Stockfish and Statistical Evaluation of the Indian Game

Phoenyx75

I put Stockfish in the title in the hopes that those who lend Stockfish (or other chess engines) no credence would hopefully bypass this thread entirely. It's been a while since I always believed that Stockfish's evaluations were state of the art in the opening, which is why, for example, I continued to believe that the Indian Game was the best response to 1.d4 despite the fact that lichess.org's Stockfish engine decided that Queen's Pawn's Game (1...d5) was the best response at a depth of 43.

 

In essence, I've generally come to believe that if Masters play 2 openings a fair amount and tend to do better with one of them, that opening is probably better, and trumps Stockfish's evaluation. Still, I thought, perhaps Stockfish simply needed a bit more time to determine what was the best opening. So I left Stockfish 9 calculating overnight for around 6 1/2 hours to see if it would come to a different evaluation if it went further then depth 43. It turns out, it did- after around 6 1/2 hours, and reaching a depth of 47, it determined that the Indian Game was the best option after all, with an evaluation of 0.00. 

 

Anyway, would be appreciated if any comments or criticisms were of the constructive variety.

ThrillerFan

You aren't going to get the critics to stay away just because of your title.

 

Stockfish and all other engines are horrible at openings.  This is why you have to invest in add-ons like the Opening book when using an engine.  Engine evaluations as a whole are total cr@p when it comes to openings.

 

And there is no such thing as the "Indian Game".  There is the Nimzo-Indian Defense, King's Indian Defense, Bogo-Indian Defense, Queen's Indian Defense, Old Indian Defense, but there is no "Indian Game".

 

In fact, 1...Nf6 does NOT even guarantee an opening with the word "Indian" in it's name:

 

1.d4 Nf6 2.c4 e6 3.Nc3 d5 - Queen's Gambit Declined

1.d4 Nf6 2.c4 g6 3.Nc3 d5 - Grunfeld Defense

1.d4 Nf6 2.c4 c6 3.Nc3 d5 - Slav Defense

1.d4 Nf6 2.Nf3 e6 3.Bg5 - Torre Attack (2...g6 3.Bg5 is also a Torre Attack)

1.d4 Nf6 2.Nf3 g6 3.Bf4 - London System (one of many variations)

 

 

So once again, you aren't going to take it as constructive criticism because you always think you are right, but I am telling you the truth.  Quit using bots to try to evaluate the first move of any opening.  After 1.d4, the moves 1...d5, 1...Nf6, 1...e6, 1...c6 (the latter two assume that you are ok with the French or Caro-Kann, and a 2.c4 response will see Black then control e4 with 2...d5 or 2...Nf6) are all legit, as is 1...f5.  The moves 1...c5, 1...d6 and 1...g6 are slightly riskier, but there is no refutation.

 

The rest are pretty much garbage:  1.d4 e5? 2.dxe5!, 1.d4 b6? 2.e4!, 1.d4 a6? 2.e4!, 1.d4 Nc6?! 2.Nf3!

Phoenyx75
ThrillerFan wrote:

You aren't going to get the critics to stay away just because of your title.

 

Stockfish and all other engines are horrible at openings.  This is why you have to invest in add-ons like the Opening book when using an engine.  Engine evaluations as a whole are total cr@p when it comes to openings.

 

And there is no such thing as the "Indian Game".  There is the Nimzo-Indian Defense, King's Indian Defense, Bogo-Indian Defense, Queen's Indian Defense, Old Indian Defense, but there is no "Indian Game".

 

In fact, 1...Nf6 does NOT even guarantee an opening with the word "Indian" in it's name:

 

1.d4 Nf6 2.c4 e6 3.Nc3 d5 - Queen's Gambit Declined

1.d4 Nf6 2.c4 g6 3.Nc3 d5 - Grunfeld Defense

1.d4 Nf6 2.c4 c6 3.Nc3 d5 - Slav Defense

1.d4 Nf6 2.Nf3 e6 3.Bg5 - Torre Attack (2...g6 3.Bg5 is also a Torre Attack)

1.d4 Nf6 2.Nf3 g6 3.Bf4 - London System (one of many variations)

 

 

So once again, you aren't going to take it as constructive criticism because you always think you are right, but I am telling you the truth.  Quit using bots to try to evaluate the first move of any opening.  After 1.d4, the moves 1...d5, 1...Nf6, 1...e6, 1...c6 (the latter two assume that you are ok with the French or Caro-Kann, and a 2.c4 response will see Black then control e4 with 2...d5 or 2...Nf6) are all legit, as is 1...f5.  The moves 1...c5, 1...d6 and 1...g6 are slightly riskier, but there is no refutation.

