I am sure after 250 games don't means that what I wrote is right. Often I am wrong with my first impressions. See my message with a bit enthusiasm I had with me. After the round robin from both versions we know it with more precision in detail. More important is that PeterPan have additional information about his changes. So he can try to make his own compile better if needful.
End of this test, I am sure, the different from both versions will be very small. So it's for the users only a decision, which style is more interesting for own things. From technical point of view both versions works fine and after all what I saw it's quit clear that both versions plays stronger as the latest version B47cB I tested.
After round 13 / 40.
I added 35 opponents in both round robins. ELO is very clear after 200-400 games only. So I think I will stopped here and added perhaps each two days the next results.
More as +-10 isn't to await.
B46fB is playing more solid, 999946f more risky. Engines which played more risky made in a rating list more points becuase they have to play vs. more weaker opponents (the TOP-7). Its clear that PHQ or Exp. 42 are playing a bit stronger. I think we can explain the IvanHoe results very easy.
In the near of Houdini = more risky
spekulativ is the secret I think!
after all my information I think only PeterPan can give you an answer in detail. I don't know his special settings he added in the compiles. KLO compiles ... no changes in the IvanHoe main versions.
In the past different persons compiles the main IvanHoe versions. At the moment most known are the compiles from KLO and PeterPan. Who is KLO and who is PeterPan, this information can be read found in my news ticker.
After my opinion we have the same situation with these compiles around IvanHoe as with different good settings from other engines.
Rybka 4.0 EXP. 42 is stronger as Rybka 4.1
Stockfish 2.1.1 PHQ is stronger as Stockfish 2.1.1
It seems that IvanHoe 999946f is stronger as B46fB.
But only if you produced the SWCR conditions (more weaker as stronger engines in round robins). I can used only what I have, and all TOP engines are included.
This are very good examples I think !!!
Rybka 4.0 EXP. 42, Stockfish 2.1.1 PHQ and IvanHoe 999946f playing more speculative, more aggressive. In my rating list (more weaker engines as stronger engines in the round robins of TOP-35) with the results that the performance by more speculative / aggressive engines are higher. This engines produced a clearly lower remis quote. For Bayesian is this important because engines with a lower remis quote will get a higher rating.
Most of the chess player like the solid style. Rybka 4.1, Stockfish 2.1.1 and IvanHoe B46fB are playing more solid.
You are the programmer. You can give me more information about it. Speculative in NullMove or aggressive settings seems to be the secret for a higher rating??!
Stockfish 2.1.1 PHQ problem:
SF 2.11 PHQ give to many own material for an attack in the middlegame. Vs. weaker engines with more success and vs. stronger engines with two possibilities.
1. A fantastic game
2. Remis very fast or Stockfish will lost if the opponent is playing very precise.
EXP. 42 is good to compare with PHQ
And IvanHoe?
Not easy at all ... have a look on the results vs. Houdini 2.0c x64. First engine which have a chance to beat the number 1 is the IvanHoe KLO. But vs. the other strong engines the results are not better as from the PeterPan version.
We know the EXP 42 settings of Rybka. We know the PHQ settings of Stockfish and we can donwload the main sources from IvanHoe. This information can help to find out more about programs on this high level.
The topic isn't esay for me, perhaps it's wrong what I have here in my brain but most of my opinions I generate with in combination with statistics.
Move average from PeterPan compiles is 4 moves higher.
Remis quote is 7% higher.
More short lost games by the PeterPan version.
But all in all vs. the most stronger opponents PeterPan's compile is playing more solid. In blitz I have the information by Stefan Pohl that the version by PeterPan is around 5 ELO stronger ... after many many games.