gambits harmful for the prospects of improving ?

It sounds a little bit like you are arguing that players should not be creative, nor favour sharp play.  Some prominent SuperGMs have based their whole careers around unusual openings that attempt to trap their opponents: Bent Larsen, Duncan Suttles or more recently Alexander Morozevich and Alexei Shirov.

I agree that you should probably not play openings where the refutations are well known or easy to guess, but there is certainly nothing wrong with preparing new unusual moves that may catch your opponent offguard.  Players who can produce novelties of that sort on a regular basis are perhaps some of the most effective.

Estragon, according to Hooper and Whyld, a gambit is "an opening in which one player offers up material, usually a pawn, sometimes a piece or more, in the expectation of gaining a positional advantage."  In the Queen's Gambit, obviously the pawn can be recovered immediately, so I don't think it really matters if it is a sham sacrifice or not.

SuperGMs like Kasparov, Morozevich, Shirov and Michael Adams base their repertoires largely on gambits, and they seem to do well enough.

 The GM is pointing out that questionable gambits ( he even names several ) are bad for developing players, I agree. The GMs you named do NOT play such questionable gambits as the OP named...... there's a world of difference between say the Latvian and the Marshall counter attack.....

Reb, I was responding more to Estragon than the original poster.  As I was hinting earlier in this thread, I agree that the Latvian is no Marshall.

There are tons of strong gambits: Ruy Lopez 5.0-0, Perenyi, Alekhine-Chatard, Caro-Kann Fantasy, Caro-Kann Modern Exchange, Benoni Mikenas, Queen's Indian Franco, Slav, Shabalov, Semi-Slav Marshall, Semi-Slav Botvinnik, Nimzo-Indian Leningrad Averbakh, Slav Winawer, King's Indian Saemisch Be3 c5, Four Knights Marshall, Ruy Lopez Open Karpov, English Bellon-Lopez, English Flohr-Mikenas, to name just some of the more common ones that are played at the GM level.  It sounds like some people here are underestimating the wide variety of interesting gambits a strong player can choose from.  I am sure that there are many more just waiting to be discovered.

Yeah but the majority of sound gambits in the opening are white-gambits. Almost all black gambits are somewhat dubious. You can't expect a gambit repetoire be equally good with both colors.

Well said.

Long-term skill development above and beyond mediocre club player will require study of classic, hypermodern, and modern openings such as the Spanish, French, Russian, English, Sicilian, ...

Offbeat openings, such as the Latvian, Polish, Englund will retard one's development if relied upon too heavily.

Of black gambits, the Ruy Lopez Marshall, Benko, Caro-Kann Modern Exchange, Four Knights Marshall/Rubinstein, King's Indian Saemisch Be3 c5 and Petroff Pillsbury are common enough among SuperGMs.  Probably the Slav Winawer and English Bellon-Lopez are playable to a fairly high level.  I agree though that white has a much wider choice of gambits available. 

"When you play 1.e4 e5 2.f4, learning starts immediately. If you're playing the Colle or something that's just not always possible."

this, this this. yes, i agree that most of the risky gambits aren't sound at higher level, but for learning tactics, they are VERY good to play as you are learning. Lines of the Kings Gambit accepted and declined were some of the very first I was taught. They have influenced me and my tactics are very sharp due to playing so many lines where I had to use my initiative. I even enjoy the more brutal lines of kings gambit such as the muzio from time to time.

Playing gambits to sharpen your tactics is like playing hockey to improve figure skating. If you want to improve your tactical vision, do exercises. There isn't a single opening that is tactics-free. Even in the most dull symmetrical lines, there are traps and tactics. 

The way I see it, playing gambits to win games is like trying to pick-up women with Johnny Bravo lines. Sure, you'll eventually snag one, but you'll get a lot of slaps before you do. Why not act sensible and play something that has stood the test of time? Trust me, there are plenty of lines where you can sacrifice pawns early on in almost all the main theoretical lines. Look at the recent popularity of the Catalan, which is basically a hyper-modern gambit.