 

The rest are pretty much garbage:  1.d4 e5? 2.dxe5!, 1.d4 b6? 2.e4!, 1.d4 a6? 2.e4!, 1.d4 Nc6?! 2.Nf3!

 

I'm fine with criticism, as long as it's relatively civil. I think you passed the test this time around. As to engine openings as a whole being "total cr@p", I haven't come to that conclusion myself. At a depth of 43, Stockfish thinks that 1...d5 is the best response to 1.d4. At a depth of 47, it decides that 1...Nf6 is better. Both are recognized to be solid responses to 1.d4. As to 1.d4 Nf6 being called the Indian Game, I'm simply taking it from chess.com's naming of the position. You don't like it, take it up with them -.- I'm fully aware that once more moves are added to the mix the name can change.

DeirdreSkye

One more scientific article from the master of Stockfish! 

Needless to say that everything you say just betrays once more that you actually understand nothing!

    Does it matter if 1...Nf6 is better than 1...d5 by 0.01 when you understand nothing?

You actually have absolutely no idea. You can't even tell when a side has a  important plus that is more than 200 times bigger!       

 

     

So your game fluctuates more than 3.00(a whole piece) and that at your best(in another game you needed 6 moves to go from +0.3 to -6.9) and you are unable to evaluate a position in which you are +2.5 ahead(250 times bigger than the mere 0.01) but you claim it will be beneficial for you to choose 1...Nf6 just because it is +0.01 better than 1...d5. 

Let me ask you again.Do you even know what this +0.01 means?Can you explain it?

Can you explain to a total beginner why the position after 1.d4 Nf6 is better for Black than the position after 1.d4 d5?Please enlighten us. 

DeirdreSkye
BobbyTalparov wrote:

@DierdreSkye While he is researching for that essay, he should also state what the margin of error for Stockfish's minimax function is, and what that means.

    I have said that in other posts. Stockfish evaluations are the result of human parameters and they might change in the next version. Basically , on the first moves Stockfish "understands" less than humans but he doesn't accept that.For him Stockfish is the result of divine virgin birth , there is no human hand involved in it's creation. He thinks  Terminator was a documentary. Someone must tell him Skynet doesn't exist ,it was a movie and machines can't reproduce themselves.At least not yet.

Strangemover

I don't understand the point of attempting to use Stockfish to determine the best opening moves. You are an amateur player, not a super GM. The joy of being an amateur player for me is that you can play whatever you want because you play purely for fun and recreation. Ok it's nicer to win, but if you're rated 1500 and are playing others of similar level for funzies, what difference does it make what a 3500+ rated machine says is the best response to 1.d4 or whatever?

Phoenyx75
Strangemover wrote:

I don't understand the point of attempting to use Stockfish to determine the best opening moves. You are an amateur player, not a super GM. The joy of being an amateur player for me is that you can play whatever you want because you play purely for fun and recreation. Ok it's nicer to win, but if you're rated 1500 and are playing others of similar level for funzies, what difference does it make what a 3500+ rated machine says is the best response to 1.d4 or whatever?

 

You actually suggested the answer to your own first question. You stated "The joy of being an amateur player for me [emphasis mine] is that you can play whatever you want because you play purely for fun and recreation."

 

That is what -you- find nice. I personally like to be able to play very well for the very first moves at least. After that, it's much harder- I can't look at databases or the like, just have to come up with moves on my own. I do my best, but I'm fully aware that my best is frequently a long ways off from Master moves, as people like Deidre like reminding me.

Strangemover

Fair enough, different strokes for different folks.

Phoenyx75
DeirdreSkye wrote:
BobbyTalparov wrote:

@DierdreSkye While he is researching for that essay, he should also state what the margin of error for Stockfish's minimax function is, and what that means.

    I have said that in other posts. Stockfish evaluations are the result of human parameters and they might change in the next version. Basically , on the first moves Stockfish "understands" less than humans but he doesn't accept that.For him Stockfish is the result of divine virgin birth , there is no human hand involved in it's creation. He thinks  Terminator was a documentary. Someone must tell him Skynet doesn't exist ,it was a movie and machines can't reproduce themselves.At least not yet.

