chess ratings

Backgammon Online

Divide and you will discover!
Nutty name, real matters
[ Sign up | Log in | Guest ] (beta)

Have you ever played Backgammon online? It is a lot of fun. So come play Backgammon online with us!

PLAY BACKGAMMON ONLINE
Sign up now to play backgammon online for FREE!
backgammon online

i_play_slowly 353 ( +1 | -1 )
Black's best responses to 1.e4 These thoughts are based on a quick study of Gameknot's two databases, one for completed +1500 games, and the smaller one for +1900 games.
*
DATABASE +1500
*
Looking at the larger of the two databases, I note that there are four responses to 1.e4 that tend to result in more victories for Black than White. In descending order of their success for Black, they are:
1) 1...e5, which yields victories for Black in 45% of the cases and victories for White in 39% of the cases, i.e., there is a difference of +6% in Black's favor
2) 1...Nc6, with a difference of +4% in Black's favor
3) 1...g5, with a difference of +4% in Black's favor
4) 1...e6, with a difference of +2% in Black's favour
*
Now comes the tipping point. The next four responses to 1.e4 tend to favor White. In descending order of their successes for Black, they are:
*
5) 1...c6, with a difference of -6% in Black's favor
6) 1...d5, with a difference of -6% in Black's favor
7) 1...Nf6, with a difference of -7% in Black's favor
8) 1...c5, with a difference of -16% in Black's favor
*
CONCLUSIONS
*
A) Four of Black's responses tend to win more than they lose. Of these four, the French requires knowing relatively less theory, perhaps making it the ideal opening for a +1500 player.
B) The Sicilian ranks 8th in the hierarchy of successful openings for Black, doing significantly poorer than the next-ranked choice, 1...Nf6. While 1...e5 favors Black by +6%, the Sicilian favors Black by -16%, a 22% disparity. It's poor showing is perhaps due to the amount of theory that must be learned in order to play it adequately.
c) The Scandi, like the French, requires comparatively less knowledge of theory. The Scandi also outperforms the Sicilian by a difference of 10%. Nevertheless, the Scandi still ranks sixth in the hierarchy, tends to lose more than it wins, and appears 12% less effective than the front-running 1...e5.
*
DATABASE +1900
*
When I look at the smaller Gamknot database for +1900 games, none of the responses to 1.e4 favor Black. Ranking Black's responses in descending order of success, they are:
*
1) 1...e6, with a difference of -7% in Black's favor
2) 1...c5, with a difference of -7% in Black's favor
3) 1...d6, with a difference of -8% in Black's favor
4) 1...Nf6, with a difference of -8% in Black's favor
5) 1...c6, with a difference of -12% on Black's favor
6) 1...e5, with a difference of -17% in Black's favor
7) 1...Nc6, with a difference of -19% in Black's favor
8) 1...g6, with a difference of -20% in Black's favor
9) 1...d5, with a difference of -28% in Black's favor
*
CONCLUSIONS
*
A) The two lists vary widely.
B) Compared to its ranking in the first list, 1...e5 does much more poorly, perhaps in part because Black's opponents at this level have mastered the Ruy Lopez.
C) 1...g5 and 1...Nc6 also do more poorly, perhaps because they are basically unsound openings anyway, and they have little 'shock value' against players at this level.
D) The Sicilian fares much better among players whose ratings are +1900, perhaps because they have the time, enthusiasm, and aptitude for the amount of theory that is involved in mastering it.
E) The French ranks highly in both lists.
F) In the +1900 list, the Scandinavian ranks 9th in the hierarchy of successful openings for Black. While 1...e6 favors Black by -7%, the Scandinavian favors Black by -28%, a 21% disparity.
G) Perhaps +1500 players should not burden themselves with the Sicilian unless they have the time, enthusiasm, aptitude, and confidence to aim towards a rating of +1900.
H) Perhaps the French is a good opening for players of all ratings.
I) Perhaps +1500 players should avoid 1...Nc6 and 1...g6, as these openings will lead nowhere in the long run.
*
Have I managed to offend everybody?
chuckventimiglia 17 ( +1 | -1 )
The only thing that stands..... out in what I read above concerns the French.

