Choosing an Anti-sicilian.

You might have a valid point there. Remember that to simplify the analysis I assumed that the most popular moves were the best, which is a good approximation but doesn't always hold true, especially with smaller sample sizes. These charts are time-consuming to do by hand, but following your question with my approach to analysis suggests I could do a few more charts for the most popular defenses against the Open Sicilian (1. e4 c5 2. Nf3).

Per 365chess stats, below are all the known responses to 2. Nf3 in the Sicilian. Names in quotes are temporary names I assigned, definitely not official...

The 3-element vectors show: [<White wins>%   <draw>%   <Black wins>%]

2...d6 {"Ftacnik Defense."}{202868/505289 = 40.1% popularity on 365chess.}

[37.2%   29.2%   33.7%]
2...Nc6 {"Sveshnikov Defense."}{159902/505289 = 31.6% popularity on 365chess.}

[36.8%   30.3%   33%]
2...e6 {"Cramling Defense."}{118876/505289 = 23.5% popularity on 365chess.}

[34.8%   31.2%   34%]
2...g6 {Hungarian Variation.}{10161/505289 = 2.01% popularity on 365chess.}

[35.5%   30.6%   33.9%]
2...a6 {O'Kelly Variation.}{7803/505289 = 1.54% popularity on 365chess.}

[37.2%   25.7%   37.1%]
2...Nf6 {Nimzovich-Rubinstein Variation.}{4620/505289 = 0.914% popularity on 365chess.}

[39.1%   30.8%   30.1%]
2...b6 {Katalymov Variation.}{714/505289 = 0.00141% popularity on 365chess.}

[39.4%   25.2%   35.4%]
2...Qc7 {Quinteros Variation.}{119/505289 = 0.000236% popularity on 365chess.}

[44.5%   28.6%   26.9%]
2...d5 {"Fernandez Variation."}{73/505289 = 0.000144% popularity on 365chess.}

[53.4%   21.9%   24.7%]
2...Qa5 {Stiletto Variation.}{65/505289 = 0.000129% popularity on 365chess.}

[47.7%   32.3%   20%]
2...Qb6 {"Rodchenkov Defense."}{30/505289 = 0.0000594% popularity on 365chess.}

[60%   16.7%   23.3%]
2...h6 {"Drazic Defense."}{24/505289 = 0.0000475% popularity on 365chess.}

[41.7%   33.3%   25%]
2...f5 {Brussels Gambit.}{14/505289 = 0.0000277% popularity on 365chess.}

[71.4%   7.2%   21.4%]
2...e5 {"Holden Defense."}{12/505289 = 0.0000237% popularity on 365chess.}

[58.3%   8.4%   3.3%]
2...f6 {"Poirier Defense."}{5/505289 = 0.00000990% popularity on 365chess.}

[40.0%   0%   60.0%]
2...a5 {"Groz Defense."}{2/505289 = 0.00000396% popularity on 365chess.}

[50.0%   0%   50.0%]
2...Na6 {"Fargere Defense."} {1/505289 = 0.00000198% popularity on 365chess.}

[0%   0%   100%]

REFERENCES
http://www.365chess.com/opening.php?m=4&n=4&ms=e4.c5.Nf3&ns=3.3.4 (6-24-15)
https://gameknot.com/chess-opening/sicilian-defence-b27?nd=523 (6-24-15)
http://allchessopenings.blogspot.com/ (6-24-15)

It looks to me like only the top three responses are worthy of analysis because the popularity drops off rapidly after those, down to a tenth of the previous popularity percentage. Three charts won't take *too* long to make, so if that sounds like it would significantly help identify any good anti-Sicilians, I'll do that. Please let me know if you can think of other analytical approaches that would be convincing. By the way, yesterday I figured out how to fit a line to such data where the smaller sample sizes are progressively ignored in proportion to their relative weight as the number of ply incease, so I'm eager to try that out.

I haven't been trying to ask you a whole lot of questions because I know how time consuming and painful lol doing analysis can be lol.

I saw a flaw in your analysis. Which is why I asked the question. My purpose was to make sure you was aware of the flaw. I don't know how to say what I'm thinking lol. So it might come out stupid. I don't want you to think I'm talking bad about your analysis. I just want you to be aware of the subtle flaw so you don't get mixed up like I did.

For example:

After the moves 1.e4 c5 2.Nf3

2...d6

The percentage as

[<White wins>%   <draw>%   <Black wins>%]

[37.2%   29.2%   33.7%]

The flaw is the above percentage is not the actual percentage.

