Training, more tolerance with movements

I think that there is a middle line here, and I'm not usually one for middle lines!

First, I think that to call the puzzles "training" is a misnomer. There is no "training" going on at all - so they should be called "puzzles".

Second, while the puzzles obviously should require one to solve mates without being waylaid by hanging queens etc., I think that the effect on individuals' "puzzle rating" is quite strange and bears only a slight relation to the difficulty I have found in solving them. This could definitely be improved.

Third, the popularity system is all well and good and may give a rough indication of how good a puzzle is, but I imagine there are many people who just move on without "marking" a puzzle, especially those who have lost.

Compared to the gameplay and overall features of lichess, the puzzles (aka "training") are definitely a weakness rather than a strength - in my humble opinion as a user.

Thanks, @Fenris1066. (They're "dubious" to me now. Maybe in ten years, they won't be.) I'll try your suggestions on my next [s]masochisting[/s] puzzle follow through session. ^,^ (Yes, there's no bbcode but the intention should be obvious.) As for 8592, if you play the starting position against AI8, it will use the bishop to eat an edge pawn (can't recall if it was A or H, sorry) and check the king and secure the queen afterwards. Like you said, it is a hobby--just a hobby for me. I know my limitations and even if I surpass those, I very likely will never even reach near the level of a master. Oh, well. Anyhow, I enjoy playing (and learning about) chess (albeit quite slowly since I'm a slow learner). That's all there is to it. :3

Before those material puzzles finally got mixed in (which was the original design intention of the dev/s), my training rating hovered at around 2300. It tends to go below 2000 nowadays. =,=

@JustTraining, that reminds me of a turn-based strategy game I played where I was flashed with a snide ending prompt: "You have eliminated all your enemies. Now, you can take your time to look for the treasure. Mission failed." In the older chess book puzzles, I remember being required to provide all the branches (i.e. to write them down somewhere). In lichess, we only need to follow/solve the line the AI puzzle(r) presents us with even when there might be two or more "best solutions" with the same number of steps. It is a puzzle convention that we take the shortest path to the win. Thus, although branches with more moves can win a game, those are not considered solutions to a chess puzzle. (Hmm, I wonder if we can just be awarded a draw instead of a loss for those? Nah, might mess the scheme too much. The ratings are there to help us gauge our current strength. If they're not properly indicative, then they become nothing more than mere fluctuating numbers.)

@Unihedron, Clarkey's reuse of resources to make those was a grand novelty.

@Toutatis, after retrying retrying a failed puzzle, I upvote if the solution gives me a "Wow, that was cool ... why didn't I think of that?" feeling. On the other hand, I might downvote a (non-checkmating) puzzle I solved if the ending position seems awkward. You're right though that I usually might not bother voting regardless. ~,~'

Some very good points brought up in these last few posts! And I agree that there could be ways to improve the puzzles...

The problem with the rating system is that people take ratings too seriously and all the blitzers want to race the puzzles. Also the puzzle ratings themselves...I'm not entirely sure how this is determined, but I have found some 1600 rated puzzles more straightforward and easier to solve than some 1300 ones, so I imagine this trend continues up the ranks (in fact I was able to solve a 2000+ rated puzzle the other day and yet I lost to a few 1200 ones!)

The problem with the upvoting system is that players downvote puzzles just because they didn't beat them or they *think* their solution was better (even though it wasn't) and so forth, so good puzzles get buried for subpar ones that are easier for people to beat, even if the puzzle itself isn't all that spectacular.

Ultimately, there should be some sort of vetting process that makes more sense than a popularity contest...for instance a team of higher rated players that go through and rate the puzzles, evaluate the rating associated with them, etc.

Also, I think there should be, as mentioned before, a time-attack version of the puzzle training that is slightly different. Then the blitz and bullet players can race the clock to find the best solution rather than get frustrated that they couldn't get the absolutely best move in a sequence meant to be pondered for a long time and every line analyzed.

I find again one example:
http://es.lichess.org/training/22378

The correct solution is:
Ke3 Kf1
Qd1#

That's nice, check mate in 2. However, when I wanted to solve this training puzzle, I saw this check-mate:
Ke3 Kf1
Qe2+ Kg1
Kf3 Kh2
Qxg2#
http://es.lichess.org/6kbUWZ4fioH6

Yes, its not the best checkmate sequence, but c'mon, it counted me bad and took away 20 points whereas I made forced checkmate.

Again, I would agree than computer says (as he says sometimes): "Not the best solution"; instead of just making me: "wrong movement"

@JustTraining, yeah, harsh. The current setup is that it's only lenient up to one extra move ... and you have two.

Hmm, an idea arose from seeing your most recent frustration. You probably know how the opening trainer asks for multiple "good moves" from a certain position. Imagine a (more comprehensive) puzzler which asks us for multiple solutions, not just the best solution, and we get credited/penalized for each line separately. Of course, the best solution will still get the most points/penalties.

Again, I found another one:
http://es.lichess.org/training/20794

Correct solution:
http://es.lichess.org/LWILXfHmjFul

The solution I found which was checkmate anyway:
http://es.lichess.org/kKnpS6RB

Actualy, I understand than they want we found the best way to checkmate, but again, I think than the computer should be more tolerant; if we are talking about:
1) "Or taking pieces or checkmate, and you decide to take pieces"
I totally agree than is a bad movement, because you decided to take pieces except of win, and this may be losing at the end for you if you make one mistake
2) Checkmate in 2 instead of checkmate in 5; That should be totally acceptable, while you didnt lose the checkmate sequence, I do not agree than computer find you are doing wrong movements.

Training it's really cool; and I'm really glad by the "automatic" system which makes puzles from custom parties; but, would be a problem if we add the condition of: "if checkmate is still available: it says: {best move but not the best}".

I think that would be great, and everyone will find it interesting and profitable, what do you think guys?

I really think the reason it does this is because when the puzzles were generated it was before the engine upgrade and when you played a mate that wasn't the best mate, it marked it as an inaccuracy even if it was still a forced mate line (which I seem to remember this being adjusted by thibault so that it no longer does this).

IMO, as obnoxious as it might be, I think someone needs to go through and re-generate all the puzzles with the new engine tolerance so that these alternate mates will not fail the puzzles outright, only missed mates. It would likely result in far less topics being posted about "bad puzzles."

It's so frustrating... again happens the same, and I lost rating. (I know the training scores are not relevant, but I like to have them because it shows my progress, and this stuff make me mad)

Puzzle:
http://es.lichess.org/training/17819

Solution:
http://es.lichess.org/H0mvBt4Xz69Q

The checkmate I did because I had in mind (which is only 1 step more) and it makes me loss points:
http://es.lichess.org/SvczHu0RcMJY

Here is the same problem than commented above; okey right, I didn't see the chekmate Qh7; but I see the checkmate Nf4 (which was more natural because it was my first instinct, to attack the Queen, and it just allows the king to go g4 to can mate him later.

So frustrating than it takes away my ranking whereas I had a checkmate solution. It's really unfair because when you see a forced checkmate sequence, you automatically think you are right and stop looking for other checkmates, because you know you'll win anyway.

@JustTraining, if you're too conscious of the training rating, then you will keep getting frustrated. ~,~' That number is just a rough (usually rather wavering) measure of your current strength in tactics. In my firm opinion, what's more important is what you learn from playing these puzzles: mating patterns, tempo, pins, sacrifices, etc. :3