TM Dendogram - Christmas 2013 Edition

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

Moderators: hgm, Harvey Williamson, bob

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
User avatar
Marek Soszynski
Posts: 499
Joined: Wed May 10, 2006 5:28 pm
Location: Birmingham, England

Re: TM Dendogram - Christmas 2013 Edition

Post by Marek Soszynski » Wed Dec 18, 2013 7:34 am

ThomasJMiller wrote:Everyone can nowadays read a dendogram, so you can draw your own conclusions by yourself.
So, the most independent-minded (so to speak) engine of all those tested is Shredder, which hasn't been updated for years.
Marek Soszynski

ThomasJMiller
Posts: 96
Joined: Fri Jan 11, 2013 11:22 am

Re: TM Dendogram - Christmas 2013 Edition

Post by ThomasJMiller » Wed Dec 18, 2013 9:07 pm

michiguel wrote:The confidence of the branches cannot be assessed if a bootstrap analysis is not done. Quite likely, this protector similarity is meaningless.

Miguel
You are right Miguel. Unfortunately I'll be very busy till the end of the year. When possible I'll make some analysis of the data.

Nonetheless, results <= 5 are very interesting, we have VERY high percentages there (>70-80-90% ....while the average is about 49%).

ThomasJMiller
Posts: 96
Joined: Fri Jan 11, 2013 11:22 am

Re: TM Dendogram - Christmas 2013 Edition

Post by ThomasJMiller » Wed Dec 18, 2013 9:10 pm

Marek Soszynski wrote:
ThomasJMiller wrote:Everyone can nowadays read a dendogram, so you can draw your own conclusions by yourself.
So, the most independent-minded (so to speak) engine of all those tested is Shredder, which hasn't been updated for years.
all branches >= 15 are surely meaningless...

mar
Posts: 1832
Joined: Fri Nov 26, 2010 1:00 pm

Re: TM Dendogram - Christmas 2013 Edition

Post by mar » Wed Dec 18, 2013 9:36 pm

Nice tree Thomas. Did you use fixed time/move or fixed depth?
I have to admit that cheng cheats a bit as it uses 100msec less time in fixed time mode :)
Next version (if any) won't use this "feature" anymore.

elcabesa
Posts: 679
Joined: Sun May 23, 2010 11:32 am
Contact:

Re: TM Dendogram - Christmas 2013 Edition

Post by elcabesa » Wed Dec 18, 2013 9:37 pm

velmarin wrote: Marco, do not follow this path.

I indigestible theme clone.
Vajolet is a derivative by Winglet , educational engine by Stef Luijten.

Keep up the good work, but not entangled.
Greetings Jose.
Hi Jose I really don't understood what you mean. :D
??I indigestible theme clone.??

Vajolet is a derivative by Winglet , educational engine by Stef Luijten. I started studying Stef Luijten code, and some of his code is still here (magic bitboard code for example), then studied Fruit , Cheng, Stockfish. If you had read my code yuo'll see lot of of idea from stockfish.

Keep up the good work, but not entangled. what do you man by entangle? bluff , deceive, cheat? It's not my purpose.
I don'tknow what you mean by derivatve work. if you are talking about forking a project I don't think I have forked it. I started my project from scratch after playing around with winglet code for some time.

User avatar
velmarin
Posts: 1579
Joined: Mon Feb 21, 2011 8:48 am

Re: TM Dendogram - Christmas 2013 Edition

Post by velmarin » Wed Dec 18, 2013 9:56 pm

elcabesa wrote:I started my project from scratch after playing around with winglet code for some time.

HEHE, from scratch, well, congratulations. :D

Uri Blass
Posts: 8037
Joined: Wed Mar 08, 2006 11:37 pm
Location: Tel-Aviv Israel

Re: TM Dendogram - Christmas 2013 Edition

Post by Uri Blass » Wed Dec 18, 2013 11:08 pm

ThomasJMiller wrote:Will programmers receive toys and candy or coal this Christmas? To find out, let's check Santa's Naughty or Nice List :D
Latest releases of many commercial and free engines have been included together with the most important open source ones.

Testing all the engines and preparing the dendogram has taken a lot of time so I wouldn't like this thread to be moved to the Engine Origins subforum. Therefore, please, avoid making comments like XXX is clone of YYY or so.
Everyone can nowadays read a dendogram, so you can draw your own conclusions by yourself.