Its a fake percentage. It is an illusion.

I've been playing the Delayed Alapin for a few years now without any real theoretical knowledge of it, and have been really enjoying the resulting positions, it seems White can get the nice 2-pawn center, though it helps if you also don't mind Advance French type positions (d4, e5)

Is anyone familiar with any books devoted to the Delayed Alapin?

I don't know what you mean. The best I can interpret your question is that there is a difference between the theoretically ideal value and the statistically sampled value, such as the difference between the "theoretical mean" and the "sample mean" (http://www.chem.ox.ac.uk/teaching/Physics%20for%20CHemists/Errors/Expectations.html).

I don't play the Alapin (quite easy for Black to handle), but I really like playing the delayed one after 2...d6.

You will find quite a bit of material in "Experts on the Anti-Sicilians" (Quality Chess) mainly about the critical line 1.e4 c5 2.Nf3 d6 3.c3 Nf6 4.h3.

Ok ok. Delayed Alapin is great! I played it for a while myself. I liked playing the variations where you go Bd3 before going d4 (pull the bishop back to c2 obviously), and you end up getting some Ruy Lopez variations if black tries to play e5. However, I feel like there is an obvious line that people are not discussing in this forum called the Smith Morra Gambit. It's kind of Alapin's cousin. If black tries to ignore c3 you get to play cxd4, side stepping a lot of annoying anti Alapin lines.

If white accepts the Smith Morra Gambit of course then the fun begins! It may not be particulary popular at GM levels, but there's all sorts of fun ways to play that leave black cowering in fear of your multiple threats.

Side note, the Nf6 variations really give the Smith Morra gambit and the Alapin a run for their money.

alapin

Interesting, I've never even considered 4. h3, I've been playing the 4. Be2 lines willing to play the 2-pawn sac after 4...Nc6 5. d4

Ah yes, the 'Kopec System'? :) 

https://www.youtube.com/watch?v=zbmcoohIWjY

4.Be2 is interesting if Black enters the 4...Nc6 5.d4 mess, but he may not comply, e.g. 4...g6 when the e2 bishop is actually misplaced.

4.h3, which limits the c8 bishop is the indicated move, usually followed by Ruy-like buildup Bd3/Bc2 etc.

I'll try to explain what I mean best I can.

The flaw was that you was looking at percentages based only on possible move 2 options. Which gives a flawed account for actual lines.

If I was to ask you what is your percentage of winning or losing in the Sicilian Dragon.

The below percentage would be wrong.

2...d6 -  [37.2%   29.2%   33.7%]

The move 2...d6 is played in the Sicilian Dragon;however, there are also other lines played which come out of 2...d6.

They are also influencing the over all percentage above.

The Sicilian Dragon starts at move 5...g6

If you use the below link which you provided before you see the percentage of the Sicilian Dragon.

http://www.365chess.com/opening.php?m=10&n=3&ms=e4.c5.Nf3.d6.d4.cxd4.Nxd4.Nf6.Nc3&ns=3.3.4.4.3.3.4.4.3

5...g6 -  [41.5%   25.6%   32.6%]

Now this percentage above here is only the Sicilian Dragon as black. If you look at one of the critical lines in the Sicilian Dragon by White. You will see the true percentage of the line.

(Yugoslav Attack it is considered one of the critical lines in the Dragon)

6.Be3 -  [44.2%   25.1%   30.7%]

If we was to look at this percentile 2...d6 -  [37.2%   29.2%   33.7%]

We would think the Sicilian Dragon is a great.

In reality this is what really is happening 6.Be3 -  [44.2%   25.1%   30.7%]

The difference between 2...d6 line

37.2% - 33.7% =  3.5%

The difference between 6.Be3 line

44.2% - 30.7% = 13.5%

It is a 10% jump in favor of white in this line.

The reason I bring this up is become I sometimes play the Closed Sicilian lines.

2.Nc3 - [35.8%   27.1%   37.1%]

If I look at the above stats only. I would get the impression white is doing terrible.

Black has a winning percent of 1.3%.

Also others on the forum have gotten the same impression see below.

However, It is a great opening. The lines that can arise from it have nice scores. If you are willing to manipulate the position. If you are willing to play differently to meet the needs of the position. I believe you will get good scoring chances.

The problem people have is they want a 1 shoe fits all kind of a chess line.

