Excellent and very valuable work. :) I hope you keep it up. Having a website for querying the latest ratings calculated this way would be awesome. Not a futile effort at all.
@bubbleboy said in #72:
> Hey @justaz , I'm a mathematical physicist more than a statistician by training, so it would be great if in this or future blogs you include a reference to how Ordo works, as it will be easier for casual/new readers like me to jump into the content faster :)
Thank you for your response! Unfortunately I do not fully understand how Ordo works. However, I'm working on my own program that seems to be 1000x faster. Which I also do not understand how it works haha
github.com/jazzzooo/Lichess-Ordo/tree/master/jazzo
Here is a link to the WIP version. I've also posted some cool graphs on the Lichess discord discussion here: discord.com/channels/280713822073913354/1190085963339272344
> From a logical perspective, I would have a lot of the same comments as @MyPoorRook around hand-wavy statements, but I am happy to see your open and measured responses :)
I was never expecting to gain that much traction with my post. In fact I wrote it in one sitting during my one day off from work on Christmas. I told my friend "I probably should sleep and write a second draft tomorrow, but I won't have time, so let's just post it". That is why a lot of it seems handwavy, and also why the graphs are just whatever matplotlib spit out without any adjustments.
December data is coming out any second. I will be testing the Ordo, Lichess and my own ratings from the November data and seeing how well they can predict the games. I'll try to be more rigorous and have better graphs. You can see my proposed methodology in comment #55
Here is a teaser, a graph produced by optimizing ratings in my program:
> Hey @justaz , I'm a mathematical physicist more than a statistician by training, so it would be great if in this or future blogs you include a reference to how Ordo works, as it will be easier for casual/new readers like me to jump into the content faster :)
Thank you for your response! Unfortunately I do not fully understand how Ordo works. However, I'm working on my own program that seems to be 1000x faster. Which I also do not understand how it works haha
github.com/jazzzooo/Lichess-Ordo/tree/master/jazzo
Here is a link to the WIP version. I've also posted some cool graphs on the Lichess discord discussion here: discord.com/channels/280713822073913354/1190085963339272344
> From a logical perspective, I would have a lot of the same comments as @MyPoorRook around hand-wavy statements, but I am happy to see your open and measured responses :)
I was never expecting to gain that much traction with my post. In fact I wrote it in one sitting during my one day off from work on Christmas. I told my friend "I probably should sleep and write a second draft tomorrow, but I won't have time, so let's just post it". That is why a lot of it seems handwavy, and also why the graphs are just whatever matplotlib spit out without any adjustments.
December data is coming out any second. I will be testing the Ordo, Lichess and my own ratings from the November data and seeing how well they can predict the games. I'll try to be more rigorous and have better graphs. You can see my proposed methodology in comment #55
Here is a teaser, a graph produced by optimizing ratings in my program:
