Development of Onno ends
Posted: Sun Mar 13, 2011 8:26 pm
Development of Onno stops. The will not be any new version. Also
Onno is no longer available for sale.
When starting Onno, I was convinced that by using mostly known
algorithms, but coding from scratch more efficiently, using software
engineering techniques such as unit tests, assertions, and different
kinds of logfiles, and adding new algorithmic ideas here and there, I
could make something much stronger then Toga.
As Toga was among the top engines in those days, I believed that this
might mean the strongest engine in the world. In deed, Onno is now
significantly stronger than Toga 2.0 or any other Engine that existed
when I started (and about the same strength as Rybka 1.0 Beta which
came out about the same time when Onno started). This is in spite of
the fact that almost no ideas from other people's newer engines are
built in in Onno.
While my initial assumption on the achievable playing strength has
proven true, I was wrong with the achievable ranking: I considerably
underestimated the time required to write a good chess engine and I
underestimated the speed of the progress in the market (which BTW has
increased over the last years). When Onno 1.0 was released, it was
ranked #7 in the world, just a few Elo after #4 (CEGT
rating). Multi-core took much longer than expected. When it was
finished, Onno had fallen back to around #20.
I am confident that I would eventually reach the level of today's top
programs. But by the time I had reached that level, they would be
even stronger. I learnt that writing a chess engine is primarily a
game of time. To keep pace, I would definitely have to invest even
more time per week than I did. Other obligations, most notably my
main job, detain me from doing that.
I still believe in the software engineering approach. It works, brings
many additional Elos. Some more remarks on this will go to the
technical discussions forum. But this approach is too slow in today's
rapid development, at least for a single part time developer.
Before I leave you, I will post several ideas that made Onno stronger
to the technical discussions forum.
I have not listened to the discussions in the fora over the last
years. So it is possible that somebody else has invented the same
independently. Also of course these ideas might have been present
before Onno and I might just have overlooked them. I did not do
intensive research now if my ideas are known. I will just post
them. If they are known, somebody else will be able to point out this
in a few minutes, with much less effort than a research in the forum
history would take me. The only thing I did is a quick check if my
ideas are already contained in Stockfish. Some are to some extend; I
will comment on this in the technical discussions.
Of course I hope that you will find my ideas useful, but if they are
not, I would not be too surprised or disappointed. If they are not yet
known, they should be tried in Stockfish.
If my ideas are any good, Vas will read and implement them too. But a
community has more manpower than Vas and his few employees. So if the
community is well organized, they might win the game of time. Good
luck fighting Rybka and Houdini this way!
Onno is no longer available for sale.
When starting Onno, I was convinced that by using mostly known
algorithms, but coding from scratch more efficiently, using software
engineering techniques such as unit tests, assertions, and different
kinds of logfiles, and adding new algorithmic ideas here and there, I
could make something much stronger then Toga.
As Toga was among the top engines in those days, I believed that this
might mean the strongest engine in the world. In deed, Onno is now
significantly stronger than Toga 2.0 or any other Engine that existed
when I started (and about the same strength as Rybka 1.0 Beta which
came out about the same time when Onno started). This is in spite of
the fact that almost no ideas from other people's newer engines are
built in in Onno.
While my initial assumption on the achievable playing strength has
proven true, I was wrong with the achievable ranking: I considerably
underestimated the time required to write a good chess engine and I
underestimated the speed of the progress in the market (which BTW has
increased over the last years). When Onno 1.0 was released, it was
ranked #7 in the world, just a few Elo after #4 (CEGT
rating). Multi-core took much longer than expected. When it was
finished, Onno had fallen back to around #20.
I am confident that I would eventually reach the level of today's top
programs. But by the time I had reached that level, they would be
even stronger. I learnt that writing a chess engine is primarily a
game of time. To keep pace, I would definitely have to invest even
more time per week than I did. Other obligations, most notably my
main job, detain me from doing that.
I still believe in the software engineering approach. It works, brings
many additional Elos. Some more remarks on this will go to the
technical discussions forum. But this approach is too slow in today's
rapid development, at least for a single part time developer.
Before I leave you, I will post several ideas that made Onno stronger
to the technical discussions forum.
I have not listened to the discussions in the fora over the last
years. So it is possible that somebody else has invented the same
independently. Also of course these ideas might have been present
before Onno and I might just have overlooked them. I did not do
intensive research now if my ideas are known. I will just post
them. If they are known, somebody else will be able to point out this
in a few minutes, with much less effort than a research in the forum
history would take me. The only thing I did is a quick check if my
ideas are already contained in Stockfish. Some are to some extend; I
will comment on this in the technical discussions.
Of course I hope that you will find my ideas useful, but if they are
not, I would not be too surprised or disappointed. If they are not yet
known, they should be tried in Stockfish.
If my ideas are any good, Vas will read and implement them too. But a
community has more manpower than Vas and his few employees. So if the
community is well organized, they might win the game of time. Good
luck fighting Rybka and Houdini this way!