Some sort of unofficial ranking off the top of my head... most frequent reason for underpromotion...
1) As a joke because they know it will be captured anyway
2) To avoid stalemate
3) To taunt the opponent (or an extra queen isn't available)
4) A knight is needed to fork
5) A knight or bishop is needed to win that specific endgame situation (knights can't lose tempi but can jump over pieces, bishops are only on one color but can mate when paired with another minor piece, etc).
Some sort of unofficial ranking off the top of my head... most frequent reason for underpromotion...
1) As a joke because they know it will be captured anyway
2) To avoid stalemate
3) To taunt the opponent (or an extra queen isn't available)
4) A knight is needed to fork
5) A knight or bishop is needed to win that specific endgame situation (knights can't lose tempi but can jump over pieces, bishops are only on one color but can mate when paired with another minor piece, etc).
So interesting you say that about a knight forking, I came up with a really nice underpromotion puzzle myself: Here it is:
Oh yeah, those bishop underpromotions are pretty cool.
Sorry the original post here was deleted the link was broken, here is the bishop puzzle. Amazing, if they promoted to a queen or rook, the position would have been stalemate. If a knight, would have been impossible to win.
I don't know how people come up with stuff like that.
99% of the time, one will promote to a queen. However, some may promote to a rook to still have a powerful piece, and have a better time avoiding stalemate. ....
Never ever underpromote on the basis of some general consideration like you describe. It's chicken. It's like opening 1.e3 instead of 1.e4 because you are afraid someone will attack your pawn. You only underpromote after precise calculation. For the Rook and Bishop it always involves stalemate; knight underpromotions commonly involve forks.
Engines like Stockfish and AlphaZero do underpromote and have a special reason for it. They tempt you not to eliminate the promoted piece, giving them an easier win and thus increase their overall win expectancy. This only works against competing machines. Humans know why machines do this and waste no time on calculating alternatives (if IQ>80). Humans should not copy this engine behaviour as they are more prone to make reading errors. Humans do not play better than engines but they are wiser.
Really? No one has mentioned the romantic-style Lasker Trap? xD
From one of my games:
By far the two commonest reasons to underpromote are to A) Avoid stalemate by promoting to a rook instead on a queen. And B)Promote to a knight to queen with check and avoid mating. This is a very common theme in king and rook versus king and pawn end games
From the famous Saavedra Study
The only time I've ever had occasion to underpromote to a knight in a game wasn't to give check -- it was to *avoid* getting checked:
I was about to resign, when I noticed b8=N!
I wish I could report that I managed to hold the draw, but even with the underpromotion, I couldn't quite save it. Still, it bought me an extra ten or twenty moves.
An extreme case of underpromotion is shown here https://www.chess.com/forum/view/more-puzzles/fide-changed-the-rules.
I once read that the rules were updated after some master (Reti?) extracted himself from a sticky situation by promoting to a knight of the opposite colour. Could have been apocryphal and I can't find it any more.
I made this one. Simple but ok.
I made this one. Simple but ok.
There is more to this position than meets the eye. After (incorrect) Q promotion white can capture pawn e7 but still not break the stalemate threats. When need be, black can sacrifice his bishop on b2 or b4. But there is room for an interesting composition. First we'd have to get rid of the illegality though which is not so simple.
Avoiding forced stalement, usually immediately on promotion to a Queen, is the only stalemate-related reason for under-promotion. Having a 'better time avoiding stalemate' is a very bad reason. If stalemate is a possibility it usually means you would be winning easily with promotion to a Queen (assuming no instant stalemate) so it is also by far the easiest way to checkmate. And if you are that worried about future stalemates just always keep your opponent in check as many moves as possible and when you need to move your king or play a non-check move it doesn't take long to assure yourself it isn't stalemate. If you just end up with Q+K vs K and you are a beginner keep the Queen at a distance while forcing your opponent's king to the edge. Stalemate is pretty much impossible then.
Or to gain time on the clock due to a moment's hesitation by the opponent.
I have auto-Queen on so I'd lose time more than I gain it likely, though I guess that wouldn't matter if I have enough time and my opponent doesn't.
Here's a rather simple study that usually takes beginners many tries to solve.
White to play and win
After discovering that trading the rook for the a-pawn is only good for a draw, the solver will often try 1. Kc3 a1=Q+ 2. Kb3, expecting to mate next move with the rook, but then Black wins with Qa8. Next the solver will try 1. d5 to block the long diagonal. But then 1...exd5 2. Kc3 d4+ 3. Kb3 a1=N+ draws. So because of the underpromotion possibility, that try also fails to win.
99% of the time, one will promote to a queen. However, some may promote to a rook to still have a powerful piece, and have a better time avoiding stalemate. Some may promote to a knight in puzzles or if a REALLY specific move comes up where a knight and only a knight can checkmate or force checkmate on the next move. Some promote to a rook just to style on their opponent. This is an open forum where you can post situations/puzzles where underpromotion is a key aspect or needed, or just discussion on situations of that matter.