Database snapshot

Discussion of anything and everything relating to chess playing software and machines.

Moderators: bob, hgm, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
User avatar
Master Om
Posts: 330
Joined: Wed Nov 24, 2010 9:57 am
Location: INDIA

Re: Database snapshot

Post by Master Om » Thu Oct 17, 2019 4:20 pm

noobpwnftw wrote:
Wed Oct 16, 2019 9:05 pm
Master Om wrote:
Wed Oct 16, 2019 2:19 pm
How can u confirm this. Am curious to know. AFAIK even numbers have a blind spot .
Roughly, at this depth it can hit "best" move with 70% probability, then among top 5 moves I have a statistically sound chance to be correct.
Then this process is done recursively so generally it wouldn't miss important lines, in practice, better moves can be discovered by self-play using deeper searches.
So depth 22 in MPV =5 or SPV ?
Always Expect the Unexpected

User avatar
Ovyron
Posts: 2699
Joined: Tue Jul 03, 2007 2:30 am

Re: Database snapshot

Post by Ovyron » Sat Oct 19, 2019 3:47 am

From what I can see it's single PV initially (for leaf nodes; unless he's putting the second move on the PV), then MultiPV if the lines are extended.

The point of Depth 22 is the speed, because if 70% of the time Depth 22 outputs the same move as Depth 70, then for those positions analyzing beyond Depth 22 is a waste of time.

For human interactive analysis of positions I think Depth 24 is the bare minimum, Depth 23 is very easy to beat so having those weak lines in analysis might be a waste. Depth 24 has been surprising in that I've seen it outperform Depth 27 (you put an equal position with White D24 against Black D27, at say, MultiPV=4 and D27 has to backtrack to play a different move first; what you've found is a position where white has the edge regardless of what engines claim.)
Great spirits have always encountered violent opposition from mediocre minds.

User avatar
Master Om
Posts: 330
Joined: Wed Nov 24, 2010 9:57 am
Location: INDIA

Re: Database snapshot

Post by Master Om » Sat Oct 19, 2019 4:44 pm

Ovyron wrote:
Sat Oct 19, 2019 3:47 am

if 70% of the time Depth 22 outputs the same move as Depth 70, then for those positions analyzing beyond Depth 22 is a waste of time
That is vety difficult to know if its waste of time or not. One has to go beyond that. many position requires one to go beyond depth 60 and for some instantly 22-15 is enuf. very much depends on position.
Always Expect the Unexpected

User avatar
Ovyron
Posts: 2699
Joined: Tue Jul 03, 2007 2:30 am

Re: Database snapshot

Post by Ovyron » Sun Oct 20, 2019 10:44 am

Master Om wrote:
Sat Oct 19, 2019 4:44 pm
That is vety difficult to know if its waste of time or not.
I can logically prove that it's a waste of time:

1. All positions where the top move is found at depth 69 or before have no reason whatsoever to go beyond it. Therefore, for those positions going to depth 70 is a waste of time.

2. If there's a move switch at depth 70 to a better move, time isn't wasted.

3. Depth 70 takes a very long while to reach.

4. If you can find this better move (from 2.) with any other analysis method (say, putting another engine to analyze the position and it likes the move much sooner), then waiting for depth 70 is a waste of time.

It follows that it's only productive time if reaching depth 70 is the fastest way to find the move.

Depth 70 is an extreme case, but in practice, even depth 40 is a waste of time. Unassisted engine at high depth is so bad that I don't leave any position analyzing overnight while I'm sleeping, even though I could, because I waste my time refuting those lines to that relative depth, it's better to just stay at low depth and find that "better move."

It'd be easy to disproving me by showing just a single position (from an actual game) where the engine needs depth 70 to find the best move (only one winning or drawing) and no other interactive analysis method finds it. I won't hold my breath while someone finds such a position.
Great spirits have always encountered violent opposition from mediocre minds.

User avatar
Master Om
Posts: 330
Joined: Wed Nov 24, 2010 9:57 am
Location: INDIA

Re: Database snapshot

Post by Master Om » Sun Oct 20, 2019 4:54 pm

Ovyron wrote:
Sun Oct 20, 2019 10:44 am
Master Om wrote:
Sat Oct 19, 2019 4:44 pm
That is vety difficult to know if its waste of time or not.
I can logically prove that it's a waste of time:

1. All positions where the top move is found at depth 69 or before have no reason whatsoever to go beyond it. Therefore, for those positions going to depth 70 is a waste of time.

2. If there's a move switch at depth 70 to a better move, time isn't wasted.

3. Depth 70 takes a very long while to reach.

4. If you can find this better move (from 2.) with any other analysis method (say, putting another engine to analyze the position and it likes the move much sooner), then waiting for depth 70 is a waste of time.

It follows that it's only productive time if reaching depth 70 is the fastest way to find the move.

Depth 70 is an extreme case, but in practice, even depth 40 is a waste of time. Unassisted engine at high depth is so bad that I don't leave any position analyzing overnight while I'm sleeping, even though I could, because I waste my time refuting those lines to that relative depth, it's better to just stay at low depth and find that "better move."

