New release Revenge 3.0

[Fabio Gobbato]

I am pleased to announce a new version of my chess engine: Revenge 3.0

There are a lot of improvements compared to Revenge 2.0:
- New neural network
- Some speed optimizations in neural network calculation
- Improved extensions
- Speed optimizations in search
- Improved reverse futility pruning
- Enhanced hash replacement scheme
- Improved internal iterative deepening
- Better quiets move ordering
- Improved parallel search
- Small improvement in chess960
- Stronger MultiPV analysis
- Added in the package executables for AVX512 and AVX512VNNI

All these changes results in about +80 elo compared to Revenge 2.0 measured in selfplay at 1:00+0.5 time control.
This new version compared to Revenge 2.0 is stronger in late middlegame.

The engine is available for Windows x86_64, Linux x86_64, Linux ARMv8 and Linux ARMv7.

Although the electricity costs have increased, the engine is sold at the same price as the previous version: 19.99€

For buying the engine or getting more informations visit www.revengechess.com

[Vinvin]

Great improvement for such a strong engine !
Congratulations !!

[Krzysztof Grzelak]

I am pleased to announce a new version of my chess engine: Revenge 3.0

There are a lot of improvements compared to Revenge 2.0:
- New neural network
- Some speed optimizations in neural network calculation
- Improved extensions
- Speed optimizations in search
- Improved reverse futility pruning
- Enhanced hash replacement scheme
- Improved internal iterative deepening
- Better quiets move ordering
- Improved parallel search
- Small improvement in chess960
- Stronger MultiPV analysis
- Added in the package executables for AVX512 and AVX512VNNI

All these changes results in about +80 elo compared to Revenge 2.0 measured in selfplay at 1:00+0.5 time control.
This new version compared to Revenge 2.0 is stronger in late middlegame.

The engine is available for Windows x86_64, Linux x86_64, Linux ARMv8 and Linux ARMv7.

Although the electricity costs have increased, the engine is sold at the same price as the previous version: 19.99€

For buying the engine or getting more informations visit www.revengechess.com

And how the program tests for a long time look like. For example - 5'+3'' or 15'+5".

[Fabio Gobbato]

I am pleased to announce a new version of my chess engine: Revenge 3.0

There are a lot of improvements compared to Revenge 2.0:
- New neural network
- Some speed optimizations in neural network calculation
- Improved extensions
- Speed optimizations in search
- Improved reverse futility pruning
- Enhanced hash replacement scheme
- Improved internal iterative deepening
- Better quiets move ordering
- Improved parallel search
- Small improvement in chess960
- Stronger MultiPV analysis
- Added in the package executables for AVX512 and AVX512VNNI

All these changes results in about +80 elo compared to Revenge 2.0 measured in selfplay at 1:00+0.5 time control.
This new version compared to Revenge 2.0 is stronger in late middlegame.

The engine is available for Windows x86_64, Linux x86_64, Linux ARMv8 and Linux ARMv7.

Although the electricity costs have increased, the engine is sold at the same price as the previous version: 19.99€

For buying the engine or getting more informations visit www.revengechess.com

And how the program tests for a long time look like. For example - 5'+3'' or 15'+5".

We haven't tested at longer time control because it needs a lot of time but of course could be something less.

[Wolfgang]

And how the program tests for a long time look like. For example - 5'+3'' or 15'+5".

You can simply find out: buy and test it
No developer, really no one , tests with such long TC

[Krzysztof Grzelak]

You can simply find out: buy and test it
No developer, really no one , tests with such long TC

Unfortunately, this is a serious mistake a lot of programmers. So what if the programmer will give me millions of games in a short time. And when it comes to playing for longer periods, the program loses. That's not what it's about. This is not the way.

[Wolfgang]

Do you really believe what you say/write??
LOL (for you: laughing out loud)

What you write is simply wrong, not to say nonsense. With longer TC the difference between the programs may become smaller, but the sequence remains the same. Maybe not every single place in a rating list, but no program is e.g. No.3 with short TC and suddenly No. 10 or so with longer TC. That's simply not true

[Graham Banks]

With longer TC the difference between the programs may become smaller, but the sequence remains the same. Maybe not every single place in a rating list, but no program is e.g. No.3 with short TC and suddenly No. 10 or so with longer TC. That's simply not true

Agreed.

[Krzysztof Grzelak]

Do you really believe what you say/write??
LOL (for you: laughing out loud)

What you write is simply wrong, not to say nonsense. With longer TC the difference between the programs may become smaller, but the sequence remains the same. Maybe not every single place in a rating list, but no program is e.g. No.3 with short TC and suddenly No. 10 or so with longer TC. That's simply not true

Disagreement. And he doesn't write nonsense Wolfgang. You think silly. I prefer to have games for longer times than for shorter times. Unfortunately, you also have no tests for a long time for games.

[Fabio Gobbato]

Do you really believe what you say/write??
LOL (for you: laughing out loud)

What you write is simply wrong, not to say nonsense. With longer TC the difference between the programs may become smaller, but the sequence remains the same. Maybe not every single place in a rating list, but no program is e.g. No.3 with short TC and suddenly No. 10 or so with longer TC. That's simply not true

Disagreement. And he doesn't write nonsense Wolfgang. You think silly. I prefer to have games for longer times than for shorter times. Unfortunately, you also have no tests for a long time for games.

Of course testing a longer time control is better than testing at short time control. The problem is that with the hardware we have today playing 10000 games or more to have small error margin at 5:00+3 or 15:00+5 takes a lot of time also if you have an high end cpu. Developers have to find a good compromise between testing more things and use a suitable tc.