fear and loathing...ok you get it i was in las vegas...(warning, looong post ahead)

Avatar of wango
| 0

Yes kids, yours truly was in Vegas for 3 weeks!  But it was for business not pleasure.  I was able to hit a chessclub and play a game with sane time controls (G90) but more on that later.

I did want to start talking about my upcoming chess plans.  A few things that I've realized.

1. I'm getting old.  Not "Get off of my yard!" old, but suffice to say I'm getting older.

2. I've been doing this chess thing for 5 years now, I started at a rating of 1008 I'm currently rated 1560, all things considered not bad.

3. I will never be a master.  My ultimate goal is expert, I will have to satisfied with that as I have other shit to do in my life besides chess.

It should come as no surprise to some of you that I have had a devil of a time getting my arms around a suitable opening repertoire, especially with the white pieces.  My score with black is actually better than with white!  I have played the Alekhine and Modern against 1.e4 and I have played the Semi-Slav against 1.d4.  But my white repertoire has been a rather sad affair.  I have tried in vain to play 1.e4 but alas the Sicilian has proven to be beyond my grasp.  I have experimented with the Reti and it has given me good results but sometimes I want something more direct.  Alas chess opening preparation is a zero sum game, compromises must be made.  You can't get everything, I have finally found something that I think will work.  First let's see how I came to this conclusion.

When I was in grad school I didn't have time for chess, and when I did get a few minutes to think about the game I would often fiddle around with Chessbase.  I researched different openings and looked at their statistics.  To see if there is one theoretically best opening for white.

**WARNING: THIS WAS A NON-SCIENTIFIC DIVERSION, I IMPLORE YOU TO NOT READ TOO MUCH WHEN LOOKING AT OPENING STATS***

I'll cover that later on though.

Here's my methodology.

  • I looked at games between 2001 - 2010, that's when my database ends. I never upgraded it, once again other things to spend my money on.
  • I looked at desirable outcome differential for white based on 100 games, I assumed a desirable outcome would be white wins.  As an example if an opening has a decisive game percentage of 68% and white wins 54% of those games we get 100(.68)(.54) = 37.62 so white wins 37 (round up) out of 100 games.  Black would win 100(.68)(.46) = 31.28 or 31 games.  White is better by 6 games.  This was done to eliminate deceiving openings where white wins 56% of the games but only 52% of them are decisive, in that case white has an expectation of 6 or just about average.  On the surface it looks like white is doing much better but in reality its just an average opening.
  • I separated games of those <2200 to those of professionals.  For arguments sake  I consider anyone over 2200 a professional although I realize that many GM's have other non - chess full time jobs and I'm sure the vast majority of masters do as well.
  • The reasoning for this is that as the skill level increases you get a clearer picture of how games "should" go with better skill and a lower probability of being fooled or tricked by an unusual opening.

Shortcomings of methodology

  • Assuming a positive expectation for white only when s/he wins.  If I'm playing someone 200 points higher than me then a draw is a perfectly acceptable outcome even with white.  After all the opening is only one phase of the game, and it doesn't matter how booked up I am my much higher rated opponent should win.
  • 10 years is convenient but does not tell the whole story, there are variations which are out of fashion and simply not played much, there are undoubtedly good lines that are just being ignored right now, these will be under if not unrepresented in this survey.
  • Transpositions, I didn't have an unlimited amount of time for this so there are transpositions that I could not cover, so I'm keeping it as simple as possible, I have an example of this later on.
  • The professional filter is used on both players, so I am not tracking where one player is rated >2200 and the other is <2200, this should not have a great affect on the data, but it bears mentioning.
  • I'm sure there are others but that's all I can think of now.

Results:

Overall in the past 10 years there are 2,499,604 games in the database, over half of the games in the database are from this period!.

The number of decisive games is 71%, with white winning 54% and black winning 46%.  So overall white has an expectation of 5.68 or 6.

There are 680,238 professional games from this same period with 59% of them being decisive white wins 55% and black 45%.  White has an expectation of 5.9 or 6, just what it is at the less exalted levels of chess.

OK I hear you all saying or at least thinking it.  So what's better 1.e4 or 1.d4?  Fear not dear reader I'm coming to that now.  All of this data comes at the earliest point in the game, in other words after black has moved; the earliest possible point where you can tell what opening is being played.

