Chess will never be solved, here's why

@12192

"an assumption that cannot be proven about underpromotions not being relevant"
++ Underpromotions to rook or bishop only make sense to avoid a draw by stalemate.
That could make sense for one side, but not for both sides.
In the perfect games we have, some promotions to a 3rd and/or 4th queen occur, but promotions to a 3rd knight, bishop, or rook do not occur. If underpromotion were limited to pieces previously captured, then all games would be the same.

"This previous step takes a 0.001% bite"
++ 1 in 10,000 is 0.01%. Tromp conjectured it to be only 1 in 10^6 i.e. 0.0001%.

"Not proven to be possible for chess as of yet, but true for checkers" ++ Alpha-Beta search is a universal method to prune search trees and does not depend on the game.

"The two 10^17 numbers represent entirely different things"
++ Yes, one is the number of positions to be considered to weakly solve Chess and the other is the number of positions actually considered during the ongoing ICCF WC Finals.
That they turn out the same shows that the effort is commensurate with the required effort.

Alpha Beta pruning *when optimal* can effectively reduce the game tree complexity to its square root. Checkers has an average branching factor of 10, chess has an average branching factor of 35. In order to achieve optimal pruning, the branching moves must be evaluated in the best possible order. This means that the best moves are to be considered first.

Slight problem: you have no way of ordering the average 35 moves perfectly (and, as you might imagine, ordering 35 in a more complex game correctly would be a very tall order compared to ordering 10 in a much simpler game), only via a vague approximation based on human and engine-derived valuations. So, I will maintain that it is not at all proven to be possible to achieve the 17 orders of magnitude downward leap you are positing.

@12216

"chess has an average branching factor of 35" ++ Where is your proof of that?
Here is proof that the average branching factor without transposition is 3.
The perfect ICCF WC Finals games draw in average 39 moves. Chess without underpromotions to pieces not previously captured has 10^38 positions. 10^38 = 3^80 = 3^(2*40).

Some positions have more branches: 20 in the initial position, but easy to order at least near perfectly. Some positions have only one legally forced move like Qxd8+ Kxd8, or only one practically forced move like a forced recapture. Apart from that Chess has many transpositions.

"This means that the best moves are to be considered first." ++ Yes, Best-first search

"you have no way of ordering the average 35 moves perfectly"
++ The ordering does not need to be perfect.

lmfao tygxc acting like all he saidhasnt been long debunked

""The two 10^17 numbers represent entirely different things"
++ Yes, one is the number of positions to be considered to weakly solve Chess and the other is the number of positions actually considered duri the ongoing ICCF WC Finals.
That they turn out the same shows that the effort is commensurate with the required effort"

no, thats's you claiming that two entirely different processes are the same type of work.

we have explained many times how it is completely different. A node is not a full positional evaluation. you just wave your hands and pretend that the objective rebuttal to your ridiculous claims doesnt exist.

"The perfect ICCF WC Finals games draw in average 39 moves. Chess without underpromotions to pieces not previously captured has 10^38 positions. 10^38 = 3^80 = 3^(2*40)."

that doesnt count as "proof" because your 10^38 has not been sufficiently demonstrated, and it ignores the fact that positions can be reached by multiple independent trees.

but of course rigor isnt something you are fond of considering

""This means that the best moves are to be considered first." ++ Yes, Best-first search" your calculation has no room for deciding which moves are best.

tygxc really does reach anti vaxxer levels of logical inconsistency.

tygxc, why havent you addressed the fact that dozens of math majors and math professors all agree with the fallacies i point out in your illogic?

Dio, elroch and MAR, i feel like we have to unite our approaches to MAYBE get tygxc to shut up and understand basic logic.

because right now he's just selectively ignoring rebuttals that he cant handle, and "addressing" the ones that he can take out of context and lie about.

I got 35 from here: https://en.wikipedia.org/wiki/Branching_factor

You know, the definition of branching factors...? Chess is their main example. The wikipedia definition references two other works:

https://www.wired.com/2014/05/the-world-of-computer-go/

https://www.gamedev.net/tutorials/_/technical/artificial-intelligence/chess-programming-part-iv-basic-search-r1171/

If the moves are not ordered correctly, then you are not going to get down to the square root. The further from perfect, the more off your premise is...

