The no solution, solution

[mindbreaker]

It is a little unnatural that we always have a win.  I think it would be cool if one out of 10 or so was just a position not a combo. You have to click "no solution" to get that puzzle correct or at least not mindlessly throw away material. 

This may not be for everyone, but I think it would be a nice mode to rein in those first move deducers and the itchy fingered folks.

Some players may think this is a solution looking for a problem, but in the real world you have to be committed to make these sorts of moves and that commitment should be justified.

These non-tactics positions can be rated just like any other.  The more advanced ones would have nice combos but have a slick defense that refutes the combo and leaves you lost.

[richard]

Hi,

Tomohawk had the same suggestion a few weeks ago.  I think it's a good idea as it adds a nice reality check that in real games tactical opportunities are usually just not there. I have this on the todo list but there are still quite a few other tasks above it.

Regards,
Richard.

[boromir]

on a related note, I think it would be great if there was a number of problems where the solution is much smaller material gain, e.g. a single pawn.
As it is now, it's sometimes a little bit too easy knowing that you have to find something that wins at least a minor piece or thereabout. Even a quite small number of these problems (say, 500) would have a big impact on the overall set, as you could no longer be sure that you need a big material gain to solve the problem.

[richard]

Hi Boromir,

I'd like to have more single pawn winning tactics, unfortunately the way the generator is currently written it is difficult to deal with single pawn tactics as I rely quite heavily on the engine evals to direct my search and it is hard to know for material that small if engine evals are due to positional or material concerns.  I do use straight material counts in some situations and I'm eventually going to rewrite the current generator in a manner which would allow a lot more experimentation and testing of other strategies, unfortunately this is probably even further away than I thought (too many other feature suggestions that are higher priority).  Hopefully I'll have time to revisit single pawn tactics in the future.

Regards,
Richard.

[tama]

I in fact think this is a poor idea for a couple of reasons, the first is this site is a (Tactics Site) if you get the drift. Second, winning pawns is very good in real games but wont help my tactical skills.

I think that adding problems were the user must fined the Best defense in a position to be better because most people have a harder time seeing there opponents tactical opportunity's and a easier time seeing there own.

But all the ideas mentioned are good if they were offered to Payed Members
(Un Rated).

[richard]

Hi tama,

I don't think winning a pawn is (always) equivalent to "non-tactical". The higher rated players can correct me if I'm wrong, but I think at the higher level it is quite common for a very nice tactic to end in the winning of a single pawn and for that won pawn to translate into a won game. Pawns are lot more valuable to titled players. If it was easy to do, I'd add pawn winning tactics right now but it will be a lot of work to modify the generator (or produce an alternative generator that can find these types of tactics).

I also think part of tactical vision is understanding when there is no tactic, so as long as there are not too many of these I think having "no tactic" problems could make tactics training here more effective (and might prevent people from making "guesses" without understanding the position).

Regards,
Richard.

[newlook]

I think it would be great to have positional and strategic problems to supplement the tactical ones. I imagine that overdosing on tactics exercises has an unbalancing effect when playing real games, which have other elements. However, finding or making good non-tactical puzzles is hard. You can know that doubled pawns are bad and a strong center good, but the payoffs are so uncertain compared to immediate material gain or checkmate. I am sure that the "find the best move" method would be great for improving positional and strategic play, but from what I hear, computer analysis is not yet up to par in those areas.

[richard]

Hi newlook,

I have longer term plans to provide more strategic problems. I agree that computers are not really up to this yet, so my current plan is to provide a way of users generating strategic/positional problems and allow voting to choose the high quality ones.  These would probably be presented separately from the tactical problems.

Regards,
Richard.

[d-man]

awesome idea. one of the things that really blew me away on this site was that i was so used to doing tactical problems i'd spend 3 minutes mesmerized in a host of variations attacking the king only to find the problem was simply taking a hanging piece. it prioritizes my mental search pattern and exposes the blind spots.

maybe have problems resulting in getting a draw from a losing position... those are helpful. and i've had quite a few in real games.

[boromir]

I in fact think this is a poor idea for a couple of reasons, the first is this site is a (Tactics Site) if you get the drift. Second, winning pawns is very good in real games but wont help my tactical skills.

How exactly would you win a pawn without using tactics (apart from straight-up blunders or sacrifices from your opponent)? pawn-winning tactics are probably the most common of all, at least when you move past the complete-beginner stage, and they tend to be way under-represented. I would say that including these kinds of problems is vital to the quality of this site (not that isn't really good as it is  ), but that's of course just my opinion.

