Scotch Four Knights... What would you do after this move?

The current revolution in chess, as I see it, is largely driven by computer engines. All sorts of moves previously screened out by "master intuition" are now up for grabs.

pfren went on about Giri playing 4.h3 a move earlier in this Four Knights . However many hours later, my Stockfish-6 now has 5.h3 as the first choice in our current line. I wouldn't be surprsied if Giri got the idea for 4.h3 by letting an engine run for a day or two.

Right now, according to Stockfish, White can play h3, a3, Be2, Be3, d5, dxe5, Bb5 and they all converge within 80 centipawns. GM practice, should fashion allow, might tell a different story, but I really wouldn't be surprised if 5.d5 works to a mild White advantage as well as the other options.

As ThrillerFan said, 5.d5 gains space. It kicks the c6 knight and helps block the f8 bishop. Sure Black has an f5 break, but it's all trade-offs. Who would say White's position after 5.d5 is bad?

I would hate to agree with Pfren.

However, I just don't know what ThrillerFan is thinking with his 5.d5 pawn push.

I do not agree with Pfren's statement "positional blindness".

Not yet any way.

I haven't heard of ThrillerFan follow up!

He might not be positionally blind.

He might be creative!

If Thrillerfan says he is going to play on the queenside here than I call him positionally blind.

If Thrillerfan says he is not going to play on the queenside than I call him creative!

In a normal position after the move 5.d5.

The center becomes locked.

Both sides would than follow a pawn pointing strategy.

The black d6 + e5 pawn are pointing to the kingside.

Black will continue playing kingside pawn pushes.

The next black pawn break will be with f5.

In response white would than follow his pawns.

The e4 + d5 pawn are pointing to the queen side.

White will continue playing queenside pawn pushes

The next white pawn break will be with c5.

However, here is the catch in this position!

If ThrillerFan adopts the normal position strategy I said above than he is positionally blind because he should know he will lose that race!

The reason why he will lose the race is because he has already commited his knight on c3.

Thriller Fan has the knight on c3 before the pawn push d5.

Which means after the pawn push d5 Thriller Fan will have to move the knight on c3!

Than play the pawn push c4!

Than play his knight back to c3.

Which means he is positionally blind because he is wasting to much time.

So in other words he can no longer do this strategy any more.

He has already commited himself to a knight on c3 that queen side vs king side race will cause him to lose.

If he wanted to do this queen side vs king side race he should of played d5 sooner!

It is to late now!

So now lets get creative!

We can't win in this arms race so were do we play?

We can't play on the queen side because we would lose the race by our faulty piece placement.

We can't play in the center because it is locked.

The only place we have left to play is on the kingside.

However, black is playing on the kingside!

How do you play on the kingside when black is playing on the kingside?

The creative juices are flowing!

Well work backwards!

What is black doing?

Black is going to play f5 than he is going to push forward with his pawns to continue attacking the kingside!

If we can't beat black in the queenside vs kingside race than we have to use blacks own kingside momentum against him!

How do we do it?

Eureka! What if white waits until black castles kingside.

Than encourages black to play on the kingside!

Once this begins happening white than can castle queenside and place to 2 fat rooks on the G and H file to meet black's king side pawn storm?

It seems like a creative idea to me.

If your opponent wants to run at you with spears.

Set up 2 machine guns on the G and H file to meet them.

Why should we discourage this from happening.

Well here's Stockfish-6 after running all night on the Four Knights position:

Stockfish shows 5.d5 at the top by one centipawn and several other reasonable looking moves within nine centipawns, which is insignificant.

My point is not that 5.d5 must be the best move -- hardly -- just that it's a playable move not to be dismissed as "positional blindness."

This "positional blindness" charge likely reflects another blind spot in current master inuition which could change in a heartbeat if some hotshot player starts burning the shirts off his opponents with the line.

How many times in chess have we seen the old conventional wisdom turned on its head like this?

All I can see is that the silicon patzer suggests 5.d5 Nb8 as best (by a fairly insignificant margin). I also happen to know that good players rarely play 5.d5.

Of course 5.d5 Ne7 is the way to play, when there are very few games between strong players: one OTB blitz Grandmaster game, and three correspondence ones between masters. It should come as no surprise that white is scoring slightly less than average. Here are the "serious" games, patzer ones filtered out:

But anyway...time for yet another useless screenshot.

You can see 5.Bb5 as joint first choice, and 5.d5 Ne7 as fourth. All that mean... err, nothing at all.

pfren: For goodness sake. Have you learned nothing from chess history? Have you not noticed that the conventional wisdom of one generation of masters is often overthrown by the next? Lines that weren't played ten or twenty years before suddenly become all the rage?

There is much groupthink going on with grandmasters and for good reason. It's safer to go play what other players than branch out on your own, unless you have really done your homework. That's what the leaders do. Top level players like Fischer, Kasparov, Kramnik. They'll play lines other people won't, succeed, and then everyone else files in behind.

Anyway, to say few GMs are playing something is interesting, but far from conclusive.

