Marco steps down as Stockfish maintainer

[Henk]

I was wondering if there was going to be a name change now (like glaurung -> stockfish), but I guess this would be a bad idea because Stockfish is already quite well known and has been growing a lot this year as a brand, so I would keep the name.

With all due respect,Stockfish is an extremely stupid name for a monster computer chess entity and I said that many many moons ago....

I hope a new powerful fork will appear soon with a proper name....
Dr.D

For Dutch readers only: I think I call my next engine "Koppie Koppie".

[Red Hood]

I think Marco is behaving a little childish. If it's not my way than it's the long way. I think he should be more flexible. I do agree that he is best leader at the moment. Maybe some one can replace him in the future, but for a reason not like this. All this attention seeking tantrum is not good for anyone. His selfish ways are putting Stockfish in jeopardy.

He has a very pragmatic approach to programming which i adhere to. Sometimes he takes it too far.

He should stay and that is my point. He should curb his impulses a little and people should fallow some rules and discipline. A little bit of effort from both parties and SF will go humming along! Marco should stop this nonsense and go back helping SF become even stronger!

[tttony]

Come on!!!

Now that SF5 is top marco is giving up

Total mess!!!

Now who will be in charge?

[mwyoung]

This is just sad...

Now the Stockfish project is in disarray. There will be dozens of different forks by dozens of random people all claiming to be the "official heir" to stockfish.

It is not like in 2008 when glaurung --> stockfish was more of a private affair and probably 3-4 people actually paid attention to both projects. This time in 2014, stockfish is the top engine and the whole world has eyes on it.

I predict the next few months to have no progress for SF, with competing ideas and models all fighting to replace stockfish as the official fork.

Since fishtest can only run official forks, I don't know what will happen since now there is no official fork.

Probably lots of chaos.

Sometimes when people reach their goals. They look for reasons to leave what they have been doing. It happens, but this is how progress is also made. If not by Stockfish with fresh minds and new ideas. It will be by someone else and their program.

Someone will always want to be the next King. The question is who the next King will be, and when will they topple Stockfish.

And the cycle continues, I have seen them all rise and fall.

[Uri Blass]

Some background may be found here with the details of his latest dispute with other Stockfish developers :

https://github.com/mcostalba/Stockfish/pull/243

Interesting.

I personally understand Marco's unwillingness to commit that patch. For starters it removes future flexibility which just doesn't seem worth it. It's a hacky hack with unclear ramifications and it was being pushed in ugly ways.

I think Marco should have made very clear that he has veto powers on fundamental design decisions and that anyone not happy with a decision can start his own fork.

In a way that's what he has done now. If the official SF repo simply tracks his tree, then all is well

I did not care to understand the patch so I am not going to comment about the specific patch(I never cared to understand how stockfish use hash tables) but I also do not understand your comment and I do not understand what is the problems that you see with the patch.

1)What do you mean by
"remove future flexibility"?

If it means that there is something that you can do today and you will not be able to do it in the future then I want to understand what is the relevant thing.

2)"hacky hack with unclear ramifications"

I do not understand this comment
if unclear ramifications mean that there is a risk that the patch does not perform well at longer time control then it is possible to test at significantly longer time control like 5 minutes per game or 10 minutes per game to check it before accepting it
but based on knowing marco in this case he is going to complain that people waste resources because he does not have patience to wait and I am not sure if people who give computer time have the same opinion(He never asked the people who give computer time to vote if they want to use significantly longer time control in some cases).

3)"it was being pushed in ugly ways."
Again I do not understand this comment and do not understand what is wrong.

[Uri Blass]

I think Marco is behaving a little childish. If it's not my way than it's the long way. I think he should be more flexible. I do agree that he is best leader at the moment. Maybe some one can replace him in the future, but for a reason not like this. All this attention seeking tantrum is not good for anyone. His selfish ways are putting Stockfish in jeopardy.

He has a very pragmatic approach to programming which i adhere to. Sometimes he takes it too far.

He should stay and that is my point. He should curb his impulses a little and people should fallow some rules and discipline. A little bit of effort from both parties and SF will go humming along! Marco should stop this nonsense and go back helping SF become even stronger!

I think that marco has the right to stop and does not have to give reasons.
Nobody pay him for working on stockfish so blaming him about it is not the right thing to do.

The main problem that I have with his behaviour is that he complain about people.

This is typical of marco:
"Actually you ignored almost all of my comments, so I really don't know why now I should invest my time to rewrite your patch in a standard way."

I can see that Ron explained that he did not ignore marco's comments and it seems that marco simply ignore other people comments.