The French is one of the most complicated openings
to play well and understand.

I do not agree that it requires less knowledge etc.
mattdw 68 ( +1 | -1 )
I am mortally offended. ;)

Interesting read, I'm probably not in a position to comment as I have only been playing about a month or so but do you think it is wise to read so much into statistics? I think there must be many more relevant factors like which openings suit a particular type of player, personally (at my stage anyway). I would rather not just learn openings move for move, but instead learn general principles (tempo, space, etc..) that can be applied to each move individually. I guess I will end up learning many openings when I have got to grips with the basics of strategy, but for now the principles will be fine for me. ;)
i_play_slowly 80 ( +1 | -1 )
mattdw I would like to steer for the middle course in which statistics neither rule my life nor are dismissed without consideration. If it wasn't for stats, I'd still be smoking cigarettes.
*
I completely agree that there are many other relevant factors, like which openings suit a particular type of player, and I applaud your commitment to learning the general principles of chess.
*
Still, given, say, that I am a positional-type player who wants to learn an opening, perhaps I would do well to start looking at positional-type openings that tend to be relatively successful for players with ratings similar to mine, rather than positional-type openings that tend to be relatively unsuccessful for players with ratings similar to mine.
i_play_slowly 46 ( +1 | -1 )
chuckventimiglia "Against 1.e4, the French Defense is an excellent choice. Itís a sound, positional system with plenty of scope for counter-play. It avoids the dense theory of the Sicilian or Spanish while still containing enough variety to hold interest after years of play." -- S. Evan Kreider
-> chessville.com
*
If I implied anything else about the French, I erred.
peppe_l 101 ( +1 | -1 )
mattdw "I would rather not just learn openings move for move, but instead learn general principles (tempo, space, etc..) that can be applied to each move individually."

I agree 100%

Memorizing moves never helps. You will run out of book sooner or later anyway, and then chess understanding (tactics, strategy, endings, etc) will either win or lose the game. My rule of the thumb is "having +/= can only help if you fully understand WHY you have the advantage and how to use it". For example, once I had clear (opening) advantage vs one of the top players here, but he still beat me easily. I choosed wrong strategy, then missed few tactical threats, and boom - it was all over. IMO nuances of opening theory are GM/IM level stuff. For players like me, that's like going for rocket science before learning basic calculus :-)

I applaud I_play_slowly for making some interesting conclusions.

But, personally I never choose moves based on statistics.

PS. It is good to remember each first move has lots of lines...
leo_london 63 ( +1 | -1 )
Interesting.. I agree, purely memorizing a sequence of moves will not improve your play.
However, hopefully thats not what you are likely to do.
If you study opening theory, you will learn why certain responses have proved successful for many years, you will learn new ideas that have become prevalent with modern players etc etc.
I always think of chess openings as a kind of " dance ", the opening moves do dictate to some extent the music you will be dancing to...probably well into the middle game. So, if you have a fiery Latin temperament...dont end up having to do a sedate waltz.
;-)
ironbutterfly 90 ( +1 | -1 )
Reuben Fine rules! Well, as someone who has played the French almost exclusively v. 1.e4 for the past 30+ years, with pretty decent results both otb and in correspondence, here are some observations:
1. I agree with chuckv - the French has some pretty "dense theory" of its own.
2. White fashions change repeatedly over the years: I have seen most of my games be Winawer, then Tarrasch, then Advance, then Exchange. And now - at least at my level - many GK players try 2. Nf3 and some other less traveled roads.
So the French player has to be prepared for variations which lead to quite different positions and possibilties.
3. I still think Reuben Fine had the best idea about studying openings - see his "Ideas Behind the Chess Openings." If you understand the basic ideas and plans for both sides that lie behind whatever openings you play, you are much better off than memorizing variations.
mattdw 49 ( +1 | -1 )
"I always think of chess openings as a kind of " dance ", the opening moves do dictate to some extent the music you will be dancing to...probably well into the middle game. So, if you have a fiery Latin temperament...dont end up having to do a sedate waltz.
;-) "

I like that analogy! Except maybe it falls down in the middle game where dance starts becoming a little violent, introducing the aim of destroying your dancing partner...AKA ultimate dancing! ;)
bonsai 130 ( +1 | -1 )
Well, even if we take these statistics seriously and want to draw conclusions from them, then we have to be careful whether 1% difference in results may not just be random fluctuation.

