It's perfectly likely that this is very, very spotty information on this defense. It remains my current favorite however. Discuss!

I looked into it recently, but made a quick opening judgment that basing a substantial amount of your plan on attacking a knight that can be defended, is rather artificial. It just doesn't seem like the pressure on that knight is really that annoying, and it also seems like something that could gradually be neutralized by white, when afterwards black's pieces might look a little dumb. For example, after 7 cxd5 Nxd5 8 Qb3!? (Qd2 is more common) white intends Rc1, and seems to be able to meet black's plan pretty well, and may contemplate a3 later. Moves like ...c5 and ..e5 seem to leave the knight vulnerable, but taking time out with ...Ndb6 could be a bit too slow.

Even if black could win material on c3, it still wouldn't necessarily be that bad for white since he'd only be losing queenside pawns, meanwhile black's queen would be exposed and white may win the center.

Of course, considering strong players have played it, black probably has a lot of resources (to continue to keep the pressure on white), as is to be expected with dynamic plans, but personally I would rather not play the type of position where I would have to find them.

6...Qa5 would be weak if it merely was an attack on the Nc3. But it isn't. It's a multifunctional move serving several purpuses at once:

1) It mobilizes the Nf6.

2) It demobilizes Nc3 (rather than attacking it).

3) It contains a hidden attack on Bg5.

If Cambridge Springs has a real weakness it is the fact that white can avoid it by playing 5.cxd5 forcing black into the orthodox exchange variation. This is why I prefer to play 2...c6 instead of 2...e6. In the Slavic Defense Cambridge Springs is a good way to avoid the complicated Anti-Meran Gambit (5.Bg5 dxc4 6.e4 b5 7.e5 h6...).

This is actually a very old defense, one of the early attempts for Black to get a more active game from the Queen's Gambit Declined.  "Cambridge Springs" refers to the tournament at that town in Pennsylvania, USA in 1904, where the defense was played a number of times and was accepted by the top players.

 

As IM pfren notes, the progress in openings over the last century has taken most of the sting out of the defense for any prepared opponent, and while it isn't bad, it offers more of a passive equality at best.  Modern players want positions that offer more counterchances for Black.

It has always been popular at the club level, though, because if White is unfamiliar with it there are a number of traps which might ensnare him.

---

There's an ultra-aggressive line black can play that even holds up under heavy computer scrutiny, though it would take some balls to play OTB:

 

 

Black will artificially relocate the king to the queenside, while kicking the bishop out of c7. Then Nd5 will come and look to trade off on the dark bishop. As I said, it takes some serious balls and prep to play this line, but I can see it crushing white off the board if white isn't prepared for it.

I used to like it until I had to face the Exchange Var of the QGD which is very strong vs 4...Nbd7. If it wasn't for the Exchange Var I'd still play the Cambridge Springs Defense, but I've switched to the NID which is much more dynamic for Black IMHO.

But I don't think computers are necessarily that materialistic. When I analyze with houdini, I often see it all the time continue on down a piece for 10-15 moves, still believing that side is better (of course the other side was way down in development). Unless of course it just calculated that far! But if it calculates that far, it's hard to be wrong about almost any position, besides perhaps highly closed ones.

I mean, if there is heavy scrutiny from computers, meaning, I presume, the playing out of tons of variations from that position, it must either be fine, or really, really, hard to refute.

Doesn't it usually go the other way around? Aggressive openings computers claim are lost or something, but are hard to refute when faced with it over the board? Perhaps that happens in the dragon or something?

---

This.

The OP almost completely discounted this line but black's pieces just don't make much sense to me here.

---

The main problem is some masters are still under the false impression that they can judge a position better than the top programs can. That just isn't the case anymore, and many of the top GMs readily admit this. I recall Anand having an anecdote about his meeting with Fischer, where he couldn't get Fischer to understand that computers have come to the point where their evaluation of a position is far more trustworthy than it used to be. So if a computer shows a line as being bad, there's no point in arguing with it. Fischer still wanted to argue, according to Anand.

Now I'll always make the distinction that there are OTB 'playable' and 'non-playable' positions, where both evaluate as say equal according to the computer. The difference comes in human nature. The computer is not phased by a human-wise 'difficult' position versus one where humans feel like they have good chances.

This goes back to a previous beef I had with pfren, where he wrongly claimed a middlegame position was lost for black. The position was very open and tactical, and indeed would have been extremely difficult for a human to defend as black. My point was that the position was NOT technically lost for black, as a multi-core computer with either Rybka or Houdini could in fact defend the position to a draw. Any time a computer can do that, the position is NOT lost. Pfren couldn't accept that, and actually had the audacity to try and argue the point (in addition to insulting me in the process). I've been playing on ICCF for years now (currently in the final round of the US CC championship with a 2300 and climbing rating), where computers are often used for DAYS at a time on each move. As such, I've gained a lot of experience on the subject. On ICCF, powerful software and hardware have leveled the field. A club player with a decent understanding of drawn endgame themes can fair just as well as a full-on GM. I fear absolutely no one there, not even if Carlsen started having a go at it.

In my view anyway, if of course computers confirm that there isn't some tactic to immediately smash black in that position you posted, even from a practical point of view I wouldn't be afraid of taking the black side. I guess black's king is more awkward than usual, but it isn't necessarily that significant if white can't find a way to take advantage of it. I think white is better, because black's position is not very coordinated, but I wouldn't feel scared for my king if I were black.

---

That would imply I know stuff about opening theory, which is certainly flattering, but unfortunately far from the truth.

I am playing Cambridge Springs defence and i think its OK