Thanks to Adam Hair who helped me improve selfsimilarity results.

Image
Can you explain the graph?

I understand that movei is similiar to gnuchess5.50 or fruit2.1(not sure if to both of them or only to one of them) but I do not understand the meaning of the numbers in the graph(when I say similiar I mean relative to the difference of movei with other engines).

Note that I am not surprised if you find that movei is more similiar to fruit2.1 relative to other engines because in the latest movei's before I stopped developing it I added some evaluation terms based on ideas that I understood from fruit2.1's code.

Uri

User avatar
michiguel
Posts: 6269
Joined: Thu Mar 09, 2006 7:30 pm
Location: Chicago, Illinois, USA
Contact:

Re: TM Dendogram - Christmas 2013 Edition

Post by michiguel » Wed Dec 18, 2013 11:35 pm

Uri Blass wrote:
ThomasJMiller wrote:Will programmers receive toys and candy or coal this Christmas? To find out, let's check Santa's Naughty or Nice List :D
Latest releases of many commercial and free engines have been included together with the most important open source ones.

Testing all the engines and preparing the dendogram has taken a lot of time so I wouldn't like this thread to be moved to the Engine Origins subforum. Therefore, please, avoid making comments like XXX is clone of YYY or so.
Everyone can nowadays read a dendogram, so you can draw your own conclusions by yourself.

Thanks to Adam Hair who helped me improve selfsimilarity results.

Image
Can you explain the graph?

I understand that movei is similiar to gnuchess5.50 or fruit2.1(not sure if to both of them or only to one of them) but I do not understand the meaning of the numbers in the graph(when I say similiar I mean relative to the difference of movei with other engines).

Note that I am not surprised if you find that movei is more similiar to fruit2.1 relative to other engines because in the latest movei's before I stopped developing it I added some evaluation terms based on ideas that I understood from fruit2.1's code.

Uri
To make it simple, Movei is not significantly related to any engine from this graph.

There are three clear clusters
1) Heron, Stockfish 4, Amitis, SF DD, SF 2.3.1, SF 3

2) H4, Robodini, IH, Vitruvius, Robbolito, Fire, DSaros, Bouquet, Critter, Strelka 5.5, BlackMamba, Gull2.3

3) Naum4.2, Strelka2, Murka3, Fruit2.1, Fruit2.2.1, D Onno, Hamsters, Alaric, colossus

There are 4-5 engines that require a closer analysis that cannot be extracted from what we see. The rest seems to be very unclear, or not related at all.

The farther to the right is the branch point, the less significant is the relationship of the leaves. But you cannot know from this only where to draw a line of "significance".

Miguel

User avatar
michiguel
Posts: 6269
Joined: Thu Mar 09, 2006 7:30 pm
Location: Chicago, Illinois, USA
Contact:

Re: TM Dendogram - Christmas 2013 Edition

Post by michiguel » Wed Dec 18, 2013 11:37 pm

ThomasJMiller wrote:
michiguel wrote:The confidence of the branches cannot be assessed if a bootstrap analysis is not done. Quite likely, this protector similarity is meaningless.

Miguel
You are right Miguel. Unfortunately I'll be very busy till the end of the year. When possible I'll make some analysis of the data.

Nonetheless, results <= 5 are very interesting, we have VERY high percentages there (>70-80-90% ....while the average is about 49%).
I have all the scripts and programs to do the bootstrap analysis. If you send me the .dat files, I can do those for you.

Miguel

User avatar
michiguel
Posts: 6269
Joined: Thu Mar 09, 2006 7:30 pm
Location: Chicago, Illinois, USA
Contact:

Re: TM Dendogram - Christmas 2013 Edition

Post by michiguel » Wed Dec 18, 2013 11:41 pm

Marek Soszynski wrote:
ThomasJMiller wrote:Everyone can nowadays read a dendogram, so you can draw your own conclusions by yourself.
So, the most independent-minded (so to speak) engine of all those tested is Shredder, which hasn't been updated for years.
If you look at only this, the less similar to the rest seems to be Deep Sjeng, Gaviota, Vajolet, (in that order) and then Shredder and Zappa.

But the difference is all noise.

Miguel

Post Reply