To study mainly the popular moves would, in my opinion, contradict Markovich's idea to "prove +=" for White. But I'll watch what happens. There ARE moves which are fine, only unusual and therefore unpopular. I believe that 3...Bc5 = is objectively best. I am not fond of Na5, g5, Bb4, f6, but 3...a5 deserves a look. - As arbiter we have engines, if we regard lines above, say, +0.35 (in R4) as "solved", fine with me.

Markovich wrote on 07/24/11 at 13:32:57:

3...f5 4.Nc3, I have no doubt about it. If White indeed plays optimally he/she has fine winning chances and Black none. But if the consensus is that 4.d3 is to be investigated first I'll accept. After all Radjabov is unbeaten after 4.Nc3 fxe4 5.Nxe4 Nf6. This anyhow means we should not pay too much attention to 5...d5; 4...Nd4 and 4...Nf6, even if the latter is my preference for Black.
Also it's OK with me if other 3rd moves come first. I just think we should not debate too long about the procedure, so this post is meant as a kick-off. As Markovich started this thread he should decide. I have faith enough in his persistance that all Black's Defences will get a chance. I also have faith enough in his common sense that he will postpone the really hard subjects (like the Marshall) until the off-beat stuff and sidelines are done.
So frankly I don't care too much on which criteria his choices will be based.

So yes, I am highly interested. It will be very helpful in case I decide to return to 1.e4 some day. I will also give IM Van Delft's choices as published in Dutch magazine Schaaknieuws a few years ago.
From this point of view we shoul begin with 3...a5 as Van Delft doesn't mention it.

Some simple statistics from my OM database

All OM games (including Corr.)
3...a6           271013    56,5%      2011
3...Nf6              32486     59,4           2011
3...f5             20075     52,5           2011
3...Bc5          14549     58,8            2011
3...d6            12147     67,9           2010
3...Nd4           6010      61,4           2010

ELO > 2500 both players (including corr)
3...a6               12896    56,7%       2011
3...Nf6               2464    54,4          2011
3...f5                   257    57,4          2011
3...g6                  205    60,0          2010
3...Bc5                    178    55,3          2010

The number of games is a silly measure for the value of a move. In most cases, the popular move is an error. Take the % score, or the difference between av. rating (Black) and his success rating.

One reasonable criterion seems to be the ratio of number of games with alternative move N and alternative move N+1, where alternative move N+1 is the alternative move played more rarely than alternative move N, but more often than any other alternative move which is played less often than alternative move N. Following this criterion and my database, there seem to be three reasonable "cut-off-points" between "mainline" and "offbeat":

(i) 3...Nf6 (22,521 games), compared to 3...a6 (140,731 games), ratio appr. 1 : 6.5,
(ii) 3...Bb4 (363 games), compared to 3...g6 (2,936 games), ratio appr. 1 : 8, and
(iii) 3...Qg5 or 3...Nh6 (1 game each), compared to 3...Nb8 (7 games), ratio = 1 : 7.

So a reasonable cut-off-point both from a statistical point of view and a common-sense point of view seems to be number (ii): 3...g6 is the rarest "mainline" move, 3...Bb4 the most common "offbeat" move. And yes, 3...a5, 3...f6, and 3...d5 would be offbeat too.



Zwischenzugzwang

fling wrote on 07/24/11 at 12:14:51:


I suppose we could first deal with offbeat tries. But we would need a secretary or something to consolidate our best efforts.

I would propose a few principles. 

1. We should always recommend White's most challenging and principled tries; no dodging around theory unless we  sincerely think our way is better.  

2. We should go broad before we go deep (e.g. we should have a complete, provisional outline out to move, umm, 12 before we go significantly deeper anywhere). 

3. We should never go any deeper than necessary to establish a consensus among ourselves. I have little patience with people who want to analyze out to move 57 just to prove that their pet variation actually works, and if we permitted that, we'd never finish. 

4. Finally I think our starting point is 3.Bb5 +=; anytime that seems to fail, we should look a little deeper.

fling wrote on 07/24/11 at 12:14:51:

The box with the offbeat systems is the one with 3...a5, no?