuff is missing. The result is an extremely efficient and fast searcher ... but slightly lobotomized.

I never read that rybka win finder was stronger than rybka at fast time control.
I do not believe that positional evaluation is unimportant at fast time control.

It may be the opposite.
In fast time control you need a good evaluation in order to win KRK positions with no tablebases

In slow time control you do not need it because search get the exact result so programs may have the same 5.12 pawns advantage evaluation for every big score that is not mate with no problem at slow time control but with clear problems to win won positions at fast time control and not winning won positions may do the program weaker at fast time control.

Uri

[Uri Blass]

I can add that I read in the rybka forum in the flip side that the new engine has knoweldge about the blind bishop unlike rybka.

I do not say that it is not a clone but
I did not read an evidence that it is a clone(except the fact that I read that Vas and other people that I trust claim that it is a clone that is of course no proof).

Note that Vas did not mention the engine name so we cannot be 100% sure that he is talking about the same engine.

Even mentioning the name is not a proof without the link because it is possible that 2 different links have different exe file with the same name
when only one of them is a clone.

Uri

[Graham Banks]

Note that Vas did not mention the engine name so we cannot be 100% sure that he is talking about the same engine.

His statement was about Ippolit. I know that for certain.