Chess will never be solved, here's why

Carlsen is reported to have almost perfect chess memory. He has perfect memory of at least 10,000 games.

@#40
<<<Assume 1 e4, 1 d4, 1 Nf3, 1 c4, 1 e3, 1 c3, 1 d3 were all known to be draw. White players might  avoid the former ones where the opponent knows how to draw in favor of the latter ones, where the opponent might not know how to draw.>>>

Ah there's an error there. Assume we already know that they're drawn. Players will still prefer the openings that give more chances of wins. The mistake is in the assumption that there's only one drawing line.

#42
A player may have several perfect games memorised where the Berlin draws, but he probably has no games memorised that draw against 1 c3. Objectively both are a draw, but subjectively 1 c3 may give more chances to win, i.e. more chances of the opponent to err, because it is less common / seems less logical and thus may be unprepared for.

#41
Suppose Carlsen has 10000 flawed games memorised and XYZ has 10000 perfect games memorised.

🍿

Yes it is incredible that they can fully analyze a 16 man position (such as the one I took from you ICCF tournament) down to 7 men in 3 days when it took 3 months of intensive computing to extend the 6 man Syzygy tablebases to 7 men, especially when you consider that the effort rises exponentially with each extra man.

Nice to know the TCEC organizers get their info direct from God.

They could have bypassed the channel by taking the approach that the results at the current level of computer play would likely be all draws for the reasons I enumerated, or simply projecting from previous engine v. engine results.

Then, of course, there is something wrong.

And not such an unlikely event.

It may seem that way to a mentally unprepared opponent but all black does is sift through his memory of 10,000 games, split into variations, and decide which best leads to equality. For instance, playing an immediate 1. ...g6 would seem like a very bad idea but 1. ...e5 should equalise very comfortably, against 1. c3.

#46
ICCF grandmasters and their engines have not been looking at the positions for their 3 days per move only: they have been studying the position for months. As one ICCF world champion put it: in ICCF the player who goes to bed later wins. In the opening they already analyse endgames.
Besides from 6 to 7 men table bases required to look at all 7 men positions, analysis of an ICCF game only requires to look at the relevant positions.
#47
At the beginning TCEC was played without enforced openings, but as they more and more approached all draws TCEC took action and started to impose slightly unbalanced openings.

#46

Still quite incredible. 

If they studied the 7 man position I posted in #27 without access to a tablebase the same would apply. 

I would be pretty confident that none would be able to achieve the win against random best moves from Syzygy after a year's study.

With 16 men the fact that they can't see a win means nothing. The win in this case could be thousands of times as deep as that 7 man position.

Correction: Since we're talking about ICCF rules (which include the 50 move rule, albeit in a quirky way) "thousands of times as deep" should probably read "several times as deep".

#47

Yes, TCEC did bypass the channel and base it on statistical results.

But I still think the reason for the statistical results is that intensive opening analysis has removed from play enough of the wins that the computers are up to and the computers are left out of their depth. Positions with equally matched opponents that are out of their depth seem to generally produce draws whatever the theoretical  assessment of the position is in positions where the theoretical assessment can be known from tablebases (starting with 5 men for engines and humans) . I would expect the same to hold with more men on the board.

SF (8-14) will mostly draw winning positions like the first one in #27 against itself. It has no "visual" assessment but it's initial evaluation correctly puts Black in front. Human "visual assessment" has always (up to the production of the relevant EGTB and largely beyond) been "draw".

What would your visual assessment of the second position be (if you didn't know the answer)? 

I think mine, after some thought, would have been even.

But "visual assessments" apply to chess at the level you play not theoretical chess. If you visually assess the first position in #27 as drawn you'd generally be right on that basis. If you arrive at the correct theoretical assessment you'd generally be wrong.

     Using today's technology to solve chess seems like a far-fetched proposition. Assuming that no one will EVER think up a better way to go about a proof or that computer technology will NEVER improve enough to make such an endeavor possible seems far too pessimistic. Just 100 years ago--a drop in the bucket in the history of human knowledge--things such as atomic energy, rockets to the moon, even computers themselves were generally regarded as beyond human achievement. Go back a mere two centuries from there and self-powered machinery, flight, treatments for most diseases, harnessing electrical energy, virtually instant long-distance communication, and much of the rest of today's commonplace everyday technology would have been regarded as impossible.

     Sooner or later--maybe in a century, maybe not for milllennia should climate change lead to the collapse of our present civilization--chess will be solved. That this will lead to loss of interest in the game is unlikely.

 Sooner or later--maybe in a century, maybe not for milllennia should climate change lead to the collapse of our present civilization--chess will be solved. That this will lead to loss of interest in the game is unlikely.>>

Why should that be? Don't people have better things to do?

I don't know the answers and I haven't seen the positions. Where are they?

Oh got them. I would assume the first one is probably drawn because I think white can build a fortress.

White must approach the pair of knights with the king but I'm not competent enough to see that there is or is not a way for black to win a knight, just by a visual assessment, so I think it's "probably" drawn.

The second one is drawn. Or looks drawn. The fact that it's a win in 549 doesn't imply that the game of chess may be a forced win, because the position is both statically and dynamically extremely unbalanced.

     We all have better things to do right now than play the game and write about it on an internet forum, yet here we are.

I think most competent players would agree with your assessments and be wrong theoretically in both cases (under basic rules of course). 

The second position would need some analysis before any assessment, which is probably what you're saying in your last sentence. Mainly can White turn his pawn into a queen without any disasters? Your assessment would probably change with e.g. the knight shifted to b7.

I've actually done a little practice on positions like the first. Not a lot but enough to render me incapable of making any assessment at all. (If the pawn were one square further back I'd tentatively go for "draw", but I'd still be very confident of losing to the tablebase as White.)

So here we have two relatively simple positions which players would generally regard as drawn if it weren't for the fact that we have tablebases, but are actually wins. The problem is that players are already out of their depth (as in depth to mate). If I gave SF14 any mate in 30  instead of 50 in the KNNKP ending I think it would win against Syzygy with the same time control.

In practical games the results would probably correspond with the incorrect assessments.

If Haworth's law extends to 32 men then the longest mates with 32 men on the board would be around 30 trillion moves deep and the longest mates with 16 men on the board (as in the random ICCF tournament game I posted in #22) would be a little over a million moves deep.  Any chess player, including the illustrious list posted by @tygxc in #33, would be completely out of their depth in these against a hypothetical complete set of tablebases. Their assessment if they gave one would be "draw".

At the same time the percentage of draws shown by the Nalimov tablebases which starts off at 100% with 2 men decreases with each extra man and looks set to practically disappear well before the 16 man stage. (Basic rules only of course, the draws are replenished under competition rules by "frustrated wins".)

So a common sense view would be that almost all positions near the start of the game are won under basic rules  (for one or other side) but many would appear to be drawn to beginners and grandmasters alike.

Of course extrapolation is a dodgy process and a maximum length mate is not a usual mate, but my guess is that whenever an opening  line is analyzed to "draw" it should almost always say rather, "out of my depth".  (But I'm not going to offer to win it.)