What prompts a ''Good move'' alert in the puzzles?
lichess.org/training/24126I was doing this puzzle and on move five I did not see the mate in one and went with bishop f5. Which would have ended in a mate 2-3 moves later. I thought this was a little harsh since I spent about 20 minutes thinking it through and that made me wonder what exactly was a ''Good Move'' that was not good enough to be ''THE'' move, but good enough not to have you fail the puzzle. I thought this could have been one of them.
Not bitter here, just trying to figure out what constitutes these so called good moves.
#1 Sorry, I do not know.
Currently I'm testing Stockfish developer's "mate finder" changes, so far with negative results on difficult puzzles (including a mate in 5 puzzle with multiple quiet moves):
http://github.com/jhellis3/Stockfish/tree/mate_finderThe puzzle experiment (and I call it for what it is, no one's ever done this sort of thing before using a database and computer algorithm to generate tactics puzzles) are inherently flawed. Why? Because the bulk of the puzzles were generated prior to the engine upgrade and the algorithm wasn't fine tuned enough to be consistent.
This is how it works: if you make a move that is not in one of the engine's lines of calculation, you fail the puzzle. Plain and simple. If you make a move that was being considered but ultimately rejected, then you fail. If you make a move that is considered an inaccuracy or mistake, you fail (obviously). The thing is, any mates slower than the preferred mate will have been "refuted" or considered an inaccuracy, thereby you automatically fail if you don't get the best mate, no second chances, no "good move" etc.
Likewise, if you grab some material that still gives you a +10 and obviously a win, but there is a mate, this material grab will have been refuted as an inaccuracy or mistake and you will fail the puzzle, no second chances, no "good move" etc.
The zealous "lichess can do no wrong" defenders of these puzzles in their current state will point out obvious facts, such as a mate in 4 is objectively better than a mate in 5, or that a mate is objectively better than a piece grab. No one is arguing against this. We are not saying the puzzles should be "dumbed down" and grant success to those who make a suboptimal move.
What we ARE arguing is that you should get the "good move, but you can do better" dialog whenever the suboptimal move is absolutely still going to win the game (and no, I'm sorry, the clock argument here doesn't hold water because if it takes you 2 minutes to find the "best" move and you were able to see another "winning" move in 2 seconds, then you've already lost on time if time were a factor - these are not timed puzzles). If you miss a mate in 5 for a mate in 8, you should get a "good move, but you can do better" dialog rather than a fail. And in certain instances, if you miss a mate to grab a queen or the like you should get a "good move, but you can do better" dialog (i.e. in situations where by grabbing material puts you at a +16 or +20 sort of evaluation where a forced mate is inevitable over time, not simply a +10 or less situation).
In my opinion, and I've stated it several times only to be ignored, it's time to take all this feedback from the experiment, the new engine, and the even more expansive database of imported master games and games played by IMs, GMs, etc. and generate a tactics trainer 2.0. But for some reason the developers seem very attached to what they already created and don't want to bother improving it. Apparently the way the puzzles were generated doesn't even allow them to remove bad puzzles. Ultimately, it needs to be done. The plug needs pulled on puzzles 1.0 and a version 2.0 with better tuning of the algorithm, use of the improved engine, and the much stronger database that we have needs to be implemented. I'm not going to hold my breath though, as it seems that the puzzles were an afterthought and no one cares to develop the feature further.
#3 While a couple of your specifics slightly miss the mark, for puzzles rated over 1800 (or 1700?) ZugAddict and I agree with your analysis (minus your personalization remarks; I try to remain optimistic).
Well, to be fair, I have become pessimistic because I have consistently presented this argument ad nauseam for 2 months and not received a single reply from a single developer in the roughly 10+ threads I have presented it. Thus I can only believe that the people working hard to keep improving lichess just don't see the puzzle system as any sort of priority despite the daily posts about puzzle inaccuracies and the like.
I honestly think that this second chance system is way too nice.
In chesstempo for example, you never get a second chance, and nobody complains about that.
In my opinion, if you couldn't find the very best move, you failed the puzzle. If you think not, then you are settling for mediocrity.
The main point is not solving the puzzle, but rather pushing yourself beyond limits. If you went for a mate in 8 and there was a mate in 5, then you are doing something wrong.
Although I agree that puzzles don't seem to be a priority here.
I don't disagree with your sentiment #6. The problem is the inconsistency. If you are going to give a "good move, but you can do better" then it should be consistent or not at all. I also find it a very large design flaw that I've heard there is no way to remove bad / broken puzzles.
I truly appreciate how great this experiment was, and I think we've had a lot of feedback about it and there are some things that could easily be hammered out to make the puzzles truly innovative and consistently good, but it seems that is just not going to happen any time soon. So we muddle on I guess.
#5 looked at some of your suggestions, there are developers responding here and there I just think you have no clue who they are. So here is you official response from a developer on the interwebs: ok.
Hahah flugs, believe it or not I'm pretty aware, I've been around longer than this account, since probably the first year lichess was around. And none (until now I guess, since I do indeed recall you working on a few things) have responded directly to any of my other posts about this.
But "ok" isn't much of a response. Is there ever going to be something done to try to regenerate the puzzles with all the feedback, stronger database, better engine, etc. that have been acquired since this experiment first took place, or is it just to be left sitting in the dust as a "feature" on the site that is novelty and works "good enough"? If the answer is the latter, I think there should be some sort of note made so that as you get more and more new users here they aren't confused about the fact that many puzzles could inherently be flawed. It might cut down on the daily posts about bad or inaccurate puzzles.
Fenris1066, you haven't done many puzzles yet, and you're still only ~1500 rated with them so I'd like to think you're missing some of the subtle differences between certain moves. "Good move" is good luck, if you make a mistake, not the norm. Learn the optimal sequences.