I had a closer look at the latest 1900+ data. Using a simple Chi-square test for proportions (however appropriate that may be and of course in no way adjusting for e.g. the ratings of the players involved), I did compare the best scoring black opening (1...c5) against the others (and for good measure let's say that we use the Bonferroni adjustment for multiple testing).

black black
move score games p-value vs. 1...c5
1...c5 0.460 4555
1...e5 0.405 3667 <.0001 *
1...e6 0.455 1372 0.7444
1...c6 0.430 580 0.1696
1...d6 0.455 385 0.8500
1...Nf6 0.455 311 0.8640
1...d5 0.355 307 0.0002 *
* indicates significance at the overall 0.05 level after Bonferroni adjustment, displayed p-values are unadjusted.

However blatantly the analysis would be better if we used a model including the strength of the players involved in each game (or rather the difference in strengths, or maybe P{ N(0,...) > difference} if we take it that the assumed normal distribution for the Elo system works reasonably well).

Mmmh, I'm a bit surprised that 1...e5 looks so bad in this analysis. But anyway, the question that I am really answering is not what is the best answer to 1.e4, but rather what has in the past been the most successful answer for gameknot players rater above 1900.
mattdw 38 ( +1 | -1 )
What exactly is the null hypothesis in this case? Is it that the true proportion of white to black wins would remain the same if move...x was made? I'm not exactly great at this area of maths (as is probably clear ;) ) I have only ever used Chi-squared tests in two way/contigency tables (& bonferroni in ANOVA) and only very briefly at that.
bonsai 23 ( +1 | -1 )
Null hypothesis would be that the proportions the score that black gets is equal between the two moves being compared - tested against the two-sided alternative that one proportion is greater than the other.
i_play_slowly 23 ( +1 | -1 )
bonsai I didn't understand any of your post, although I am intrigued. What exactly is meant by:
*
1...c5 0.460 4555
1...e5 0.405 3667 <.0001 *
1...e6 0.455 1372 0.7444
1...c6 0.430 580 0.1696
1...d6 0.455 385 0.8500
1...Nf6 0.455 311 0.8640
1...d5 0.355 307 0.0002 *
*
Please be patient with me.
*
Respectfully,
*
i_play_slowly
anaxagoras 37 ( +1 | -1 )
Very nice work, but it suffers from (at least) two fatal flaws: Lack of significant sample size for statistical significance, and no consideration of the relative ratings of the players. Anecdotally, deciding that 1...g5 is a good reply to 1 e4 because of its winrate here at Gameknot is laughable. What a waste. Focus your efforts on endgames and tactics, not opening fashions.
mattdw 31 ( +1 | -1 )
Bonsai, I just noticed on your profile that you are a statistician for a pharmaceuticals company, that would be quite similar to the medical statistics module I have began studying at University probably wouldn't it? I was just thinking it has so far been one of the more interesting modules I have done (or am doing).
i_play_slowly 113 ( +1 | -1 )
anaxagoras I agree completely that 1...g5 is a laughable move, hence my conclusion, "1...g5 and 1...Nc6 also do more poorly [in the +1900 database], perhaps because they are basically unsound openings anyway." Later, when I suggest, "Perhaps +1500 players should avoid 1...Nc6 and 1...g6," I have made a typo: '1...Nc6 and 1...g5' is meant, echoing the previous statement.
*
I also agree that focusing on tactics and endgames is a sure road to chess mastery, and that too much attention is often paid to openings. However, nowhere above did I insist that people pay too little attention to openings, or that openings is a topic well worth obsessing over. I am left wondering if your post is contradicting statements that I never made. Recently I even suggested in the forum at the Chess Coaching Club, "Learn general opening principles, but don't get bogged down in deep studies of any particular opening."
*
Would you agree in response that the topic of openings is not utterly abhorrent, and that a person might satisfy a little curiosity in that area without growing hair on the palms of his hands?
bonsai 76 ( +1 | -1 )
anaxagoras: Yes, you are right about the sample size being too small to say much about g5, although comparing 1...g5 to 1...b6 actually has a fairly small p-value (0.0114 - but the problem is of course that I was just going through all possible moves to compare it to find one that looked interesting and didn't do anything to account for this "fishing for p-values").

