Relative Filter Graph

[daSilva65]

Relative Filter Graph is not working for me in Windows 11. Is it working for somebody?

By the way, I would prefer the relative values in percentage (%) than in permillage (‰), because percentage is widely used and we usually think in percentage. All other relative values in scidCommunity are in percentage.

[whelanh]

Your problem with the graph was not Windows-specific. It appears to me that "Rel. Graph" doesn't work in Scid either. Somewhere along the way someone abandoned this code and the numerator and denominators used no longer made sense. I have put a fix in the v5.1.3.03 version (which will honor whether you have "all games" unchecked in your Tree View). However, I didn't change to % to stay consistent with the tradition of Scid and Scid vs. PC.

[daSilva65]

In fact the Filter Relative Graph works perfectly as well with Scid vs. PC and, by my suggestion, the graph was converted to percentage some 5 years ago. I told the programmer than I was dividing all the values by 10 by head calculation, all the time, to convert from permillage to percentage. In fact permillage in graphs is like a rare occurrence. The moves are all in percentage in the tree, so why not in the graph?

Anyway, congrats. You keep amazing us with your speed and easiness of work. And thanks a lot. Brilliant.

[daSilva65]

I feel like there should be a way to make it ignore rating. If one wants to display popularity of an opening position over time, old games have less rated players and per-millage is artificially lowered using rating in the filter. I can set rating from 0 to 4000 but I still get only 85 per mill at higher year after 1. e4 when it should be like 480 per mill by the 48,2% in the tree for the first moves.

[daSilva65]

What means moves interval in the graph configuration?

[whelanh]

@daSilva65: your questions are good ones. The moves interval is only for the 3rd graph and sets the bounds of the x-axis. The restriction on rated games only applies to the second graph by rating category (because if it doesn't have a rating, it can't fit it in a rating category). In terms of the difference between what you are seeing between the Tree View number of games and the graphs in the first (by year) graph may have to do with the fact that the Tree View will show games without a proper date, but the graph obviously can't include those games.

I've looked deeper into the math of the graphs and am not happy with it. Dividing by the total number of games in the database doesn't make much sense. Each point should be "selected game/opening" divided by total games in that period/category. So I am making some changes (and yes...I now agree with you that percentage makes more sense!). So look for changes in the next version (v5.1.3.04)

[daSilva65]

I am really a fan of the Relative Filter Graph. Here the evolution of popularity of the Caro-Kann defense in my bigger database. It shows at around 4,1% of games in the database and it's popularity has been growing consistently over the years. Went high in the early 1930's and in the past few years even higher. Caro-Kann is now at the peak of it's historical popularity. Now this is done perfectly and this Relative Filter Graph is on top of chess database features. Congratulations.

[whelanh]

Interesting. Thank you for pointing out it wasn't working. Nice to have it functional again.