from a database I checked :
best variations to play against french defence


In my next life, I would play the French Defense as Black.
The only thing I do not like about the French is that White can get a fairly easy draw with the Exchange variation. [notice I said White]
But by the time I get to my next life, chess will be played out more and it will not be so bad to get a draw with Black.

Qe2 is a bad move in that white has an opportunity to play for an advantage after 3. Nc3 or even 3. e5 and decides to let black equalize instead. It's not terrible, but the fact that it's not best means that it should not be played IMO.
This reminds me of an interview with Karpov I read years ago . Karpov was asked how is it that he never seemed to be in time trouble while his opponents often were . His answer was that while many top players search for the best move in a position he is content to find a good move and play it quickly , even though it might not be the very best move objectively speaking .

In addition to the stats given by Reb and SmyslovFan there is a website which gives stats on the French Defense and about 80 other openings and variations of openings.
It gives the French Defense as 39% wins 34% loses 27% draws.
I think it is from 500,000 games since 1990 [for all openings]
It also gives top ten openings, bottom ten openings etc etc.
I typed "chess opening stats" on google to get it. czyzewski.org

Yeah, but that's ALL games. I think it's more telling what the GM practice is.
I agree but I showed the stats with all players so that your stats cant be so easily dismissed by the " but at the amateur level " crowd . I was surprised that the french is the third most popular against 1 e4 , I thought carokann was third and french 4th but I had it reversed .

Here are Blacks 5 best defenses per the web site I mentioned
1. Nimzo Indian
2. Pirc Defense
3. Sicilian Defense
5. French Defense
6. Alekhine Defense

Well I do not know what Database Smyslov is using but the one I am using differs from those stats.
I will show you the break down of 3 different ranking groups I will also include the difference between the white winning percentage and black winning percentage at the end in blue.
2700+ vs 2700+
out of __ amount of games White wins/draws/Black wins/Differ.
3.Nc3 249 games 33.7% / 47.4% / 18.9% / 14.8%
3.Nd2 74 games 40.5% / 40.5% / 18.9% / 21.6%
3.e5 28 games 53.6% / 25% / 21.4% / 32.2%
3.exd5 4 games 25% / 25% / 50% / -25%
2200+vs 2200+
out of __ amount of games White wins/draws/Black wins/Differ.
3.Nc3 38799 games 37.4% / 37.1% / 25.5% / 11.9%
3.Nd2 26385 games 36.6% / 41.1% / 22.3% / 14.3%
3.e5 8336 games 36.7% / 32.5% / 30.8% / 5.9%
3.exd5 4868 games 19.4% / 57.5% / 23.1% / -3.7%
All games
out of __ amount of games White wins/draws/Black wins/Differ.
LOL I got tired of retyping it so I posted a screen shot for the last one since it had alot of numbers. It doesnt show the difference though so I'll do that below
3.Nc3 12.7%
3.Nd2 14.6%
3.e5 2.7%
3.exd5 -8.7%
Actually I might add another screenshot of a higher level since the 2700 vs 2700 has short game samples maybe I'll do 2500 range higher samples their.
2500+ vs 2500+
out of __ amount of games White wins/draws/Black wins/Differ.
Difference
3.Nc3 13.9%
3.Nd2 14%
3.e5 14.6%
3.exd5 -0.8%
If the Difference between the wins and the losses are greater it means the chances that white will win is a higher among that range group.
1 thing that can be said about all of the samples 3.exd5 seems to be the worse. However, 3.Nc3 surely does not seem to be the best. Whether the sample sizes have been small or big 3.Nd2 has been dominating 3.Nc3. percentage wise in all area's. In fact 3.Nd2 and 3.e5 have been competing with each other in different sample sizes as well as level ranges.

Nc3 is the most fun to play and gives white a bigger chance for an advantage, hence why it's the most popular. The KIA is a fun game too but certainly not as respected.

An for the Millions and the Millions of the KIA fans out their I will show an
All game of the KIA against the French with possible 3rd move's.
3.Nd2 8.8%
3.Qe2 22.5%

what are sample sizes?
The line 1. e4 e6 2. d3 d5 3. Qe2 is much better for White than
1. e4 e6 2. Qe2
After 1. e4 e6 2. d3 Black should not play d5. Black should play c5

what are sample sizes?
The line 1. e4 e6 2. d3 d5 3. Qe2 is much better for White than
1. e4 e6 2. Qe2
After 1. e4 e6 2. d3 Black should not play d5. Black should play c5
The sample size of 3.Qe2 is 1469 games.
It is an old idea which has gotten a new spin I believe. The line move does come from the idea of 2.Qe2 I believe. It doesn't have an offical name as of yet other than the KIA that I'm aware of.
However, 2.Qe2 has a different name entirely.
1.e4 e6 2.Qe2 is considered the French, Chigorin Variation. So I mean I believe it has been around but not many play it that is why sample sizes are small but with the current sample size it does seem 2...c5 is blacks most played response.
Those lines I believe have recently been remade popular to add spin on things

what are sample sizes?
The line 1. e4 e6 2. d3 d5 3. Qe2 is much better for White than
1. e4 e6 2. Qe2
After 1. e4 e6 2. d3 Black should not play d5. Black should play c5
While I agree with this and I do often play 2... c5 a french player is not likely to play 2 ... c5 unless he is also a sicilian player , which I am .