And I see lots of people have a problem making a distinction between a gambit and a sacrifice. A gambit usually happens early on and the pawn or the piece is actively offered , i.e. the Ulvestad, or deliberately left en prise, i.e. the Muzio. Lines like the Grunfeld Exchange with Rb1 or the Poisoned Pawn in the Najdorf aren't gambits, because the side winning the material does so on its own volition, not because it was offered directly.

In his books, Garry Kasparov often describes a pawn sacrifice as a gambit even if the two players are out of book, but one of the two sides hasn't castled yet.  Judging by the comments in this thread, it would seem that gambiteers and encyclopedia writers tend to have a very wide defition of what a gambit is, while players who disapprove of gambit play think that gambits mainly include lines like the King's Gambit or Evans Gambit.

This is not how I learned, and I'm doing okay.

Here's what I read in the above post.

Something something something Caro-Kann Fantasy, Caro-Kann Modern Exchange, something something something.

First, how is the Caro-Kann Fantasy a gambit or a sacrifice? Second, what is the Caro-Kann Modern Exchange?

I feel like I should know these things.

Ozzie, Anatoly Karpov and Mikhail Podgaets wrote a book called Caro-Kann Advance Variation and Gambit System wherein "Gambit System" refers to 1.e4 c6 2.d4 d5 3.f3.  There are a whole bunch of different gambits associated with this line.  At the GM level, the most common is 3...dxe4 4.fxe4 e5 which Alexander Bangiev calls the Prague Gambit.  3...e5 is the Thomas Gambit.  In a joint book with Mednis et al, Andrew Soltis refers to 3.f3 as the Classical Gambit suggesting that the name stems from 3...dxe4 4.fxe4 e5 5.Nf3 exd4? 6.Bc4, a line which Bangiev calls the Maroczy Gambit.  Karpov and Podgaets deal with that line first, so that may indeed be the reason for the gambit designation.

Soltis refers to 2.c4 d5 3.exd5 cxd5 4.cxd5 Nf6 as the Modern Exchange variation, although here it would appear that white can try to hold onto the pawn making it a gambit from black's point of view.  Incidentally, 3...Nf6 transposes into the Blackburne (Bangiev) or Marshall Gambit (Hooper/Whyld) in the Scandinavian Defence.

2.d4 d5 3.e5 c5 Bangiev refers to as the Botvinnik Gambit.

The Caro-Kann is clearly a favourite amongst contemporary GM gambiteers.

Thanks. Crazy.

I call 3.f3 the Fantasy variation. I call 2.c4 the Accelerated Panov. And I think white should play 3.cxd5 exactly to prevent black playing ...Nf6 against 3.exd5, with the transposition you mentioned.

1. e4 c6 2. d4 d5 3. f3 dxe4 4. fxe4 e5 is almost not a gambit when taking the pawn means facing the crushing Qh4+.

1. e4 c6 2. c4 d5 3. exd5 cxd5 4. cxd5 Nf6 is called the pseudo-panov by many. Just like the botvinnik gambit white can try to hold on to the pawn but in both cases he tangles up his pieces, which is blacks compensation.

Ozzie, some other Caro-Kann gambits if you are interested: 1.e4 c6 2.d4 d5 3.exd5 cxd5 4.c4 Nf6 5.Nc3 g6 6.Qb3 Bg7 Euwe Gambit (Bangiev) or 5...Nc6 6.Bg5 Qb6 Rejfir Gambit (Estrin) although in this latter case, it isn't exactly clear whether it is white or black who is sacing a pawn.

AtahanT, the Prague Gambit definitely is another example of a gambit that is perhaps too strong to be accepted.

First one is quite bad lol. After 4.c4 Bb4+ followed by Qh4+ just kills white.

Black's pawn is on e7, so 4...Bb4+?? would be an illegal move.  Fischer, Tal, Gelfand, Adams et al all play this line as white, so I'd be very surprised if there were a simple refutation.

Ah sorry my pgn was totally messed up while I was browsing the line. Youre right ofc.