And, yes, one of the real other problems is that I don't account for the involved ratings, so if Kasparov playing anonymously had scored 51.5% in 71 1...g5 games (most recent database numbers) on gameknot against some average players, then that would in my mind actually mean that the move is really, really bad (which I think it is, but it's not that bad, Kasparov win with almost(?) any first move against almost all of us).
mattdw 62 ( +1 | -1 )
I_Play_slowly, I don't think anyone means to accuse you of wasting your time of having incorrect priorities, it's just that this is a discussion forum and when you begin a discussion on something a bit controversial you should expect many people to 'disagree' with you or more specifically in this case with what you said (even if you never actually proposed to change your game due to your findings) it would be boring if people who were opposed to the matter didn't join in and express their opinions on the matter, otherwise you probably would have had 10 responses saying, "Yes, I agree." ;)
i_play_slowly 141 ( +1 | -1 )
mattdw I believe that you are correct in saying that I should expect disagreement when I introduce a controversial subject to the forum. By the same principle, I am free to disagree with any particular disagreement--no harm in dialectics. As happens so many times in this forum, we begin playing chess without a board.
*
Unfortunately, there are no rules for verbal chess. I would like to propose two. Firstly, I would expect that the reader focuses her disagreement on what I have actually said, and not on her own assumptions. In this case, as I never urged readers to focus on opening fashions, nor claimed that I was doing so, it's perhaps 'against the rules' to issue an imperative, "Focus your efforts on endgames and tactics, not opening fashions." By way of clarification, let me own that I do have an interest in openings, but, as my postings over time would indicate, my interests in chess are encyclopedic.
*
Secondly, and perhaps I broke this rule myself, I would hope that the reader can express disagreement without resorting to condescension, e.g., "laughable" and "What a waste." When someone responds to my post with an expression like "What a waste," I do reckon I am being accused of wasting my time.
*
I appreciate the altogether respectful tone of your posts to everyone here. If I have failed to match your tone in my dialectics, I should sharpen my own verbal chess skills.
*
Respectfully,
*
i_play_slowly
leo_london 229 ( +1 | -1 )
i_play_slowly I thought your original post was interesting, and viewed it in the manner I believe you intended...just a post to encourage discussion, not a recommendation of how to play !
To expand further, I list some opening stats that most people have probably already seen. I didnt waste much time, and again it is posted purely for the purpose of discussion.


White's Best 10 Openings (Based on White Win % minus Black Win %)

Position Opening Difference %
1 Queen's Gambit +16
=2 English Opening +14
=2 King's Indian Attack +14
4 Ruy Lopez +13
=5 Four Knight's Defence +12
=5 Giucco Piano +12
=5 Ponziani's Opening +12
8 Bishop's Opening +11
=9 Benko Opening +9
=9 Blackmar Diemer Gambit +9

White's Worst 10 Openings (Based on White Win % minus Black Win %)

Position Opening Difference %
1 Dresden Variation -56
2 Alapin's Opening -42
3 Konstantinopolsky Opening -22
=4 Anderssen's Opening -18
=4 Van't Kruijs Opening -18
6 Mason's Variation -17
7 Saragossa Opening -15
8 Grob Opening -13
9 Deprez Opening -9
10 Bird's Opening -4

Black's Best 10 Openings (Based on Black Win % minus White Win %)

Position Opening Difference %
=1 Nimzo Indian -3
=1 Pirc Defence -3
=1 Sicilian Defence -3
4 French Defence -5
=5 Alekhine Defence -6
=5 Robatsch Defence -6
=7 Fred Defence -7
=7 Nimzowitsch Defence -7
9 Rat -8
10 Caro-Kann -9