Well I do not know what Database Smyslov is using but the one I am using differs from those stats.
I will show you the break down of 3 different ranking groups I will also include the difference between the white winning percentage and black winning percentage at the end in blue.
2700+ vs 2700+ out of __ amount of games White wins/draws/Black wins/Differ.
3.Nc3 249 games 33.7% / 47.4% / 18.9% / 14.8%
3.Nd2 74 games 40.5% / 40.5% / 18.9% / 21.6%
3.e5 28 games 53.6% / 25% / 21.4% / 32.2%
3.exd5 4 games 25% / 25% / 50% / -25%
2200+vs 2200+ out of __ amount of games White wins/draws/Black wins/Differ.
3.Nc3 38799 games 37.4% / 37.1% / 25.5% / 11.9%
3.Nd2 26385 games 36.6% / 41.1% / 22.3% / 14.3%
3.e5 8336 games 36.7% / 32.5% / 30.8% / 5.9%
3.exd5 4868 games 19.4% / 57.5% / 23.1% / -3.7%
All games out of __ amount of games White wins/draws/Black wins/Differ.
LOL I got tired of retyping it so I posted a screen shot for the last one since it had alot of numbers. It doesnt show the difference though so I'll do that below
3.Nc3 12.7%
3.Nd2 14.6%
3.e5 2.7%
3.exd5 -8.7%
Actually I might add another screenshot of a higher level since the 2700 vs 2700 has short game samples maybe I'll do 2500 range higher samples their.
2500+ vs 2500+ out of __ amount of games White wins/draws/Black wins/Differ.
Difference
3.Nc3 13.9%
3.Nd2 14%
3.e5 14.6%
3.exd5 -0.8%
If the Difference between the wins and the losses are greater it means the chances that white will win is a higher among that range group.
1 thing that can be said about all of the samples 3.exd5 seems to be the worse. However, 3.Nc3 surely does not seem to be the best. Whether the sample sizes have been small or big 3.Nd2 has been dominating 3.Nc3. percentage wise in all area's. In fact 3.Nd2 and 3.e5 have been competing with each other in different sample sizes as well as level ranges.
The "difference" he is talking about is the difference in winning %s between White and Black if you disregard the draws. Well, chess doesn't disregard draws.
So, for his +2700 table, translating it to the same numbers I posted:
There are 355 games mentioned.
3.Nc3: 249/355 games= ~70% of the games. White scored 57.4% (all the wins plus half the draws. Black scored the other half of the draws)
3.Nd2: ~21% White scored: 60.75%
3.e5: ~8% White scored: 66.1%
3.exd5: 1% White scored 37.5%
These numbers are very much in line with the numbers I posted earlier, except for the stats for 3.exd5 which is easily explained by the small number of games (4) played in that line.
A word on "sample size": This isn't actually a sample, it's 100% of games that fit the description. However, the number of games played with 3.exd5 is not statistically significant enough to draw conclusions about the soundness of the opening. This is a common problem with selecting very small groups. That's why I tend to use all GMs (+2500) rather than elite GMs (+2700) for most of my database searches.
I hope this helps clear things up.

I have consulted the One True Database™ and will issue the final words on this matter.
- Nc3 - 4820 57.1%
- Nd2 - 3088 58.2%
- e5 - 731 56.8%
- exd5 - 202 47%
Interesting note about the exchange variation; Black wins more than White.
White wins: 47 (=23%), Draws: 96 (=48%), Black wins: 59 (=29%)

If all you care about is winning, here are the best options:
- e5 White wins 38%
- Nc3 White wins 37%
- Nd2 White wins 36%

Who has a better record against the french defense than Karpov ? In 58 games as white he played 3 Nd2 33 times , winning 21 games and drawing 12 and losing none ! He played 3 Nc3 21 times , winning 12 , drawing 6 and losing 3 . He played the exchange 4 times , winning 3 and drawing 1 and no losses . Amazing record !

If all you care about is winning, here are the best options:
e5 White wins 38% Nc3 White wins 37% Nd2 White wins 36%Statistically, yes. But I have found that a very uncomfortable variation for black to play against is the Tarrasch.

Who has a better record against the french defense than Karpov ? In 58 games as white he played 3 Nd2 33 times , winning 21 games and drawing 12 and losing none ! He played 3 Nc3 21 times , winning 12 , drawing 6 and losing 3 . He played the exchange 4 times , winning 3 and drawing 1 and no losses . Amazing record !
Naidistch:
29 wins, 14 draws, 2 losses. +152 elo performance over expected.
Btw, the higher the ratings, the better White's score. This suggests something that Carlsen said: 2...e6 may be a second-class opening at the GM level.
Here's the stats for games between +2600s:
3.Nc3: ~61% of games, 58.9% score for White
3.Nd2: ~29%; 57.2%
3.e5: ~8%; 62.3%
3.ed5: 2%; 55.3%