It's still great. Maybe even more fascinating that this was studied so long ago.

By the way, didn't you have a game where you actually had to promote to a bishop to win?

Idk... Probably? I have had some stalemates (because of queen promote), and those could have been fixed with a promotion to bishop.

In the critical line, yes. Unfortunately, my opponent chose to play a different move and lose in pedestrian fashion, but in the defensive line that was hardest to crack an underpromotion would have been required in order to win.

Here's an underpromotion puzzle to help understand further

That's the one. Thanks.

OP, there you go, an example where it is necessary to promote to a bishop.

It's a nice puzzle sadly the solution (mate in 12) has been refuted by Nalimov. With best play Black can hold out for 26 moves and White must promote to Queen to achieve the win.

Interesting, of course if black plays that, white will not promote to a rook. That study is still correct in the case black tries to trick white with Rd4.

Nope, this very famous study certainly hasn't been "refuted" by Nalimov. On the contrary, tablebases confirm that the study is sound by showing that White wins no matter what Black does. 5...Rf3 allows White to promote to a Q, reaching a Q vs R position that's already known to theory as won for White. As such, it's an uninteresting side variation, regardless of how long it takes White to give mate. 5...Rd4 is the interesting main variation, because it eventually forces White to underpromote as the sole way to win.

In any endgame study, Black will have alternative moves at every stage of the solution, but they just create normal side variations. What's important is that White doesn't have alternative moves in the main line. When White has such an alternative, it's called a "dual", and it can be a serious or a minor flaw. This study actually has a dual after 1.c7 Rd6+ 2.Kb5 Rd5+ 3.Kb4 Rd4+, when 3.Kc3 works as well as the intention, 3.Kb3. But it's not a serious flaw because 3.Kc3 Rd1 forces 4.Kc2 anyway, transposing back to the main line that requires White to underpromote. 

The only purpose of promoting to a rook or bishop is to prevent stalemate.-

It's a nice study, I enjoyed it and I would agree with you. However you're taking my comments out of context and that is that the study has been "refuted by Nalimov. With best play...". The refutation is that with best play White can only win by promoting to a Queen and since the theme of the study is under promotion it's been refuted.

A Mate in 4 study wouldn't be a Mate in 4 study if best play resulted in a mate in 5! Same here the theme of the study, under promotion, relies on an inaccuracy.

Not all the time. See this puzzle for example:

Underpromotion sometimes is used to fork promoting to a knight.

Your post shows a common misunderstanding of two types of composed problems: endgame studies and mate-in-n problems. An endgame study involves the task of "White to play and win (or draw)" with no limit on how long it takes to actually give mate. A mate-in-n problem (see my blog for an introduction to this type, called directmates) does require White to mate by the specified number of moves. The position we're discussing is an endgame study composed by Barbier and Saavedra, not a directmate. It's true that Black's "best play" in a directmate necessarily involves moves that delay White's mate the longest, but that's simply not the case in studies. If you look at this study closely, you'll notice that not only 5...Rf3 but countless other black moves also require White to promote to a queen in order to win. But the intended solution is not refuted at all as long as at least one Black move forces White to promote to a rook.

I only said rook or bishop, I did not include the knight.

Oh sorry didn't see that

Except the study is presented as an under promotion when clearly under promotion is not required and there are many solutions to the preferred solution. We'll have to agree to disagree on this one.

Yes

It fine.

Well, clearly the underpromotion is required in the main variation, as explained. 5...Rd4! and other black moves in the main line are called thematic defences, as opposed to mundane moves like 5...Rf3 that lose regardless. We're not having a debate; I'm just explaining how endgame studies work. No worries if you're not interested in learning, but since we're in a public forum, I don't want other beginners to be misled by you into thinking one of the greatest studies of all time is incorrect.

True, but I can still get a queen. I just have to make a waiting move with my king first and then promote without stalemate.