Ordo 1.0.9 (new features for testers)

[drj4759]

I think name alias substitution is very useful in ordoprep. I use ordoprep very heavily in my Owl rating list which is where I extract subsets of engines for selective rating. At this moment I had a custom python script to make the hybrid rating list possible from inconsistent naming of engines which could be made faster if the name substitution will be incorporated in ordoprep.

[Ajedrecista]

Hello again:

Suppose a player played against 3 opponents (1,2,3) the following amount of games: n1, n2, and n3.

(total games) N = n1 + n2 + n3

fraction of games played
f1 = n1/N
f2 = n2/N
f3 = n3/N

Then the following summation:

I = - (f1 * log(f1) + f2 * log(f2) + f3 * log(f3))

I is the Shannon information content and log is natural logarithm

OppDiv = exp (I)

Miguel

That's it! I have just found it here, once you explained that:

https://en.wikipedia.org/wiki/Entropy_( ... Definition

I can reproduce Ordo's value now. Thanks for your answer!

Regards from Spain.

Ajedrecista.

[michiguel]

I think name alias substitution is very useful in ordoprep. I use ordoprep very heavily in my Owl rating list which is where I extract subsets of engines for selective rating. At this moment I had a custom python script to make the hybrid rating list possible from inconsistent naming of engines which could be made faster if the name substitution will be incorporated in ordoprep.

Yes, it is in the to-do list.

Miguel

[Ozymandias]

regarding players with few games. It's a bit inconvenient, that you have to use the -g switch when some players are under-represented. It should be possible to specify the minimum number of games for a player to be rated, and after the program runs, Ordo could list those that were left out and how many games they had. In a related request, I would also include an option for the user to specify the percentage of games a cluster needs, to be rated. Below that mark, I wouldn't rate games from marginal clusters.

ordoprep can clean up a file removing players setting some limits.

When I mentioned "players with few games", I thought it would be clear, from context, that I was referring to those that link clusters; which I'm guessing are the ones responsible for Ordo, refusing to process a particular PGN, unless you run it with the -G switch.
At least the request, about ignoring relatively small clusters, has been noted?

[michiguel]

regarding players with few games. It's a bit inconvenient, that you have to use the -g switch when some players are under-represented. It should be possible to specify the minimum number of games for a player to be rated, and after the program runs, Ordo could list those that were left out and how many games they had. In a related request, I would also include an option for the user to specify the percentage of games a cluster needs, to be rated. Below that mark, I wouldn't rate games from marginal clusters.

ordoprep can clean up a file removing players setting some limits.

When I mentioned "players with few games", I thought it would be clear, from context, that I was referring to those that link clusters; which I'm guessing are the ones responsible for Ordo, refusing to process a particular PGN, unless you run it with the -G switch.
At least the request, about ignoring relatively small clusters, has been noted?

What it could be removed are the small clusters themselves. There are no players who link the clusters, otherwise, they won't be a separated cluster. I am considering to provide rankings for each of the clusters in separate files (which is one of the suggestions from Dann Corbit). I guess this is what you mean.

Miguel

[michiguel]

Two new features

https://github.com/michiguel/Ordo/releases/tag/1.0.9.8

Code: Select all

-i <file> include only games of participants present in <file>
-x <file> names in <file> will not have their games included

In certain situations, the user may want to include/discard in the calculation only a subset of the games present in the input file/s. Switches -i <file> and -x <file> are used for this purpose. Switch -i includes only the games of participants listed in <file>. In this file, each participant name has to be in different lines. Also, each of those names may or may not be surrounded by quotes. Both are accepted. For this reason, if a .csv file is provided as a list of participants, only the first column is read. In addition, -x could be use to exclude games of participants listed in <file>.

The file that will contain the list of participants will look like this

Code: Select all

Gaviota 1.0
Crafty 25.0
Stockfish 7.0
etc.
etc.

or

Code: Select all

"Gaviota 1.0"
"Crafty 25.0"
"Stockfish 7.0"
etc.
etc.

And a modification in the format:

In the output, the column that contains the score (%) was removed of the symbol %.

I hope it is clear what I meant.

Miguel

[Ozymandias]

regarding players with few games. It's a bit inconvenient, that you have to use the -g switch when some players are under-represented. It should be possible to specify the minimum number of games for a player to be rated, and after the program runs, Ordo could list those that were left out and how many games they had. In a related request, I would also include an option for the user to specify the percentage of games a cluster needs, to be rated. Below that mark, I wouldn't rate games from marginal clusters.

ordoprep can clean up a file removing players setting some limits.

