TalkChess.com

More info on:
https://sites.google.com/site/nebulachess/

I spent the last 6 months working full time on Nebula. Culmination of all that hard work is this version. It has almost everything I planned to implement when I started a year ago (first 6 months were on and off).
There is still a lot that can be added. Especially in the pruning area, but I want to take a little break - meaning working 4 hours a day instead of 8
Hope you enjoy testing this version.

Cheers, Dragan

Dragan wrote:More info on:
https://sites.google.com/site/nebulachess/

I spent the last 6 months working full time on Nebula. Culmination of all that hard work is this version. It has almost everything I planned to implement when I started a year ago (first 6 months were on and off).
There is still a lot that can be added. Especially in the pruning area, but I want to take a little break - meaning working 4 hours a day instead of 8
Hope you enjoy testing this version.

Cheers, Dragan

Dragan,

Congratulations on getting Nebula out. I see from your webpage that you added SE to it.
How much of a strength increase did that give you? It's on my TODO list.

regards,
--tom

Tom Likens wrote:

Dragan wrote:More info on:
https://sites.google.com/site/nebulachess/

I spent the last 6 months working full time on Nebula. Culmination of all that hard work is this version. It has almost everything I planned to implement when I started a year ago (first 6 months were on and off).
There is still a lot that can be added. Especially in the pruning area, but I want to take a little break - meaning working 4 hours a day instead of 8
Hope you enjoy testing this version.

Cheers, Dragan

Dragan,

Congratulations on getting Nebula out. I see from your webpage that you added SE to it.
How much of a strength increase did that give you? It's on my TODO list.

regards,
--tom

From some testing results I've seen, looks like a decent improvement in strength, but that is obviously a combination of all the things that Dragan has implemented.

Graham Banks wrote:

Tom Likens wrote:

Dragan wrote:More info on:
https://sites.google.com/site/nebulachess/

I spent the last 6 months working full time on Nebula. Culmination of all that hard work is this version. It has almost everything I planned to implement when I started a year ago (first 6 months were on and off).
There is still a lot that can be added. Especially in the pruning area, but I want to take a little break - meaning working 4 hours a day instead of 8
Hope you enjoy testing this version.

Cheers, Dragan

Dragan,

Congratulations on getting Nebula out. I see from your webpage that you added SE to it.
How much of a strength increase did that give you? It's on my TODO list.

regards,
--tom

From some testing results I've seen, looks like a decent improvement in strength, but that is obviously a combination of all the things that Dragan has implemented.

Probably true, but usually we try to be somewhat scientific in these things, i.e.

• 1. add one change,
  2. test
  3. verify
  4. record the results
  5. rinse and repeat

My guess is Dragan likely has the elo increase for each change (or a good estimate).

regards,
--tom

Yes I do

Version before the SE scored 721 wins and 1130 losses in 2500 total games against 10 different opponents.
SE version got 821 wins and 1013 losses.

So the improvement was +30 ELO

The thing I like the most about the SE is that improves engine's tactical abilities.
You should definitely have it as a high priority improvement.

Dragan wrote:Yes I do

Version before the SE scored 721 wins and 1130 losses in 2500 total games against 10 different opponents.
SE version got 821 wins and 1013 losses.

So the improvement was +30 ELO

The thing I like the most about the SE is that improves engine's tactical abilities.
You should definitely have it as a high priority improvement.

Thanks Dragan,

I figured you had the numbers. BTW, I downloaded Nebula 2.0 and it's beating the latest version of Djinn by about +40 elo after 1000 3sec.+1sec games, which is a very nice improvement by Nebula since my latest version of Djinn is much stronger than 0.979.

Well done

SzG wrote:

Tom Likens wrote:my latest version of Djinn is much stronger than 0.979.

Aiming the stratosphere, Tom? 0.979 was already about 60 Elo better than 0.969. Please slow down, I cannot keep pace.

Hey Gabor,

That's one of the advantages of having a weak engine, there's basically only one direction to go! No worries though, Dragan has now set the bar much higher so I think you and the other testers are safe for a while!

BTW, I still need to release version 1.0 (it's only been about 12 years since I started working on Djinn, so it may be time)!

best,
--tom

Tom Likens wrote:

SzG wrote:

Tom Likens wrote:my latest version of Djinn is much stronger than 0.979.

Aiming the stratosphere, Tom? 0.979 was already about 60 Elo better than 0.969. Please slow down, I cannot keep pace.

Hey Gabor,

That's one of the advantages of having a weak engine, there's basically only one direction to go! No worries though, Dragan has now set the bar much higher so I think you and the other testers are safe for a while!

BTW, I still need to release version 1.0 (it's only been about 12 years since I started working on Djinn, so it may be time)!

best,
--tom

Good luck Tom

Djinn is one of the mile stones in computer chess history......
Dr.D

Thanks for the effort!

Thanks Tom

I tried using Djinn as a test opponent but the current public version doesn't do very well in bullet (40/5" or 40/10")
This is not very surprising considering the fact xboard doesn't send time in milliseconds. Arasan fares even worse so I wasn't able to use it either.
Dirty is the only xboard engine handling bullet well.

I hope you make a great improvement but not too great