Part of developing your tactical skills is also improving your ability to decide when there is or isn't any tactics available or when the best you can get is a single pawn or a miracle draw, and I think that having a greater variety of tactical positions, including no-solution and drawing problems, would help improving that.

The problem, from the problem-generating point of view,  is of course ambiguating between an actual pawn won, and one that wasn't worth it. Did you really just win a pawn with a clever sacrifice or did your opponent give it to you in exchange for a powerful kingside attack, did you completely mess up your own pieces to win that pawn, etc. Then again, Richard seems to be reasonably good at figuring out that stuff...

[richard]

d-man: I'd like to add draws, at the moment draws get interpreted by the generator as completely even positions which combined with a "tactics must win at least something" rule means there are almost no (possibly none) drawn positions in the set.  This is one of the things I'd like to fix in future generators, it doesn't lead to bad problems getting in but it does keep some interesting problems out.

boromir: The "have I just won a pawn or have I just exchanged a positional advantage for a pawn" is quite a tricky problem. A simple approach is to simply ignore positional issues and take the attitude that a pawn-is-a-pawn.  This is somewhat dissatisfying (and still has technical issues) but might be ok as long as all the "tactics" positions are looked at with the mindset that you are only looking for tactics and positional issues are irrelevant.  Currently positional elements do have some influence on the tactics.  Often the choice between best and second best move (when the second best is an allowed alternative) is down to positional elements the engine has seen, although the new generator demands a more clear cut distinction between best and second best , so for better or worse this positional subtlety will be lost (most users didn't seem to like it , especially when different engines or the same engine run for a longer time had a different assessment of the positional elements).

Further generator changes (beyond the latest change which will hopefully see the current problem set updated in the next 2-3 weeks) are at the moment only medium term priorities so it will still be a while before I can investigate more of these ideas.

Regards,
Richard.

[cyanfish]

For the winning-a-pawn problems, I think that perhaps the generator should look for positions in which a) obviously, a pawn is won, and b) the eval is above a certain threshold ( e.g. 1.1, 1.2? or maybe as low as 0.8 ). This would stop problems in which the materially down side has some compensation.

Of course, this is further complicated by the fact that the engine may be looking at a line that involves giving the pawn back ( which would otherwise be a weaker position if it kept the pawn ).

[richard]

Hi cyanfish,

That sounds like a workable idea.  The generator currently uses the hybrid approach of examining the engine evaluation as well as my own material evaluations, but I haven't tried applying that idea to pawn takes yet.  I'm not that keen to put further development effort into the current generator as it is becoming increasingly difficult to add and test new features.  I'm hoping to find time for a complete generator rewrite in the future which will keep all the same heuristics of the current generator, but do so in a much more flexible way. This will hopefully allow me to more easily experiment with new ideas. When I first wrote the generator I thought it would be about a weeks work, but I completely misread the difficulty of some parts of the problem (I incorrectly figured that the engine would be doing all the work for me and post-processing the engine results would be trivial - the thresholds for good moves etc are not that tricky, but things like pruning a move sequence at the right spot was hard [and still not completely satisfactory, although I still think there are more moves that are required to be played out here compared to CTS]).

Regards,
Richard.

[slacker00]

boromir: The "have I just won a pawn or have I just exchanged a positional advantage for a pawn" is quite a tricky problem. A simple approach is to simply ignore positional issues and take the attitude that a pawn-is-a-pawn.  This is somewhat dissatisfying (and still has technical issues) but might be ok as long as all the "tactics" positions are looked at with the mindset that you are only looking for tactics and positional issues are irrelevant.  Currently positional elements do have some influence on the tactics.  Often the choice between best and second best move (when the second best is an allowed alternative) is down to positional elements the engine has seen, although the new generator demands a more clear cut distinction between best and second best , so for better or worse this positional subtlety will be lost (most users didn't seem to like it , especially when different engines or the same engine run for a longer time had a different assessment of the positional elements).

This was also my first thought.  I'm a King's Gambit player as white and I try to carry that attitude into my other openings, middlegames and endgames as well.  I regularly try to sacrifice a pawn to get other positional or strategic advantages, especially if I suspect my opponent is a "pawn pig".  I think that's where chess really becomes interesting, when there is a material inequality, yet positionally equivalent.  If winning-a-pawn problems are introduced, I'm not sure how the Pandora's box about positional compensation can stay closed.  Of course, the answer is to also introduce problems which focus on tactical motifs purely for positional improvements, which is way off in the horizon at the moment.  The current state of the art seems to be set at winning a piece, which is much tougher to justify in terms of positional compensation.  (But I have done it in real games!)