 

Sigh -.-. No, I don't think that Stockfish is perfect. As a matter of fact, I now have several opening moves that I have labelled "Exception". As in, I'm not following Stockfish's choice because I don't think it's the right one. But I label them exceptions because they are exceptions to the rule. For the most part, I have found Stockfish's choices to be state of the art. Now, this may be in part because I am an amateur. But I think most would agree that while Stockfish may not always come up with the best opening moves, it will at least stay away from bad moves. And I also find it noteworthy when it just happens to hit on opening moves that Masters do best with, statistically speaking, such as 1...Nf6 in response to 1.d4 (aka Indian Game). 

Phoenyx75
DeirdreSkye wrote:

One more scientific article from the master of Stockfish! 

Needless to say that everything you say just betrays once more that you actually understand nothing!

    Does it matter if 1...Nf6 is better than 1...d5 by 0.01 when you understand nothing?

You actually have absolutely no idea. You can't even tell when a side has a  important plus that is more than 200 times bigger!       

 

     

So your game fluctuates more than 3.00(a whole piece) and that at your best(in another game you needed 6 moves to go from +0.3 to -6.9) and you are unable to evaluate a position in which you are +2.5 ahead(250 times bigger than the mere 0.01) but you claim it will be beneficial for you to choose 1...Nf6 just because it is +0.01 better than 1...d5. 

Let me ask you again.Do you even know what this +0.01 means?Can you explain it?

Can you explain to a total beginner why the position after 1.d4 Nf6 is better for Black than the position after 1.d4 d5?Please enlighten us. 

 

I suspect you are asking questions you already know the answer to. No, I don't know why Masters win more often when they play 1...Nf6 as opposed to 1...d5. Perhaps there is no one who does. But you don't need to know why a move tends to win more to take advantage of the fact.

 

As to my games, as you can see, I'm an amateur. Looking at my rating should be able to tell you that. In the game of mine you bring up, it looks like I started out well in the game, probably because I was using my mainly Stockfish compiled database of moves, then apparently messed up because the Stockfish database moves ran out, and then apparently salvaged the situation without being aware of it and thought things weren't going in my favour and resigned. There is nothing all that surprising there.

 

I've said it before and it looks like it bears repeating- my main focus is the opening. Ofcourse I would like to win more games, but what's easy is to simply look at my openings and see "what would stockfish have done?" and add a move or 2 to my database. That way, I won't make the same opening mistake twice. 

 

I also frequently take the 10 lessons a day I get daily from chess.com with my subscription and I think that teaches me something valuable as well.

ThrillerFan
BobbyTalparov wrote:
ThrillerFan wrote:

You aren't going to get the critics to stay away just because of your title.

 

Stockfish and all other engines are horrible at openings.  This is why you have to invest in add-ons like the Opening book when using an engine.  Engine evaluations as a whole are total cr@p when it comes to openings.

 

And there is no such thing as the "Indian Game".  There is the Nimzo-Indian Defense, King's Indian Defense, Bogo-Indian Defense, Queen's Indian Defense, Old Indian Defense, but there is no "Indian Game".

 

In fact, 1...Nf6 does NOT even guarantee an opening with the word "Indian" in it's name:

 

1.d4 Nf6 2.c4 e6 3.Nc3 d5 - Queen's Gambit Declined

1.d4 Nf6 2.c4 g6 3.Nc3 d5 - Grunfeld Defense

1.d4 Nf6 2.c4 c6 3.Nc3 d5 - Slav Defense

1.d4 Nf6 2.Nf3 e6 3.Bg5 - Torre Attack (2...g6 3.Bg5 is also a Torre Attack)

1.d4 Nf6 2.Nf3 g6 3.Bf4 - London System (one of many variations)

 

 

So once again, you aren't going to take it as constructive criticism because you always think you are right, but I am telling you the truth.  Quit using bots to try to evaluate the first move of any opening.  After 1.d4, the moves 1...d5, 1...Nf6, 1...e6, 1...c6 (the latter two assume that you are ok with the French or Caro-Kann, and a 2.c4 response will see Black then control e4 with 2...d5 or 2...Nf6) are all legit, as is 1...f5.  The moves 1...c5, 1...d6 and 1...g6 are slightly riskier, but there is no refutation.

 

The rest are pretty much garbage:  1.d4 e5? 2.dxe5!, 1.d4 b6? 2.e4!, 1.d4 a6? 2.e4!, 1.d4 Nc6?! 2.Nf3!

While I agree with just about everything here, there is one point of clarification: 1. d4 b6 2. e4 e6 transposes into the Owen Defense (a side line of the French). While I am not a fan of it, there is no direct refutation.

