A better Tactics Trainer

As everybody knows, one can be great at solving Tactics Trainer puzzles, and still make oversights during games. Part of the problem is that in tactics puzzles, one knows that there is a tactic. Another is that people use their tactics skill wrongly in games.

One improvement that has been suggested a lot is a percentage of problems where there is no working tactic, and a "No Tactic" button to press.

In his latest "Novice Nook" column, Dan Heisman says that "The primary use of learning these basic tactics is to reject your candidate moves that are not safe."

That suggests another idea: show a position and a candidate move, let the user say whether the move is safe (e.g., show a diagram, ask "Is 17.Rad1 safe?"). If the user says yes, let him play it and make a few moves (possibly showing the hole in some seeming threat), if he says no, let him play the other side of the board for a few moves to show why the move is wrong. You get points if you got the right answer and the right moves.

One big problem is the difficulty of generating enough interesting positions where there is no working tactics; I don't see how it can be done automatically, and that is needed to get the thousands of problems we'd like to have (ideally there'd be a 50/50 split between safe and unsafe moves).

Comments? Ideas for the last problem?

That suggests another idea: show a position and a candidate move, let the user say whether the move is safe (e.g., show a diagram, ask "Is 17.Rad1 safe?"). If the user says yes, let him play it and make a few moves (possibly showing the hole in some seeming threat), if he says no, let him play the other side of the board for a few moves to show why the move is wrong. You get points if you got the right answer and the right moves.

That is training correct chess thinking! It is really a great idea IMO. The downside, and I think it is considerable, is emotional/psychological. Ordinary chess problems have a built in reward stimulus, the pleasure of finding a clever set of moves and then mating or winning material. Your puzzles put us psychologically in the position of: figure out why you +can't+ play a move. Not so sexy for the casual player.  But as a serious training discipline, IMHO, it's right on the money. I'm a chess-addicted casual player and I personally would love to train that way.

I like this idea too.  Of course the obvious problem is finding the problems for them.  I think the second part is probably eaiser to accomplish, there are already books or at least problems that ask you to make the wrong move, (The Ultimate Chess Puzzle Book, by John Emms comes to mind)  so I think the "Is this move safe" problems are at least attainable.

Good idea by the way!

I have a practical suggestion:

We can all collect such problems from our own games.  That's a source of inspiration for everyone, and can be just as spectacular.  I know because I'm having that experience myslef: I've been collecting chess tactics problems from my own games over the past few months and by now I have over 100 of them.

It'll not be hard to augment the collection with problems of the new variety that Remko suggests here.  I'll remember to do that from now on.  Then, if we figure out a productive way to gather all that we've collected, I'd be glad to contribute to the common pool.

I like Remco's idea too. Maybe people that made the Deep Rybka3/Aquarium softwares are already working on it!

I'd say that the problems I get right knowing there's a tactic are counterbalanced by the problems I get wrong thinking that there must be a more complex answer than what I've seen- situations I would probably have no trouble with in a game...

I love this idea and it's really what all of us need. To know when to look for a tactic and also learn to refute candidate moves, especially sacrifices because in real games there mostly is some way to defend against an attack. The problem is that you need to be able to see you opponents defense 3 moves deep and then refute your queen/rook sac. In the tactics trainer it is too obvious what the tactic might be (usually some odd sac) and most fails are because of wrong move orders and alike or being too greedy and grabbing a queen instead of the mate.

I agree with this. I did several 2600+ problems today where the solution was what we'd all play in a game, but they're rated so highly because there's a plausible looking tactic. I passed a couple and failed one that I can remember.

So in a way Scarblac's idea does exist on TT already... maybe such problems are hidden towards the top end.

http://www.chess.com/tactics/server.html?id=35453

Answer is simply Rdxe6, nothing more, nothing less.

http://www.chess.com/tactics/server.html?id=29175

Answer is Qxf7, not the plausible Rh8+

http://www.chess.com/tactics/server.html?id=30956

A real prime example - the decoy is 1.Qxh7+, the proper move is simply 1.Rxb2.

Oh one more thing. I don't think it is that hard to generate positions that are rich but are void of tactics. Just let the search engine evaluate positions and if there is no tactical blow that nets more then half a pawn there is no tactic.