A Newbie drew Stockfish at highest level (2 draws so far)

Thank you for your post.

I believe that the possibility to beat a machine is quality over quantity. Machines strong point is more quantity over quality.

I still don't have enough quality knowledge. I mean... I know the general theory principles because I read the literature part in the books (skipping board demonstrations. Sorry, I still didn't have enough time for that) and understood it, but still didn't try all of those principles in a match. I tried many, like for example "the two weakness principle": to create them far away from each other, so that in the endgame a knight  can't defend both at the same time because is to slow, a bishop or rook would be better for that and you need good activation and mobility for threatening one and the other repeatedly. First you create the weaknesses if they don't exist, like pushing a passed pawn  or increasing preasure in a backward pawn...

But I still didn't try all theory principles. If I understand them is not enough, I need to see them in a match and test them. That's the reason at the same time I'm learning theory from books, this time looking at the diagrams, and analyzing the match to see the mistakes. But I still didn't have so much time for that. Only one or two hours per day, some days nothing.

I know I'm very slow, in blitz games my queen is taken by any pawn always because I didn't think enough time. But I think the magic of chess is in its strategy, and I think machines have good positional programmation, but maybe they are not so good at long-term strategies and creative plans, which requires a soul (I think that can't be programmed on a machine, maybe I'm wrong).

But even though I'm slow, I can think up to 20 moves ahead in middle game, and even opening if that would be useful in opening. Stockfish in my phone also thinks 20 moves deep. So, I think I can do it.

The problem is that from those 20 moves per variation, only 3 or at most 6 actually happens in a match (some times even only one move actually happens ). I believe that improving my strategy knowledge I would be able to see more moves that actually happens in a match (I think that's called accuracy, to improve my accuracy). 

A position is either a draw or a win for one of the sides, brute force is what showed us that information in our tablebases not any sort of "soul".

On another note, we have neural networks with a positional understanding that possibly superseeds us, just look at AlphaZero, it sacrificed pawns or the exchange for positional benefits that us mere mortals wouldn't have seen.

Those scrifices that AlphaZero can do are impresive. Can Stockfish do that too?

Anyways, that's more like positional play. What I was thinking about is very long term strategy. Something that can't be calculated with moves, because after a pawn sacrifice the machine calculates the position and one open file occupied by a rook is worth it. 

But what I mean with strategy is something more deep than positional play. They are plans that can only be thought in general, but not calculated move by move. Something like... I'll attack in the queenside because I can see most of the pieces are in the kingside and I think they will not be able to come to the the queenside to defend, because I have space advantage in the middle of the board and there is not a good flow of the pieces. The concept of space and restriction can't be calculated by a machine, I think... and I don't think it can feel danger and vulnerability... Things like that, things a human can see but not calculate, and a machine can't calculate too.

Those concepts are incorporated in the evaluation function of engines (at least a good one) and a sound sacrifice can always be calculated if enough time is provided, but then we ourselves can't be sure of the soundness of sacrificess unless we can calculate as well.

SF can also sacrifice material but it rarely does so for very long term compensation which is positional play.

So, do you mean engines can detect space, evaluate it and design plans accordingly? and the flow of the pieces...? That would be really impressive.

I don't know specifically how engines are programmed, I just guess what they can't do. 

hahah, to move them with the mind, that's funny   That's what chess players actually do, but they screw it all up by finally touching the pieces with their hands.

I think one must think tactical, positional and strategically at the same time. I think in chess all those concepts need each other.

They evaluate king safety and piece activity, they also design plans in the sense that they create a main line of play assuming their oponent plays like them. I'm sure there are more parameters involved in the eval function of engines but those two are crucial for positional play.

This is only true if you have a chance of winning - e.g. by playing opposition ~200 rating points above you own rating. I believe there are drawbacks to playing super strong engines - although these might be mitigated by playing against other opposition too.

After 13.e4 your position was practically lost. Engines can play these standard positions with one pawn up better than any human in the world given that the depth is high enough. Continuing positions like that is only useful for learning. Also you got to keep at least one pair of rooks on the board to maintain drawing chances.

1. You can check out Stockfish source code to get a feel for how it searches @ https://stockfishchess.org

2. You can also run its "eval" command to ask its position evaluation.

3. I assume you know about tablebases, which is simply a solution set for positions.

So, combined, stockfish has these 3: tactical, position, and solutions lookup.

That's right, it could have taken the pawn, hahahhah

@Prometheus_Fuschs:

That's interesting. What an engine can think about is more impresive than I thought... 

@mgx9600:

I don't think I'll have time to take a look at the source code. And I don't know if I would understand it. It's a long time I don't code anything.

It would be a good idea to look for a website that explains what stockfish or other engines are able to think about. I'll do that later. Maybe I could find something, some weak point, things they are not able to think about... and exploit that weak point directing my plans on the board following that path.

How can I run its "eval" command? I will search in arena to see if I can find a button like that. Or do I need other GUI? It is very interesting to know how they evaluate a position. I always wondered what value they assign to an outpost, or to an open line...

Just type "eval" at its command prompt after setting the position.

Here's some info:

https://hxim.github.io/Stockfish-Evaluation-Guide/

Play 10 sec bullet at home offline computer. You will never beat Stockfish. 

In chess.com online, there is 1 sec per move lag and SF will be lucky to play 10 moves without loss in flag. XD

I regularly beat Stockfish from set positions. This position comes from a game of a beginner who achieved a nice position and then claimed no ability to find a plan for Black. I played out the position as a way to offer help.

An extra pawn while being up the exchange? No wonder you won.

It wasn't easy to win this a blitz time controls. Took several tries before I succeeded.

The point of the post:

Playing Stockfish from the starting position, you're gonna lose. Playing Stockfish from a position where you have a theoretical win, but plenty of play remains, can be a useful challenge and you might succeed.

That's a good one. I don't know chess.com technical details, but I agree.

After some months not playing chess for real life hard work and university reasons, I came back to strike again!

This time I accepted an exchange in which I had an endgame with my rook against Bishop and knight. I know by theory books that in the endgame a Bishop and knight worth more than a rook because they coordinate and the rook is alone. But I never believed that book until I played this match. We newbies usually believe it's better a big piece than two little ones.

So, from now on, I learnt a lesson. Here is this second game:

Now I'm in the middle of the next match, the third match. This one is being played with the new version Stockfish 11 (50 elo points stronger than the previous version).

I must rush to win before they release the new version Stockfish 12, or this will get more and more difficult with each engine update.

I will catch you this time Stockfish!

(Next is the picture of the board in which I'm thinking. I'm stuck in that position for three days now, and still thinking variations)