@12220

"I got 35 from here" ++ Yes, that 35 are the also mentioned 31 may be plausible,
but does not take transpositions into account.
It often happens that moves A and B are available, but at the next moves AB or BA are available, collapsing to the same position.
A position should be considered once only, regardless of how it was reached.
Transpositions are no exception, but the rule.
10^38 = 31^25
With a non transposing branching factor of 31 all chess games would be over in 13 moves.
There can be no more chess positions than there are chess positions.
That is the pigeonhole principle

ur talking abt bruting 10^x layouts. know that AI is gonna bring that # way-WAY dn. i know mosta u beauzeauz cant for the lyfa u bleeve that ?...but trust me...just watch. its coming. yee !

tygxc you were just objectively demonstrated to be incorrect and this has had zero effect on your claims

tygxc why arent you addressing the fact that the 10^17 nodes used in the correspondance chess are not full positional evaluations?

it takes 10^17 nodes to complete 100 games with reasonable confidence of no errors. thats 10^-15 of the way there.

u guys are tryn2 completely describe chess. u approaching it WRONG. know why ?...cuz AIs gonna remove the need to. s/t in the future ur gonna hafta accept that not EVERY position will need a full eval. ur tiny mind is looking 4a solution from what u know (iows the past). u needta take urself beyond that. i know thats hard for those w/a lactose backbone. dont be one a them.

u can thank the sci-fi renaissance for where were at today. isnt this fun ?!

Lukeys sooo right...x will find us before we find it. and then who knows whats gonna happen ?

nah bro tygxc is downvoting all of your comments lol 💀

Yes the down votes don't even matter I only down vote when I see a bunch cause I like seeing numbers go up

It's funny how quickly you go from "Chess has branching factor of 35? Where did you get that from?" to "Yes, chess has a branching factor from 31 to 35, but [blah blah blah...]". You routinely get caught out in these types of mistakes, and then have to rapidly integrate them into your narrative.

Bringing up the pigeonhole principle as if it somehow cements your new rationalization is like adding "...and remember, ice is cold" to the end of an argument about glaciers melting...it only impresses the children in the audience that don't know any better and will happily accept anything written at a certain level of vocabulary.

You can either evaluate the pigeonhole boxes standalone, going backwards from mate, or going in a forward direction, you can evaluate them, but only with information in hand that you don't have until you can perfectly evaluate a single position (and that perfection going backwards from mate is why tablebases are created that way, such perfection of single position evaluations does not exist in the opening or middlegame). Whether AB or BA are actually identical in result is not something you can determine without calculation. If you are calculating to determine whether a transposition is actually the same in either order, then you are not actually working with 10^44, you are working with the larger game tree. Move order matters...ask any super GM that has forgotten their prep...

Here's where I think your fundamental mistakes lie...at each step of your reductions in magnitude, you double, triple, and possibly quadruple count the elimination of positions. 10^44 already is a reduction to *unique positions only*. Saying that you can take another square root out of that number (or worse, from your dubious 10^34 which is where the triple and quadruple counting occurs) because of transpositions is incorrect, because there are no transpositions in a set of unique positions...they were already eliminated to reach the set of unique positions. Your argument tries to apply it's largest reduction twice and then toss in some vague fuzzy reductions on top, all to get you to the 10^17 number you had in mind before you started. That's the opposite of the scientific method, cherry picking and fudging numbers/results to fit your hypothesis.

So sayeth the lady who still keeps trying to mine cryptocurrency as if it had any objective value.

Pretty sure there are several posters here who know a lot more about AI's potential than you do...myself being one of them.

the harm that those posts could cause pales compared to tygxc's and optimissed's misinformation

myself being one of them

trust me. ur dummer than AI. & u know it.

That statement doesn't follow since you are actually the one being compared to, not AI...but what it does imply is that I was right, because I'm not "dummer" enough to make the mistake you just did.

u just need opti back so u can pick on s/o. and ty for re: that crypto thingy from so long ago. wow lol ! why doncha go obsess on s/o else...plz ?