When I mentioned "players with few games", I thought it would be clear, from context, that I was referring to those that link clusters; which I'm guessing are the ones responsible for Ordo, refusing to process a particular PGN, unless you run it with the -G switch.
At least the request, about ignoring relatively small clusters, has been noted?

What it could be removed are the small clusters themselves. There are no players who link the clusters, otherwise, they won't be a separated cluster. I am considering to provide rankings for each of the clusters in separate files (which is one of the suggestions from Dann Corbit). I guess this is what you mean.

Miguel

As initially stated, they're two different inconveniences. They're related, but they're two. The first one, is the need to having to use the -G switch. I could just include it, always, but I like to know when the database contains groups which are poorly connected. The second one, regarding clusters (which are a extreme case of poorly connected groups, ie. connected by 0 players), could be also addressed following Corbit's proposal, although it isn't what I had in mind. What the heck! I like his solution better

[drj4759]

1. I suggest the (%) sign for the Score field and the D% for Draw field should be changed to SCORE% and DRAW% for clarity and consistency.

2. The CFS(next) and OppErr columns are too wide. I suggest CFS% for the former and then shorten the widths of both columns. The entire 14 columnar output will not fit in a normal web page width which necessitates omission of some fields.

3. The W D L columns should be descriptive like WIN DRAW LOSS. No need to economize here. The engine names field is also a bit wide, it can be reduced to 1 or 2 less.

4. I hope you incorporate the suggested changes so that I won't have to make some modifications with the ordo source code everytime you make a release.

5. The inclusion/exclusion option is great. It makes selective extraction processing of millions of PGN games a breeze for rating list sites like mine. This feature should appear in ordoprep in addition to the name alias substitution.

Your rating utilities are great! I salute.

Thank you.
drj

[michiguel]

1. I suggest the (%) sign for the Score field and the D% for Draw field should be changed to SCORE% and DRAW% for clarity and consistency.

2. The CFS(next) and OppErr columns are too wide. I suggest CFS% for the former and then shorten the widths of both columns. The entire 14 columnar output will not fit in a normal web page width which necessitates omission of some fields.

3. The W D L columns should be descriptive like WIN DRAW LOSS. No need to economize here. The engine names field is also a bit wide, it can be reduced to 1 or 2 less.

4. I hope you incorporate the suggested changes so that I won't have to make some modifications with the ordo source code everytime you make a release.

5. The inclusion/exclusion option is great. It makes selective extraction processing of millions of PGN games a breeze for rating list sites like mine. This feature should appear in ordoprep in addition to the name alias substitution.

Your rating utilities are great! I salute.

Thank you.
drj

Hi Donald,

Many people like different things about the headers. For that reason, I decided to make it customizable.

https://github.com/michiguel/Ordo/releases/tag/1.1.0

A switch was added to control column format in the output

-b <file> format column output, each line being <column>,<width>,"Header"

This switch controls the width of each column in the output and the header of each column.
This <file> consists of a text in which each line has <column>,<width>,"HeaderName".
For instance, if you want to have the item 4 (games total, see manual.pdf), to be 5 characters wide and be named "Points", you will have to include a line like this:

4,8,"Points"

The default would be a file like this

Code: Select all

 0,  0, "PLAYER"
 1,  6, "RATING"
 2,  6, "ERROR" 
 3,  9, "POINTS"
 4,  7, "PLAYED"
 5,  7, "(%)"
 6, 11, "CFS(next)"
 7,  7, "W"
 8,  7, "D"
 9,  7, "L"
10,  7, "D(%)"
11,  7, "OppAvg"
12,  7, "OppErr"
13,  7, "OppN"
14,  7, "OppDiv"

You can edit that file and include the switch -b file.txt with the width and names that you like best.

Miguel
PS: Once the additions I included in Ordo stabilizes, I will move them to ordoprep.

[Frank Quisinsky]

Hi Miguel,

you are working comparable to Martin Blume.

Give him an idea ...
His final result after programming is better than the original idea!

Perfect for me!
Thanks again!

Best
Frank

PS: Ferdinand is such a candidat too! I am to old for programming but not to old to see that your work on Ordo is really fantastic.