What if Mega Database, Lichess opening tree and eval don't agree?

What do you think is the best move to choose for my opening repertoire if Mega Database, Lichess opening database at my (limited) level and the engine eval don't agree on a certain move for a certain position?

So for instance:
- grandmasters in mega database play a certain move by far the most and with decent success
- lichess opening database suggests (filtering to my ELO rating) winning chances are clearly higher on a whole different move that GM's don't play
- the engine eval shows that the move it prefers has an eval that is +1 higher than what the GM's prefer, but it hardly is played by GM's. (to my surprise this happens a lot!)

I should add that my goal is to understand, learn and improve in the long term. Winning in the short term is not so important to me. And you could argue that for learning it can sometimes be better to choose a line with clearly higher winning chances from Lichess that you don't see any fault with, knowing the GM's play something else. Once you play it and it gets refuted a lot in your games you might prefer the GM line and understand it better.

The important thing in selecting a move, whether in the opening or another phase of the game, is that it is your selection. That means you understand what the move does and why you are playing it, both immediately and in the longer term. In the opening, this might mean balancing things like:

Does it lead to a more drawish or a more imbalanced game?

Do I understand how to play the type of position that will result? (based on pawn structure, open vs. closed position, etc.)

Is it a gambit line with theoretically equal compensation, or a more solid line?

Is it an attacking line that is theoretically somewhat worse, but very difficult for an opponent to meet in practical terms?

A lot also depends on the amount of theory in an opening. For example, some very tactical lines might have been analyzed out to a forced draw, so are now avoided by GMs. This happens a lot in the Sicilian and Spanish/Ruy Lopez complexes.

Databases and engines are tools and a single-score move evaluation should never determine your choice, IMO. They are most useful for doing your own research and analysis, to see if you can understand and would like to play the middlegame positions that result from a particular opening sequence. That said, if you see something strongly negative in the engine, or a poor winning percentage in the database, that probably is in fact something to be avoided.

Thanks for your helpful considerations. The points you describe are exactly what I am trying to take in consideration. I think I agree that my own level of understanding on the strengths and weaknesses of different moves are very/most important in deciding on a move. The tools can help in getting those insights.

Maybe use the move that is best as long as you still understand the plan behind it, so don't use something that you don't understand or that doesn't suite your playing style at all.

For learning it is better to choose a line that you UNDERSTAND.

I've played people who play a perfect sequence of book moves out to move 15 or 16, and then blunder away a piece on their first move out of book. That's what memorizing a sequence of moves that you don't actually understand does for you.

If you don't really understand the differences between the moves the eval doesn't matter. Play something good that you also understand. The best move will be the one you can follow up with  a plan.

Show some examples of this and then people will be more able to know what to suggest doing

Good points blueemu and nTzT. Thanks!

In reply to blueemu: To me, that kind of shows the limitations of the chessable courses as well. They offer good idea's and interesting insights, but some lines and idea's stick more with me and make more sense, where other lines often don't make sense to me, even after reading and studying all the related text, comments, and other related lines. Those courses are not end-all / blanket solutions that you can drill down with spaced repetition and you're all set. Instead you need to pick and choose I guess.

If you are really 1300 rated, you should spend at most 1% of your chess time with openings. At most. Maybe 0% is better.

You should concentrate on middlegames, especially tactics.

Openings aren't overly important because the evaluations run close together.

Like a "bad" opening might be +1 worse than a good opening.

Compare that to the middlegame, where a bad middlegame gives you a -10 position, or an endgame, where a single pawn move turns a forced win into a forced loss.

So my advice to you when it comes to openings is to just pick one which gives you a position you're comfortable playing. Don't worry so much about what is theoretically better, because any slight advantage you get from better opening moves is meaningless if you reach a middlegame position you don't understand and you're about to make a -5 blunder.

To cerebov: I'm sure you are right that this probably is not the most efficient way to improve, but I enjoy exploring this as a project for now.

I would say that to choose an opening repertoire for a 1300-1400 shouldn't be based on:

opening database (Because GMs don't play openings that help them improve they play openings that help them achieve a result against another GM)

engine eval (same reason as above also engines don't understand openings all that much)

no idea what lichess tree is but ok

I would recommend some instructor on youtube like Gothamchess who will lead you to the correct choice depending on your rating

Sorry but you're wrong and I explained why

I love chess.com and prefer it... but saying it has a better engine and explorer doesn't make any sense. Stockfish is free and other websites use it as well.

Making the same engine play itself on different platforms hardly means anything, it's simply using your hardware and whatever time you gave it.

While chess.com might have more players lichess has a massive amount of players and games that far exceed the necessary for whatever sample size you need. You shouldn't even be really using that. You should use master+ level games for serious opening study.

Nice

He doesn't get the concept of "explorer" those are only titled games not online games