Also forgot to say that you must not give only the specific position in exercise, but all moves from the begining.
If you want, I can created a pgn file with the opening book. I already have a database from twic with 3M games. But it's better to create your own database with a script as reference for the best game.
I have been using the "Openings Training" feature for quite a while now. I have attempted over 100 puzzles, and I would like to present some of my feedback.
TL;DR: The Opening Trainer is practically useless (no offense)!
1. I have not seen any improvement in my opening play.
2. Half the time, the candidate moves you have to find in the position are very natural developing moves that even a 6 year old is able to find.
3. Often a capture (like bishop for knight or pawn for pawn) is either a good move or one of the candidate moves, so you can never go wrong with it unless there is a clear tactical refutation. Captures are always the first thing I try and usually ALWAYS work. This is not good because many of these captures are NOT theoretically good in that particular opening, however due to the way this system is setup, even if the move was suboptimal to about -30cp of the best move, it would still work. All these suboptimal moves that the "Opening Trainer" teaches you to play worsen your opening play instead of improving it.
4. Positions are chosen at random, and there is no way to know what opening that position came from, and in what move order. This is not how opening training should work. Opening training is all about understanding moves, move orders, and then memorising them and creating a repertoire based around your strengths and weaknesses. The Opening Trainer at lichess achieves NONE of these goals. After doing 100 opening trainer puzzles, I remember NONE of those random positions, I have gained ZERO knowledge of opening theory, I have gained ZERO understanding of those positions or any opening, and if any of those positions happened in one of my future games, I would not be any better off if I had not done the opening training.
5. As already described by others, if you make a mistake, you don't know why it is a mistake. Providing the PV in future will help a bit, but still not enough because many "mistakes" in the opening are not tactical, but rather positional inaccuracies, and simply displaying the PV does not teach anyone anything. Anyway this last point is the least significant. See above 4 points.
@run:
I agree that the lack of directed learning does limit the training aspect currently. But they will implement opening specific training at some stage which should help more focused line training. Clarkey also mentioned ideas about playing on from one of the candidate moves to allow for a better understanding/learning of a variation.
But on 2. and 3. - these ARE grandmaster played moves (from TWIC), not just engine tested moves. If you (or a 6 year old) are finding these moves, then GREAT. Often we over-complicate chess and imagine GMs play amazing moves - when they are really just mostly playing "natural development moves". If captures work, GMs have tested these captures in those openings to "show" they work.
#22
Lots of stuff still to be added, including what you've suggested.
1. Good for you.
2. Good for you.
3. First half of the paragraph => good for you. Second half => The scoring system is still being ironed out.
4. Still to get added in the future. There's only a finite amount of positions that we've analysed from TWIC games so adding a filter won't be done until it's useful.
5. Still to get added in the future.
This feature is still ultra beta.
@Clarkey I suggest you label it as "alpha" or "pre-beta" so that people know what to expect.
Cheers :)
Thanks!
I liked it a lot, exactly the way that it is now. Many times I will be forced to think about what can I play in a position that I try to avoid at all costs (coming from an opening that I hate in the first place), and I will learn "5 playable moves" in that position, because SF says so and that`s good enough for me.
Suggestion: Let us see the centipawn evaluations of each one of the best moves (no need to show the evaluations for our wrong moves, even because that would take some SF processing). After all, sometimes there`s no opening move that is capable of giving an advantage, just the smallest disadvantage as possible in terms of centipawns. So, it would be interesting to know if SF considers the position + or - for our side.
#15
You can safely assume that if something doesn't seem complete on lichess, it's not. We've never released a product that is 100% complete, never to be updated on the first release. It's an on-going process and you get to use the features as soon as they have minimum functionality.
I love the idea of this feature but being unable to chose openings that I play makes it pretty much worthless.
I don't play the Spanish as black OR white, so what do I gain by learning opening theory in an opening I don't play? Sure I might gain some slightly better understanding of positional play in drier positions, but that's really not what I'm looking for in something that's supposed to be training my opening skills.
I would much prefer if I could just pick and ECO code and go to work from there.
This would be awesome for two reasons.
1. I would get to study openings I play
2. I would get to learn the ECO codes for openings I play (kind of worthless but it looks impressive to your friends.)
Great feature but a little pointless as it is currently implemented.
Atlec, our intention is to add that functionality when we have enough puzzles to make it useful (in about 2 months). If we added that right now, a significant amount of openings would only have 1 or 2 puzzles; many with none at all.
I don't mean to hijack the thread, but I have a couple of comments/questions after trying out 20 or so opening puzzles.
* It's not really obvious (at least not to me), that two of the best moves could be with the same piece. For example in
http://en.lichess.org/training/opening/21, peasant to d1 and d2 seems to be two of the best moves.
* Some of the openings seems to be taken from the same games, just one or two moves before/later. For example in
http://en.lichess.org/training/opening/22 one of the best moves is to castle. While in this opening,
http://en.lichess.org/training/opening/15 you have the same opening, just that the castle is already done. Is this really helpful? Maybe it is. I guess that opening 22 could just have had 6 good moves instead, and the second puzzle would not be needed. Or is my logic flawed?
* Which brings me to the next question. Is it really always 5 good moves worth mentioning in these positions? I ask because I just started with chess. But for me in many of the puzzles, it would be more natural to develop other pieces before I do all the 5 moves I suggest, but they are closed in and not movable in the puzzle.
+1 (on giving out points even if you only manage 3-4 out of 5 moves. I usually fail on the 5th)
- ØØ -