For example here is a comment of me that he ignored about some magic number in stockfish.
Edit:I see that the link does not work but title of the subject in the stockfish forum is

magic number to remove from stockfish and an idea for testing so the queue is not going to be empty and the subject is from 7.6

https://groups.google.com/forum/?fromgr ... HKFcWe_bBo

Stockfish still has the number 128 in the end of material.cpp and I got no explanation why not to replace it by PHASE_MIDGAME
in the following code

Code: Select all

return  npm >= MidgameLimit ? PHASE_MIDGAME
        &#58; npm <= EndgameLimit ? PHASE_ENDGAME
        &#58; Phase&#40;(&#40;npm - EndgameLimit&#41; * 128&#41; / &#40;MidgameLimit - EndgameLimit&#41;);

Note that I understand that marco may have no time to read every post but in this case he responded to the relevant subject when he ignored my comment so it is not that he did not read the subject (and even if he does not read all the subject then I expect him in case of responding to a subject at least to read the title and understand what it is about so read the first post).

[M ANSARI]

I think SF has just grown too big for one person to handle it. I don't think Marco said he will "quit", he just wants more flexibility deciding what should go in and out without having to discuss the issue with what is now a very large group ... and the larger the group the bigger the chances of bitter disagreements. SF is open source, so most likely the progress in strength will not stop ... whether it is a new fork or an upgraded dev is still to be seen. There is still a lot of improvement to be made and I really doubt the progress in ELO strength will stop. I do agree that very quick time control games are good to a certain point, and that then you have to choose some LTC games to test some ideas that might not pass lightning TC games. The most desired goal for an engine is an engine that plays the most correct chess or strongest chess at LTC or that gives the best move in analysis mode.

[mar]

Hmm, I think it's important that Marco stays as a maintainer.
As the strongest open source engine, it's important that SF keeps the current quality, many people will study the source (thus there is some responsibility for open source authors).
If they start accepting patches like that, it will soon become umaintainable mess. Elo isn't everything!

Some remarks about the new patch (I hope the calculations are correct):
new patch uses 32 bytes/bucket, iterating over 3 entries instead of 4 (so only 3 choices vs 4 for replacement strategy),
it's no longer cacheline-aligned but this is no problem as the entry is smaller and can't cross the line anyway.
There are also some dirty micro-optimizations (the patch author tried hard).
Obviously it can accomodate 1.5x more entries than the original implementation, but it sacrifices effective resolution:
assuming a 4 megs hash table:
original TT implementation has 48 bits effective resolution (64k buckets = 16 lsbits, 32 msbits stored)
this new implementation has 33 bits (128k buckets = 17 bits, 16 msbits stored)
Add 1 bit for each doubling of TT size.
I think much more extensive tests should be done because it's unclear to me what impact it may have (maybe it's enough?):
for 1G TT, the new implementation is only 41 bits (vs 56), which is still less that 4M original...

[Henk]

I think Marco is behaving a little childish. If it's not my way than it's the long way. I think he should be more flexible. I do agree that he is best leader at the moment. Maybe some one can replace him in the future, but for a reason not like this. All this attention seeking tantrum is not good for anyone. His selfish ways are putting Stockfish in jeopardy.

He has a very pragmatic approach to programming which i adhere to. Sometimes he takes it too far.

He should stay and that is my point. He should curb his impulses a little and people should fallow some rules and discipline. A little bit of effort from both parties and SF will go humming along! Marco should stop this nonsense and go back helping SF become even stronger!

I think that marco has the right to stop and does not have to give reasons.
Nobody pay him for working on stockfish so blaming him about it is not the right thing to do.

The main problem that I have with his behaviour is that he complain about people.

This is typical of marco:
"Actually you ignored almost all of my comments, so I really don't know why now I should invest my time to rewrite your patch in a standard way."

I can see that Ron explained that he did not ignore marco's comments and it seems that marco simply ignore other people comments.

For example here is a comment of me that he ignored about some magic number in stockfish.
Edit:I see that the link does not work but title of the subject in the stockfish forum is

magic number to remove from stockfish and an idea for testing so the queue is not going to be empty and the subject is from 7.6

https://groups.google.com/forum/?fromgr ... HKFcWe_bBo

Stockfish still has the number 128 in the end of material.cpp and I got no explanation why not to replace it by PHASE_MIDGAME
in the following code

Code: Select all

return  npm >= MidgameLimit ? PHASE_MIDGAME
        &#58; npm <= EndgameLimit ? PHASE_ENDGAME
        &#58; Phase&#40;(&#40;npm - EndgameLimit&#41; * 128&#41; / &#40;MidgameLimit - EndgameLimit&#41;);

Note that I understand that marco may have no time to read every post but in this case he responded to the relevant subject when he ignored my comment so it is not that he did not read the subject (and even if he does not read all the subject then I expect him in case of responding to a subject at least to read the title and understand what it is about so read the first post).

I don't like these people telling other people what they should do in their scarce free time. Computer Chess is just a strange hobby or paid work for scientists.

[Uri Blass]

I do not tell marco what to do with his free time and I have no problem with the fact that he decided to stop to be a maintainer.

The problem for me is that it seems to me that marco use double standards.
It seems that he is complaining that other people ignore him and at the same time ignore other people.