Black's Worst 10 Openings (Based on Black Win % minus White Win %)

Position Opening Difference %
1 Damiano Defence -67
2 Lasker Simul Special -43
=3 Benoni Defence -23
=3 Gunderam Defence -23
=3 Polish Defence -23
=3 St George's Defence -23
=7 Borg Defence -22
=7 Budapest Gambit -22
9 Dutch Defence -20
10 Philidor's Defence -19

10 Most Drawish Openings (Based on openings with the highest Draw %)

Position Opening Draw %
=1 Bogo-Indian Defence 40
=1 English Opening 40
=1 Four Knight's Defence 40
=1 Grunfeld Defence 40
=1 King's Indian Attack 40
6 Torre Attack 39
7 Reti's Opening 37
8 Petroff's Defence 36
=9 Nimzo Indian 35
=9 Three Knight's Defence 35

10 Least Drawish Openings (Based on openings with the lowest Draw %)

Position Opening Draw %
1 Damiano Defence 3
2 Borg Defence 10
3
Blackburne Gambit 12
=4 Blackmar Diemer Gambit 15
=4 Grob Opening 15
6 St George's Defence 17
7 Konstantinopolsky Opening 18
8 Mason's Variation 19
9 King's Gambit 20
=10 Centre Game 21
=10 Fred Defence 21
=10 Polish Defence 21



i_play_slowly 103 ( +1 | -1 )
leo_london Thanks for your understanding response. I remember seeing those stats a few years back at: -> homepage.ntlworld.com. It's a great website, kind of like a chess almanac.
*
Encouraged by the numbers there, I learned the Nimzowitsch Defence--and lost horribly, game after game. As peppe_l observes, it is indeed a mistake to choose moves based on statistics. Encouraged by the same list, however, I also began playing the Queen's Gambit, with much better results.
*
The author of that website provides a link to Chesslab.com, so you can corroborate his data. I was surprised at how well the Rat does overall, although it also appears to have lost the last >190 games that were saved to the database. Rat fans, your days are numbered!
*
If anyone can explain bonsai's intriguing stats (above), I would be very grateful--perhaps I am not the only person who does not understand them.
bonsai 60 ( +1 | -1 )
I'll explain: Basically when pasting them above you forgot the column headers, first column is "black move" (=i.e. the black reply to 1.e4 for which you have the stats in a row), the next column is "black score" (=how well black scores, e.g. 0.460 = 46% of the possible points), then "games" (=number of games with that move) and finally "p-value vs. 1...c5" (=p-value that the move in that row is either superior or inferior than 1...c5, note there's nothing in this column in the 1...c5 row). Admittedly the table is fairly difficult to read, because I didn't manage to have any proper spacing.
i_play_slowly 88 ( +1 | -1 )
bonsai--I'm almost there! Thank you for your generosity and your patience. I have been trying to understand your stats by replicating them--no luck! If you could please just bear with me one more time...
*
For example, 1...d5 was played in 307 games. 17% of these games resulted in victories for Black, and an additional 37% of the games resulted in draws. In other words, 52.19 of these games resulted in victories for Black, and an additional 113.59 games resulted in draws. If 1 point was awarded for each victory and a half-point was awarded for each draw, Black won 52.19 points for its victories and 57.79 points for its draws, or 109.98 points in total. If Black had won 100% of the games, it would have earned 307 points. Consequently, Black earned 109.98 points out of a possible 307 points. In other words, Black earned 27.9% of the points it could have earned. Shouldn't the score for 1...d5 be 0.279? Where have I gone wrong?
*
Respectfully,
*
i_play_slowly

bonsai 8 ( +1 | -1 )
You made a typo when you added the wins and draws, it should be 108.985 out of 307 points (=35.5% ; 17 + 37 / 2 = 17 + 18.5 = 35.5).
chessplayer20 46 ( +1 | -1 )
well i think datebase what we got here dose not give info about the type of player like a french player and how well the french is going i mean expert french players not somone who just plays random opening if you know what i mean like dose database add non french player and french players that is qustion is wear they get data from not just everyone playing french