It'd be easy to disproving me by showing just a single position (from an actual game) where the engine needs depth 70 to find the best move (only one winning or drawing) and no other interactive analysis method finds it. I won't hold my breath while someone finds such a position.
U can't prove wrong with yard stick U are holding as the measurement is relative.
Your method works fine if u have no other work and only sit infront of the pc doing FB analysis saving the evals and then minimaxing it.
Some have busy schedules of work. Hence they feed the positions what needs to be analysed and leavfe the pc in overnight analysis and then back solve it.
Some play through the position irrespective of time or depth after running a deep analysis just to make sure which plan they gonna execute on the basis of their understanding of position and not what a program spits out.

Therefore depth 22 is as dumb as it sounds. Quality very much depends on time and the time given in human interaction to the position.
Always Expect the Unexpected

noobpwnftw
Posts: 360
Joined: Sun Nov 08, 2015 10:10 pm

Re: Database snapshot

Post by noobpwnftw » Sun Oct 20, 2019 11:35 pm

Master Om wrote:
Sun Oct 20, 2019 4:54 pm
Therefore depth 22 is as dumb as it sounds. Quality very much depends on time and the time given in human interaction to the position.
I assume that you can spend 100 years of your life time doing nothing else but "human interactions", and you have a relatively fast computer that can do depth 60 searches within 20 minutes on average for one position, this means 2,628,000 positions.
For these 2,628,000 positions, assuming there are 40 legal moves on average, a total of 105,120,000 moves takes about 116 hours with my current speed to evaluate every single one up to depth 22, now it only gets about 70% right, then a MultiPV=5 exploration is done for each one of them, provided that not a single duplicated position and it will take 580 hours to finish, now the correct rate in theory raises up to about 82%, and the new positions have 70% correct rate, and the total amount of time doing this is still within a month.

So I guess analyzing by hand is pretty much as dumb as it sounds.

User avatar
Master Om
Posts: 330
Joined: Wed Nov 24, 2010 9:57 am
Location: INDIA

Re: Database snapshot

Post by Master Om » Mon Oct 21, 2019 12:02 pm

noobpwnftw wrote:
Sun Oct 20, 2019 11:35 pm
Master Om wrote:
Sun Oct 20, 2019 4:54 pm
Therefore depth 22 is as dumb as it sounds. Quality very much depends on time and the time given in human interaction to the position.
So I guess analyzing by hand is pretty much as dumb as it sounds.
In that case humans are dumb as whatever the methods are ,all are done by humans only.
Always Expect the Unexpected

User avatar
Master Om
Posts: 330
Joined: Wed Nov 24, 2010 9:57 am
Location: INDIA

Re: Database snapshot

Post by Master Om » Mon Oct 21, 2019 12:04 pm

noobpwnftw wrote:
Sun Oct 20, 2019 11:35 pm
Master Om wrote:
Sun Oct 20, 2019 4:54 pm
Therefore depth 22 is as dumb as it sounds. Quality very much depends on time and the time given in human interaction to the position.
I assume that you can spend 100 years of your life time doing nothing else but "human interactions", and you have a relatively fast computer that can do depth 60 searches within 20 minutes on average for one position, this means 2,628,000 positions.
For these 2,628,000 positions, assuming there are 40 legal moves on average, a total of 105,120,000 moves takes about 116 hours with my current speed to evaluate every single one up to depth 22, now it only gets about 70% right, then a MultiPV=5 exploration is done for each one of them, provided that not a single duplicated position and it will take 580 hours to finish, now the correct rate in theory raises up to about 82%, and the new positions have 70% correct rate, and the total amount of time doing this is still within a month.

So I guess analyzing by hand is pretty much as dumb as it sounds.
" now it only gets about 70% right,"

Now how can u say its 70% right ? Is it an average for most positions or just an assumptions on observations of small data/positions ?
Always Expect the Unexpected

User avatar
Master Om
Posts: 330
Joined: Wed Nov 24, 2010 9:57 am
Location: INDIA

Re: Database snapshot

Post by Master Om » Mon Oct 21, 2019 12:15 pm

noobpwnftw wrote:
Sun Oct 20, 2019 11:35 pm
Master Om wrote:
Sun Oct 20, 2019 4:54 pm
Therefore depth 22 is as dumb as it sounds. Quality very much depends on time and the time given in human interaction to the position.
I assume that you can spend 100 years of your life time doing nothing else but "human interactions", and you have a relatively fast computer that can do depth 60 searches within 20 minutes on average for one position, this means 2,628,000 positions.
For these 2,628,000 positions, assuming there are 40 legal moves on average, a total of 105,120,000 moves takes about 116 hours with my current speed to evaluate every single one up to depth 22, now it only gets about 70% right, then a MultiPV=5 exploration is done for each one of them, provided that not a single duplicated position and it will take 580 hours to finish, now the correct rate in theory raises up to about 82%, and the new positions have 70% correct rate, and the total amount of time doing this is still within a month.

So I guess analyzing by hand is pretty much as dumb as it sounds.
Also my comment was in related to analyze positions of a corr game not to create a database of opening positions. Read what I wrote above in point 1 too.
Always Expect the Unexpected

noobpwnftw
Posts: 360
Joined: Sun Nov 08, 2015 10:10 pm

Re: Database snapshot

Post by noobpwnftw » Mon Oct 21, 2019 4:27 pm

Master Om wrote:
Mon Oct 21, 2019 12:04 pm
Now how can u say its 70% right ? Is it an average for most positions or just an assumptions on observations of small data/positions ?
It seems that I'm the one providing data(in billions of unique positions) and you are the one making assumptions out of nowhere, go figure.

Post Reply