  Opening Games Decisive White W Black W Delta
Total* 1. e 4 1,316,400 74 53.0% 47.0% 4.4
Professional 1. e 4 320,906 60 54.0% 46.0% 4.8
             
Total Open 310,077 75 56.0% 44.0% 9.0
Professional Open 64,216 55 55.0% 45.0% 5.5
             
Total Sicilian 541,379 73 51.0% 49.0% 1.5
Professional Sicilian 151,315 62 53.0% 47.0% 3.7
             
Total French 172,855 72 53.0% 47.0% 4.3
Professional French 41,653 60 55.0% 45.0% 6.0
             
Total Caro-Kann 93,174 71 53.0% 47.0% 4.3
Professional Caro-Kann 23,867 58 54.0% 46.0% 4.6
             
Total Alekhine 28,241 73 51.0% 49.0% 1.5
Professional Alekhine 5,941 61 56.0% 44.0% 7.3
             
Total Modern 35,453 74 51.0% 49.0% 1.5
Professional Modern 9,466 66 54.0% 46.0% 5.3

* Total games taken, many games in the database have players with no rating, hence they do not show up when attempting to get amateur results ie. rating range 0 - 2199 will eliminate many games and do not give an accurate total result.  I am aware that separating  the amateur games would be ideal but I can't figure out how to capture it.

So what does this tell us about 1.e4?  Overall it leads to a less than ideal 4.4 - 4.8 positive result for white.  This is due to the myriad of options available to black on move 1, allowing him/her to steer away from open games where white is 9 for amateur players and an acceptable 5.5 for professionals.

Statistically the Sicilian is the most challenging defense for white to face.  No surprises here but look at how truly awful it is for white if you're not a professional!

"Tricksie" defenses such as the Alekhine and Modern are good choices for an amateur as the Hypermodern practitioner is probably more familiar with the positions, once you go up the rating ladder though, it seems as if White players have it figured out and they get good results.  Specifically my beloved Alekhine does not fare well when those playing are above 2200.  One interesting thing about these Hypermodern Defenses is rating.  In the Modern Defense black's rating is 36 points higher than white and in the Alekhine it is almost even, with black having a 17 point advantage.

  Opening Games Decisive White W Black W Delta
Total 1. d 4 765,250 70 54.0% 46.0% 5.6
Professional 1. d 4 239,188 58 55.0% 45.0% 5.8
             
Total 1…d5 2.c4 158,144 69 58.0% 42.0% 11.0
Professional 1…d5 2.c4 52,571 54 57.0% 43.0% 7.6
             
Total 1…d5 2.Nf3 84,766 69 55.0% 45.0% 6.9
Professional 1…d5 2.Nf3 20,096 55 54.0% 46.0% 4.4
             
Total 1…Nf6 384,137 69 53.0% 47.0% 4.1
Professional 1…Nf6 133,042 59 55.0% 45.0% 5.9
             
Total 1…Nf6 2.Nf3 115,561 68 52.0% 48.0% 2.7
Professional 1…Nf6 2.Nf3 34,200 58 53.0% 47.0% 3.5
             
Total 1…c5 13,260 75 52.0% 48.0% 3.0
Professional 1…c5 2,886 68 58.0% 42.0% 10.9
             
Total 1…g6 22,379 72 52.0% 48.0% 2.9
Professional 1…g6 6,765 65 54.0% 46.0% 5.2

First off, those of you with a keen eye will realize that the total 1.d4 games for this sample is 765,250 but the sum of the variations I give is 778,247.  The reason for this is my inclusion of 2.Nf3 lines.  They can be reached via the move order 1.Nf3, u 2.d4 and we get data carryover that we did not intend.  I have a rather startling example of this later on.  So while there is some non 1.d4 opening data in here for our (or mine rather) purposes it is fine.

A few things of note, 1.d4 scores better than 1.e4.  The Queen's Gambit Declined scores better overall than for professionals although at 7.6 it's still better than any expectations for 1.e4 openings.  Once again "Tricksie" lines such as 1...c5 (heading towards a Benoni structure) or 1.g6 fare better at lower levels.  Although the 1.g6 lines can transpose easily to a Modern or Pirc depending on how White reacts.

