I'm playing Black in the position below. Take a look and try and see why I think Black is better.
White has just moved his queen to e2 and is attacking the Black knight. In the game I played
25...Nf2+? 26.Rxf2 Qxf2, 27.Qxf2 Rd1+
I call this type of error "jumping in". You see a line with a forcing sequence from a strong position that appears to win and instead of making 100% sure that it does win you dive in and 'snatch' at the full point.
What is the answer? Cold hard analysis. Dan Heisman at the marvellous Chesscafe site talks about this type of error and says you should try and analyse the moves until the position is quiet, that is no more forcing moves http://www.chesscafe.com/text/heisman55.pdf.
If you think you are close to winning then its worth spending extra time checking your analysis. If you do look at a line and can't see all the possibilities then look for other simpler lines.
25...Nf6 takes the knight out of the sights of White's queen, but the immediate attack would be over, so this line would be the last to look at.
25...Rd1. The first point is that 26.Rxd1 Rxd1+ 27.Qxd1 allows Black the knight fork 27...Nf2+ winning the white queen
Connect the rooks, perhaps White could try 26.Bxf5, but 26...Rxa1 forces 27.Rxa1 and then ...Nf2+, 28.Kg1 Nh3+, 29.Kh1 Nxf4 - Note how this discovery trick crops up again and again. - wins a piece for White and stops any queen invasion on h5
Defend f2, 26.Nh3, this is probably White's best. 26...Rxf1+, 27.Qxf1 takes the threat from Black's knight and Black can introduce new back rank threats with 27...Qb5
- Take your time and analyse as deeply as possible if you think you may have a winning line
Try to find alternatives, especially if your first line involves giving up any material. The second line may give the same result without the material risk. Which is my second point.
- Given a choice, play the line with the least risk