They want 1 line that can be played against everything. In truth it is not realistic or even possible. I am no exception to this problem. I have had this same problem and have struggled with it for months and months.

I had an online friend who use to brow beat me so to speak. I love playing the KIA and I would play it against everything. He talk so much smack to me. During a game he would say stuff like "Hey look its player J. The person who plays the autopilot K-I-A".

He would say stuff like " Hey look player has a mate in 1 but he is not going to do it. He will play g3 first than Bg2 than castle and if the mate is still there by the time he finishes all his autopilot moves than. He will mate them."

My friend talked so much smack lol. He wasn't a coach or anything but that is what it took for me to over come this problem. I was that stubborn someone had to brow beat me and talk smack lol to get me to change my terrible ways. Which got me to realize trying to use the same line against everything is not realistic.

The reasons I have told you this now is because of the following:

2.Nc3 - [35.8%   27.1%   37.1%]

How can this move be considered better for black if you have so many options here. The only lines which have been effected by the move 2.Nc3 are ones that maybe deal with the C pawn. 

Macrozy Bind, Alapin, Delayed Alapin, and Smith-Morra that is it.

We still can play some Anti-Sicilian lines here. We still can play the open Sicilian here. We still can play closed sicilian lines here.

Which was the point I was getting at. The sample length is not great enough. We can't checkmate anyone from move 2. An there is tons of lines which can be followed up from move 2 in this position. Which can be effecting the percentage above.

Here are those three charts I said I would try to create and post, showing the three top responses to White's Open Sicilian: 1. e4 c5 2. Nf3:

2...d6

2...Nc6

2...e6

Remember that the last third of so of such graphs are basically noise, and worthless for evaluation.

All these graphs are pretty similar. The Najdorf (2...d6) graph I posted before, with only a slightly different heading, and the other two are qualitatively similar. In the ...d6 variations White has a slight advantage out to about move 10, and in the ...Nc6 and ...e6 variations White has a slight advantage after about move 10. In the Kan (after 2...e6), White has a more noticably superior position after about move 15, but since that's pretty far into the game, I suspect that's just a statistical anomaly based on selection of one variation, and the outcome evens out after a while anyway. Overall, behavior is as I would expect, with draws (the red dots) gradually increasing toward 100% as the game progresses, and no major anomalies anywhere. This is in contrast to some of those unusual Sicilian variations like 2. f4, 2. d4, 2. Nc3, and 2. c4, where White took a hit that persisted in statistical outcome as soon as he played his offbeat second move. My overall assessment: all significant Open Sicilians are quite similar in behavior and statistical outcome. This suggests that an anti-Sicilian that is unusual and has good statistical outcome with reduced tactics/memorization won't be found among the Open Sicilians.

I see what you're saying now: you're objecting to accumulating statistics from a complicated tree into a single number. I will respond with these thoughts:

(1) What we're attempting to do here is summarize. Unless generalities can be made, all that's left is to wave one's hands in exasperation and say "Go look at an openings database to figure out which move is best." That's obviously thousands of times too complex to be useful for human understanding.

(2) You're essentially referring to the mini-max algorithm (https://en.wikipedia.org/wiki/Minimax#Minimax_algorithm_with_alternate_moves) by implication, which does nothing useful here that I can see. To take a database of a move tree, then to go through it and select only the best outcomes, then propagate that information up to near the root of the tree for generalization purposes is what players are already doing when they select the moves they like to play best as White or Black because it gives them the best chances. That's why the method I'm using, of choosing the most popular moves, takes advantage of the work already done by players and by the automatic sort in 365chess' database. Mini-max would just do virtually the same thing, unless *maybe* you're suggesting a greedy algorithm approach.

(3) Ultimately you're referring to what computer science calls a "greedy algorithm" (https://en.wikipedia.org/wiki/Greedy_algorithm) if you're suggesting that at each step of the move database we select the most winning or most drawish outcome, then follow that branch as the best for giving us that desired outcome. The greedy strategy fails in chess moves since at each step the encountered database value often negates the value we encountered just one move earlier. The best way to overcome that would be to use the minimax algorithm,but then we're back to (2). There's no magic method guaranteed to work here, just ad hoc methods, although if you think of another algorithm that will provide more insight into the form of a useful summary, please let us know.