So what can we learn from all of this?  Well 1.d4 is a better move from a positive expectation standpoint for white, given that s/he knows the opening in question.   As you and your opponents get better there are going to be more drawn games, but the reality of an opening at top levels can be wildly different down here for us mortals.

I also believe that if you want to become really good 2200+ someday you will probably want to take a mainstream defense.  While the Modern is a great weapon sub 2200 it loses its luster past this mark, although it appears that it is still playable and there are GM's who employ it.

Does that mean that you should switch your entire repertoire?  No!  If you like what you're playing then continue playing it.  After all the Modern works for me, but I don't mind having my pieces be a bit constricted, if you don't like this you are probably going to have awful results with the Modern.  If you like the King's Gambit, then 1.d4 followed by 1.c4 is probably going to drive you nuts.

This was just a diversionary exercise for me.  It is at the end of the day just a bunch of numbers and I can't possibly take in all of the variables for an opening, there are rating discrepancies, dozens of variations to consider.  Here is one example.

What is the difference between these two positions?

AND

Ah, you think me daft eh?  See to the untrained eye it appears as if I have merely copied the same position down.  But not so!  See the first position arises from the following sequence, 1.e4, c5 2.Nf3, d6 3. Bb5+, Nc6.  While the second position arises from the move order 1.e4, c5 2.Nf3, Nc6 3.Bb5, d6.  OK now you are thinking what the hell is the difference well I'll show you.

See, after 3.Bb5+ there are 3 reasonable responses by Black.  Block the check with the Knight with 3...Nc6 or 3...Nd7 or with the Bishop 3...Bd7.  However as the following table shows.

  Opening Games Decisive White W Black W Delta
Total 3.Bb5+ 14,763 65 51.0% 49.0% 1.3
Total  3... Nc6 5,465 68 54.0% 46.0% 5.4
Total 3... Bd7 9,324 63 54.0% 46.0% 5.0
Total 3... Nd7 3,130 67 47.0% 53.0% -4.0

The total of the responses equals 17,919 or 3156 more than the total for the sample.  So the only answer is that there must be some move order shenanigans going on.  So to address this I created a separate database using only the 3.Bb5+ move order, in other words I would only get the original 14,763 games that I was originally looking at.  Here are the results.

  Opening Games Decisive White W Black W Delta
Total  3... Nc6 2,303 69 48.0% 52.0% -2.8

So in the 3,162 games where the position was reached through a 2. Nc6 3.Bb5, d6 move order white scores better!  White's results are terrible with the Moscow (3.Bb5) move order.  What does it mean?  I'm not sure, but my guess is that the 3...Nc6 move order is used by Black players who are comfortable in these positions but what they are trying to avoid?  I'm not sure,  why not just play 2...Nc6?  The rating may give a clue.  On average the black players were 40 points higher than their opponents, perhaps its a trap meant to lure opponents into thinking they can get into the 3...Bd7 lines, which seem to score well for white.

If I make a database simply using the Rossolimo move order (2...Nc6 3.Bb5) we get the following.

  Opening Games* Decisive White W Black W Delta
Total 3.Bb5 21,787 69 58.0% 42.0% 11.0
Total  3…d6 3,159 67 58.0% 42.0% 10.7
Total 3…g6 10,661 69 58.0% 42.0% 11.0
Total 3…e6 4,240 70 55.0% 45.0% 7.0
Total 3…Nf6 1,137 67 54.0% 46.0% 5.4
Total 3…a6 842 80 65.0% 35.0% 24.0

*There are a few less variations, I didn't cover them all as combined they equaled less than 1800 games and I didn't need to get into it for the purposes of this post.

White scores better here in the 3...d6 lines and s/he is rated 16 points higher than his/her opponent.

So what does it all mean?  Statistics can only give you a broad 30,000 foot view.  once you get more than a few moves into a game there are so many transpositions and variations that it doesn't even make sense to talk about a specific opening any more at least at patzer level.  There are simply too many variables.  I did this over the course of a few weeks to see if I could scientifically come up with a way to determine the best opening repertoire.  But alas I am not a machine and cannot think like one.  I am human and there are going to be far more variable involved in my success at the chessboard than the opening.

I did use this method as a very general guide.  How you may ask?  Well that's going to have to wait until next post.

Have a good week friends.