Queen's gambit declined

I find very difficult to play for a win in the QGD. Is it a drawish opening in your opinion?

Yes, black has only 20% of wins in statistics of master games. But it's not drawish for white: the white's sharpness (=ratio of white's wins to draws) sh1=0.83 is quite decent. In conclusion, it's good for white and bad for black.

if it is bad for black, why it is recommended by strong coaches?

I don't know. Show the statistics of master games to those coaches. Statistically, the whole Queen's Gambit is bad for black.

Put simply, it's because statistics don't matter in individual games and because 1. d4 d5 is as sound as it gets.

Yeah, with 20% of black wins in 50000 master games it's as sound as buttocks.

In "individual games" U can win with 1. Nh3 too. It doesn't mean that the Ammonia attack is sound.

@Yigor: But statistics, established theory and common sense say that 1. Nh3 is a crap opening, whereas with 1. d4 d5, statistics (or at least your statistics) say it's subpar, but established theory and common sense say it's rock-sound, and I think I can prove your numbers wrong.

According to the Chesstempo database for 2200+ vs 2200+, after 1. d4 d5 2. c4 the distribution of results is 35.8%W-44.1%D-20.1%B. In other words, White scores 58%.

Then sh1=0.812 and sh2=0.456, so sev=+0.781.

However, if we jack the winning percentage of each side by 5% and reduce the amount of draws by 10%, then we get a hypothetical sharper opening with stats 40.8%W-34.1%D-30.1%B. Here White is still scoring 58%, but sh1=1.196 and sh2=0.883, so sev=+0.354.

This shows that the equation for sev is heavily biased toward sharper openings, therefore calling your entire concept into question.

EDIT: The Black result for the "sharper" opening is 25.1%. Therefore sh2=0.736, so sev=+0.625. So admittedly your eval function is (mostly) consistent. But still, if we "sharpened" the opening by 5% again, we would get something which is still scoring 58%, but should have a sev of around +0.475. And 24% draws is still a realistic figure for an opening.

This is why the concept fails: it basically knocks solid openings down a quarter of a pawn for no reason. Why? Well, if sh1 = W/D and sh2 = B/D and sev = sh1/sh2 - 1, then sev = W/B - 1. If White wins a% more than Black, then sev = (B+a)/B - 1 = 1 + a/B - 1 = a/B. 

Then White winning a% more than Black will produce results (B + a)W, (100 - (2B + a))D, BB. White's score then equals (B + a) + (100 - (2B + a))/2 = (100 + a)/2 = 50 + a/2.

We have a problem now: All openings with equal a will score equally well or poorly, but the sev of an opening with a certain value of a and therefore a certain score is inversely proportional to the percentage of wins for the weaker side. This means that solid openings with a high draw percentage are penalized heavily.

@Phoenyx75: Sorry, my post was aimed at @Yigor. Your method is by far the best one. It's numerically perfect and actually means something as opposed to the wonky and inaccurate "statistical evaluations".

Tbh I didn't really have a problem with it until I realized that he seemed to believe it over 150 years of established theory, and that the index itself was seriously flawed.

"the index" was basically a word for "sev", but whatever. As for your second argument, I would argue that part of the reason the Ruy and Najdorf, for example, do so well for both sides is because the theory is highly developed, as opposed to a less developed opening where both sides will make more mistakes, which on balance should benefit White as if White makes a mistake in the opening at high levels the position is around equal while if Black makes a mistake they will be worse.

It's not a problem, it's how sev is defined. When W>B, on a indeed sev = W/B - 1 = (W-B)/B and sev is inversely proportional to B. If U increase additively both W and B considering W+x and B+x, sev will diminish. It's normal cuz it expresses the multiplicative ratio.

Example. W=30%, B=20%, D=50% => sev=+0.5, sh1=0.6, sh2=0.4 while adding x=10% we obtain W=40%, B=30%, D=30% => sev=+0.33, sh1=1.33, sh2=1. So what? It's normal. It slightly diminished the statistical evaluation but increased the sharpness. That's why I indicate also the sharpness factor when sh1 and/or sh2 are unusually high or low.

The sev is defined as (W-B)/B when W>B in order to be compatible to engine evaluations. As I've shown in dozens of examples in my posts, it's not flawed at all and gives reasonable results.

In addition, nobody forbids U to use both sev and simple W-B, if U prefer it as Phoenyx75, as well as any other derived statistical values.

However, chesster3145 distracted us from the original question. In the case of QGD, it's bad for black not because of any definition of sev but cuz B=20%. The percentage of black wins is too low.

Btw chesster3571 has made a huge mathematical discovery: positive additive factors diminish multiplicative ratios w/b when w>b [(w+x)/(b+x)< w/b when w>b]. wow  LMAO  But it's not a flaw, the elementary arithmetics of our world just work like that.

Phoenyx75: Personally, at the moment, I prefer to use sev as well as sh1 and sh2. But, of course, U can use any other derived values that are convenient for U. For instance, when W>B, sev=(W-B)/B so, W-B=B x sev. The conversion is easy.

Phoenyx75: There are no sev in databases. I calculate it myself using chesstempo data. So, instead of converting sev, just go to the database and use the original W/B/D statistics.

StupidGM: I'm also posting test lines for all 960 Fischer's positions :

https://www.chess.com/forum/view/chess960-chess-variants/list-of-all-960-fischer-s-positions-evaluations-and-best-moves

What is the best variation for Black in the QGD to score a win?

Great! It sounds interesting.