There is also a Slav-Diemer gambit you can use against the Caro: https://lichess.org/analysis/pgn/1.e4+c6+2.d4+d5+3.c4+dxe4+4.d5+cxd5+5.cxd5+Nf6+6.Nc3+e6+7.Bg5+Bb4%20#14
As for how dumb it is, the Diemer-Duhm (the simplest version where you fork the bishop) has a probability of 53% while the one above has one of 45%. Which means it really works. BTW, those are calculated on 1800+ blitz+ games.
As a funny addition, @b5towin said in #9:
> You are almost on the right track - you try these kind of tricks either in bullet against strong players OR in rapid against medium players.
> This is because 2100+ instinctively play Bb4 in bullet, while players 1900- may need a few seconds to find out Bb4 indeed attacks d5 often enough.
> Statistics should confirm this.
In fact you are correct. The probability that people fall into either Diemer gambit actually increases with rating.
How and why does it work with higher ELO
@Dancho_BlackJ said in #13:
> How and why does it work with higher ELO
In the critical position, if I just consider blitz+ the probability of Black playing Bb4 is 59%. If I filter out anything under 1800 the probability goes to 55%. Anyway, the difference is minor. If I add the Bullet time controls the probability goes to 53% for 1800+.
However, if you take the average probability to get to that critical position and play Bb4, in All the games the value is 47%. If I remove the bullets, still 47%. The probability goes to 53% if I remove everything under 1800. And just for fun, I removed everything for under 2200, which you should agree it's a good rating, and the average probability per move to fall into the trap is 51%.
So it's not that higher rated players will move Bb4 more, it's that they will get into the critical position more often.