I read the following conversation in the nTCEC chat of nTCEC:
nTCEC chat wrote:Guest: Raimund, how do you do tuning of evaluation?
Raimund Heid: I change to value to be tuned until I have found three results with a peak value in the middle. Than find the maximum of a quadratic function with mathematica (e.g.: interpolating polynomial {{126, 50.50}, {132, 50.25}, {120, 47.50}})
Up to now, there are not new binaries of Protector 1.5, but a public release should be done in few weeks (in words of Raimund). Let us wait and see if there is a significant Elo gain from version 1.4.0... I wish him good luck.
Ajedrecista wrote:Up to now, there are not new binaries of Protector 1.5, but a public release should be done in few weeks (in words of Raimund). Let us wait and see if there is a significant Elo gain from version 1.4.0... I wish him good luck.
It's been a while. Looking forward to the new release.
Ajedrecista wrote:Up to now, there are not new binaries of Protector 1.5, but a public release should be done in few weeks (in words of Raimund). Let us wait and see if there is a significant Elo gain from version 1.4.0... I wish him good luck.
It's been a while. Looking forward to the new release.
Raimund Heid wrote:v1.5 should be at least 80 points better than 1.4.
The chat of nTCEC could be a goldmine of computer chess news!
In the case that it is confirmed, an improvement of at least 80 Elo is more than an usual, decent improvement... even in two years, given the high rating of current Protector. It should be in the top-10 of IPON list! I would like to know a better estimate from the author, such as 90 ± 10 Elo... but I am unable to post as guest in the chat of nTCEC. We can only wait.
Protector 1. 5. 0 Author Raymund Heid.
Compiled by Jose M velasco.
Not code change.
Beta distribution without warranties,
allow the distribution of the author
Sources 29-01-2013 http://sourceforge.net/projects/protector/
Ajedrecista wrote:Up to now, there are not new binaries of Protector 1.5, but a public release should be done in few weeks (in words of Raimund). Let us wait and see if there is a significant Elo gain from version 1.4.0... I wish him good luck.
It's been a while. Looking forward to the new release.
Raimund Heid wrote:v1.5 should be at least 80 points better than 1.4.
The chat of nTCEC could be a goldmine of computer chess news!
In the case that it is confirmed, an improvement of at least 80 Elo is more than an usual, decent improvement... even in two years, given the high rating of current Protector. It should be in the top-10 of IPON list! I would like to know a better estimate from the author, such as 90 ± 10 Elo... but I am unable to post as guest in the chat of nTCEC. We can only wait.
Regards from Spain.
Ajedrecista.
The best I can add to this conversation is that I ran Deep Sjeng c't against Protector 1.4.0 at 4+2 in 50 game match. Sjeng won by 28 elo.
Then I ran the same Deep Sjeng c't against Protector 1.5.0 at same control in 50 games. This time Protector 1.5.0 won by exactly 28 elo!
So 28x2 = 56 elo improvement. Before you remind me- I am aware of the error margins. But at least it is a good start.
Can't get the linux version to work correctly. With Scid it always complains that it's not a UCI-engine and then won't run. In terminal window after uci go the calculation stops a second later.