(4) Most books about openings mention that the power of the Sicilian for Black is largely (mostly?) based on Black being able to open the c-file to attack by exchanging his side (c-) pawn for one of White's central (d-) pawns, plus counterattacking White's e-pawn. I believe this is the key to a successful anti-Sicilian. That's why the Staunton-Cochrane (2. c4) interests me so much as a good candidate: by the 2nd move Black's entire strategy of opening the c-file has been thwarted by White's blocking c-pawn. I'm still surprised the statistics and position seem to favor Black in the Staunton-Cochrane, although not enough to give Black a guaranteed win. Even in the Symmetrical English White strives to play d4, so maybe a more successful Staunton-Cochrane would be based on a wise choice of when White should play d4 instead of refraining from it altogether. If that's the case, it would take a lot of wisdom and experimentation to come up with an optimal set of guidelines for that d4 push. As for counterattacking White's e-pawn in the Staunton-Cochrane, that is pretty much ruled out since it is backed up with White's pawn at d3, so even if Black were to trade off White's e-pawn, that would just open up the d-file (or maybe f-file), which would facilitate the trade of pieces.

----------

(p. 182)

     SICILIAN DEFENCE

   (1 P-K4, P-QB4)

MENTIONED by Polerio in 1594, given its name by the Italian master

Greco in the seventeenth century, the Sicilian received its earliest

practical tests and a big boost in popularity in the MacDonnell-La

Bourdonnais match (1834), Staunton v. St. Amant match (1843), and

the great London Tournament of 1851. Since then it has experienced

repeated refutations and rehabilitations. But "age cannot wither nor

custom stale its infinite variety." Barely a year passes which does not

cause us to re-evaluate some variation. Today the defence is a regular

in every tournament, and one should refrain from 1 P-K4 unless

prepared to confront it.

   The Sicilian owes most of its effectiveness to the semi-open Queen's

Bishop's file produced after White plays P-Q4 and the inevitable

exchange of pawns. With 1...P-QB4

Black immediately puts pressure on the Queen's

flanks and plans complete liberation with ...

P-Q4. The move also strikes at Q5 and

QN5, half opens a file and frees the Queen.

From the strategical point of view the Sicilian

has the advantage of allowing a large choice of

pawn-configurations. Moreover it avoids pre-

pared variations and certain drawing openings

like the Giuoco Piano and Four Knight's Game.

Evans, Larry, and Walter Korn. 1965. Modern Chess Openings, 10th Edition. New York: Pitman Publishing Corporation.

----------

(p. 66)

      Sicilian Defence:

   1 e4 c5

   Like the Caro-Kann, the Sicil-

ian begins by breaking the sym-

metry. But unlike that defence, it

does not do so merely to hold the

centre, but to institute a counter-

attack on the queen's wing. For

that reason the outstanding charac-

teristic of the Sicilian Defence is

that it is a fighting game. Both

players must necessarily seek their

objectives on different sides, which

can lead to deliciously compli-

cated and exciting variations.

   Because the Sicilian is more of

a unit than most other defences it

is possible and worth while to

lay down a number of general

principles which will be found to

be valid in a large majority of

cases.

   White almost invariably comes

out of the opening with more ter-

rain. Theory tells us that in such

cases he must attack. He does so,

normally, by g4 followed by a 

general advance g5, f5, and eventu-

ally f6. In some cases he may castle

long (in that event he must weight

the counterplay which Black can

undertake). One of White's major

positional objectives is the preven-

tion of ... d5.

   Normal play for Black consists

of pressure on the c-file, especially

c4. Coupled with this is keeping

White's e-pawn under observ-

ation. The counter-attack against

the e-pawn may also be quite

strong independently of the play

on the c-file. 

Fine, Reuben. 1989. The Ideas Behind the Chess Openings, Algebraic Edition. New York: Random House, Inc.

The best Anti-Sicilian is 1. d4. All the GMs are using it nowadays.

In all honesty, this isn't a terribly bright comment- but I guess you know it, already.

Actually, 1. d4 completely avoids the Sicilian- not like these Anti-Sicilians which allow different versions of it where white doesn't open the position. Maybe that makes it the ultimate anti-Sicilian.

I find it amusing that you actually responded to my first comment. But, truthfully, I do play 1. d4 as white primarily, and a part of the reason why is that I don't enjoy facing certain variations of the Sicilian.

And then runs into the Dutch, King's Indian, Nimzo-Indian, etc., not to mention a body of literature that probably greatly exceeds that of 1. e4. Ain't no easy way out.

@Sqod I don't think the body of literature exceeds it. And in any case, white does not have to go into anything as cutthroat as some of the Open Sicilians.