lichess.org
Donate

The xG of Chess: Shark Points

#31
@Testrider said in #30:
> This article has a lot of text, but if I understand correctly the only thing that was actually done was mapping the centipawn eval to a simple curve between 0 and 1?

I think the first part is the key. Sometimes doing comes after. I think there is a lot of thinking to be done, before the doing. And I thank the blog author for the football analogy (more than the shark one, personal taste, or understanding of shark behavior).

The probabilistic context is introduced early. And even if I got the advice direction wrong, what matters is identifying the dimensions, and that was done.. Then it might need work on the actuall mathematical framework, but I think the point of the blog was suggesting there was more to a score 1-dimensional value of a position for game outcome purposes, from the human chess point of view.. Of course I am adding my salad sauce.

There is a lot of repetitive doing running in circle being done to my taste, sometimes getting head of that water, helps doing something else. As I said.. chess does involve some research into the fog. Otherwise we might always play speed tic-tac-toe.
#32
flash-back: I pointed to the lichess user base that might be sharing the property of avid learner (ready to explore and learn from their exploration, those that can think on their feet relying on their own abilities, in part at least).

Devils advocate (I need to put warnings, less fun, but maybe needed):

Perhaps to help those that need user case number to shoot their dog (that is having rabies), I should add avid active learners.
or autonomous avid learners?

So, we could deprecate or keep running in circles, on the teaching side to keep only the passive learners. more streaming.. more imitation learning.. more oracles.... (I must have relatively bad mood humor in my blood, but this post, is taking care of that).. Deprecation galore... Ahh, the good old times of network TV channels with captive audiences....

Same argumentation or dynamic patterns for deprecation can apply to status quo (or very fast dynamic running in circle).
#33
Good points for Lichess: keeping a gamut of SF choices. acorss the main versions transitions.. now only the user control of input depth instead of seconds, and at least we could be, as users doing the research to prove the concept I propose.

not wanting the dev side to put finger in the video mentioned dev nightmare, cogs of features interactoin crumbling onto its own weight, is one thing, preventing new information to arise from actively participating proactive users using the tools with own reasoning and autonomous learning plans is being blind or letting the fear become ideology or assumption preventing new data.

Some principles need to be relaxed and made concerns or hypotheses to keep in mind. Not lead chaperons.
#34
@BenjiPortheault great post.

What are the "Save" point in the final table ? Are they just the opposite of shark points = how deep in trouble the player once was during the game, yet they managed to draw it ? So Nepo was the best at not-losing (OR, giving winning chances to his opponents, but chances that were impossible to spot)

Seems the Waste and the Save points add up to the same total so I'm guessing it's that.
#35
@lecw said in #34:
>
do you follow the relation between the box and arrow figure, the shark points and the save points? I think you are likely right as the quantities all rely on one dimension measure and either outcome (conversion) or gain not same amplitude as expected from upstream position. No sense of the basis on which the engine was attributing such upstream score and how the human departed from the assumed SF best path selected. But having confirmation from author would be nice.

zero sum. 2 players. same quantity for both players at any position. degree of freedom for variables...
#36
@dboing said in #35:
> do you follow the relation between the box and arrow figure, the shark points and the save points?

Yeah I think I followed the post. If white at any point got the eval of the game to go up to +2.5 (250 centipawns) that converts to 0.8 "expected score" (probability of winning, based on the sigmoid curve they show) and so they get 0.3 shark points if the game ends in draw.

I did not read everything you posted. If you are looking to get clarifications from the OP, I would advise you to phrase your thoughts a little bit more concisely ^^"
#37
@lecw said in #36:
>

as concise and omitting information as a 1 dimensional SF score. Different objective.. I seek information. not concision. or information has priority over concision. sorry. I do not expect everyone to read everything. I am debating out loud. While there is little debate potential left on lichess. With the blog infrastructure directional bias. This is what I get in my face whenever I land on lichess, the blogs. I like discussions, even others discussions. Short blogs approximate that (for those wanting a finish product, often propagated from elsewhere). Original lichess blogs are closest to what forum discussions with front page visibility might have been, at least I react the same way.
It might be a more consistent post to start with, important is to catch and understand the question. In that discussion light, expecting a done deal, is not what I seek, from a lichess community space. I am one lichess user.

multidirectional debate with search of knowledge in common, is not as accessible as it used to be..... If there were a better infrastructure for multiple information sources and debate then I could just be part of it, and multiple capable users could deliver each concisely different point of view or own sub dialogs, distributing the search for better understanding of difficult questions, not yet solved as computer implementations, so needing chewing still. no place on dev spaces for that.. better do it in a chess first space. Deal with what is still there.

but this blog and comment structure kind of makes people having less voice. So good on the author of the blog to not give a completely vetted report. But ask the question, and suggest an implementation to illustrate that it could work. We can work from there. Let's start with the following post.