 

As I posted in another threat on unorthodox openings, when a player argued that Owen's Defense is better than the English Defense, I highly beg to differ.  His claim was that after 1.d4 that 1...b6 was the best way to enter the English Defense as it avoids lines with Bd3 and I pointed out that transposing to Owen's Defense and not playing 2.c4 was significantly stronger for White, and the line I play against it is a large advantage for White.

 

1.d4 b6?! 2.e4 (You are about to see why I say the English Defense "might" be sound, but Owen's Defense is not!) Bb7 3.Bd3! e6 (3...f5 is crap!  4.exf5! Bxg2 5.Qh5+ g6 6.fxg6 Bg7 (6...Nf6 7.gxh7+!! Nxh5 8.Bg6#) 7.gxh7+ and White's winning) 4.Nf3 answering 4...Nf6 with 5.Qe2 and 4...c5 with 5.c3! (A move that can no longer be played in the English Defense because the pawn is already on c4, weakening d4).

drmrboss

 As long as you do 30+ ply depth, as long as 12 moves opening ends within 0.25 evaluation ( with stockfish 9 it is 0.37 evaluation cos of 0.12 contempt), all those variations are fine. Just choose whatever variation you like.

 

drmrboss

There may be some variations with 0.50+ and still playable, but exceptional.

Phoenyx75
drmrboss wrote:

 As long as you do 30+ ply depth, as long as 12 moves opening ends within 0.25 evaluation ( with stockfish 9 it is 0.37 evaluation cos of 0.12 contempt), all those variations are fine. Just choose whatever variation you like.

 

 

Playable, sure, but I'm looking for the best opening moves. Yes, I know, I know, it's not really possible to prove which moves are best, or even if there are best moves instead of a bunch that are equally good at the opening, but I do the best I can with the information I have.

drmrboss

There is no best opening move. Because chess is draw. As long as either side play perfect, no one can win.

Choose the opening where " you have the lowest chance to make mistake, that will be the best opening for you."

 

Remember, chess is like marathon, you or your opponent will likely do mistake, and catch that opportunity timely to win.

drmrboss

Even if engines show 3+ in opening, it may theoretically be possible draw ( cos you cant win your opponent with one extra minor piece, you need minimum of 5+ or a rook) to win.

A pawn can be assumed +9 in ending though.

drmrboss
DeirdreSkye wrote:

One more scientific article from the master of Stockfish! 

Needless to say that everything you say just betrays once more that you actually understand nothing!

    Does it matter if 1...Nf6 is better than 1...d5 by 0.01 when you understand nothing?

You actually have absolutely no idea. You can't even tell when a side has a  important plus that is more than 200 times bigger!       

 

     

So your game fluctuates more than 3.00(a whole piece) and that at your best(in another game you needed 6 moves to go from +0.3 to -6.9) and you are unable to evaluate a position in which you are +2.5 ahead(250 times bigger than the mere 0.01) but you claim it will be beneficial for you to choose 1...Nf6 just because it is +0.01 better than 1...d5. 

Let me ask you again.Do you even know what this +0.01 means?Can you explain it?

Can you explain to a total beginner why the position after 1.d4 Nf6 is better for Black than the position after 1.d4 d5?Please enlighten us. 

I am pretty sure there are scripts for stockfish that can give you explaination for 0.01+ in detail. For example. this is breakdown detail evaluation of stockfish for one position.eval_bot (41w) Total(-3.35,-3.14) Material(-1.23,-1.30) Imbalance(-0.28,-0.28) Initiative(0.00,0.02) Pawns(-0.36,-0.17) Knights(0.00,-0.10) Bishops(0.07,0.17) Rooks(-0.33,-0.14) Queens(0.00,0.00) Mobility(-0.73,-0.68) King(-0.12,-0.07) Threats(-0.26,-0.37) Passed(-0.11,-0.23) Space(0.00,0.00) <Stockfish>

ThrillerFan

The entire problem with what the OP is saying is that he is trying to use artificial intelligence to make human decisions at a point in the game where artificial intelligence is not very intelligent at all!

 

In many ways, chess is like sex.  Using a computer to make all the decisions for you as to what is considered the best position would be the equivalent to using a blow up doll to make all the decisions for you as to what is considered the best position!

ilusmte

There are no such thing as Indian Game ??????????????? OMG...can you even spell India or chess Mr. Thriller???? Check Google or Wikipedia or any opening book...... If you can read english then I am pretty sure you will find something known as Indian Game starting with Nf6

ilusmte

https://www.chess.com/openings/A45-Indian-Game