We try finding NEW opening concepts at official correspondence chess, since the theoretical lines are over-analysed. Actually this is the only interesting part of modern correspondence... the playing part is rather boring- engines either fall into not-so-cheap tricks (it's becoming rarer) or play sub-par in the ending, where you can take advantage.

But heck, I can safely bet that 5.d5 is not an interesting one...

But heck, I can safely bet that 5.d5 is not an interesting one...

pfren: So now you're moving the goal posts.

You called 5.d5 "positional blindness." If that's a serious charge, it means White should pay a price for being positionally blind, as in thoughtlessly accepting doubled pawns without compensation.

Now you are saying 5.d5 is merely "not interesting," which is something else entirely.

I'm sure if it was 1999, before Kramnik unveiled his Berlin Wall against Kasparov, you would be lecturing us how the Berlin Defense was too passive and uninteresting, and informing us that few GMs played it.

5.d5 is positionally flawed, and that is a fact. It is an option in other similar positions from Ruy Steinitz with a pawn on c2 and knight on c3, when white can trade the light-squared bishops. Here, white can't do this, and more than that, Black does not have to lose time by Nc6-b8-d7 etc- so the idea is ill. End of story. If hou don't believe this, well... it's not my problem!

Even with this flawed move order, white could still try (1.e4 e5 2.Nf3 Nc6 3.Nc3 Nf6 4.d4 d6 5.d5 Ne7) 6.Bb5+. However, if Black is not equally blind positionally, will certainly play 6...c6! and enjoy a perfectly good game.

I agree with fiveofswords. I have said this as pertaining to the principle: "The best move is a position is only the best move you can find." If you are in a position which is "better" but where it is extremely difficult to find the best move, go with the inferior position where it is easier to play.

Remember that, at a very practical level, chess is only a draw until your opponent makes a mistake anyways. A very real facet of chess is to play for positions where you won't make mistakes and your opponent will rather than go for positions that win with best move order.

Guess what, with best moves, everything mentioned here draws at move 5. 5: d5 is a draw, 5: Bb5 is a draw, 5: dxe5 is a draw. There isn't a win for white.

So the only thing left to do is to enter the position where it is easier for you to find the right move than it is for your opponent. Your opponent has to make a mistake for you to win, and it's important to recognize that. Don't pretend like white has a forced win at move 5, just force your opponent to think as hard as possible and make the game flow as easily as possible for you.

pfren: At this point I believe little of what you say beyond the obvious. Just about everything you say is supported only with hand-waving and "if you're not smart enough to agree with me" disparagement.

Players far better than pfren have been proven wrong in their dogmatic harrumphing about opening lines.

The argument, again, is not whether 5.d5 is a killer move that puts Black on the ropes, but whether 5.d5 is a playable alternative for White at that juncture.

So far nothing you've said demonstrates 5.d5 is unplayable other than it goes against your idea of what should be played in such a position.

I let Stockfish run to 44 hours and 5.d5 was still the top choice. Which hardly means it is the best move but it's worth consideration. Maybe it's a problem with Stockfish's programming. Or maybe it reflects a blind spot in current master intuition with pfren as an example.

5..Nb8 also remained Stockfish's choice for Black. It would be interesting to compare that line with 5..Ne7. I looked into it briefly and it seemed Stockfish was "concerned" about the pin Bg5.

Fiveofswords makes a perfectly fine argument for preferring a line one understands well over a line which might be favored by a computer engine or current GM practice for that matter.

The queen trade and textbook grind-down in that Four Knights position was one of the first things I thought of.

FiveOS also makes a good point that ThrillerFan's advocacy for 5.d5 might relate to his familiarity with the French Advance variation.

Stockfish 6 on 1 CPU with MultiPV set to 7 isn't going to get the best results.

Letting an engine run for 44 hours straight is not the best way to determine what the best move in a position is.  You need to fill the hash up with several variations into the future then return to the start position to see what it thinks.

SK5: I'm not trying to get the best results. I'm just offering a data point that 5.d5 is a playable alternative.

So far I've read nothing substantial to the contrary.

Using a chess engine isn't the best way to determine what the best move on move 5 is...period. I agree with pfren on that much.

But a chess engine is a reasonable way to get an idea for plausible -- not necessarily the best, but plausible -- moves in an early position.

Of course there is an art to setting up the hash tables, threads, and so forth. I have written a UCI interface to talk to a chess engine and so it can "grade" my Guess-The-Move games, so I have some idea of the complexity there but I wouldn't call myself an expert.

5. d5 is a legal move that doesn't lead to a forced loss for White, so sure it's "playable".

By playable I intend something more like a "good enough alternative, though not necessarily the best."

I'm trying to avoid pfren mischaracterizing my position then ridiculing it, as he frequently does.

pfren claimed 5.d5 is "positionally blind." I understand resolving the central tension immediately like that is generally frowned upon these days. Which is fine. That's part of the accumulated intuition which makes a master or grandmaster.

But those intuitions, valuable as they are, may not always be right. As chess has advanced, more and more exceptions are found. Moves that were previously discarded because they were considered positionally blind or some equivalent have become standard master practice.

5.d5 might not be such a move, but it strikes me as worth investigating. I understand that it offends pfren's positional sense, but that's not sufficient reason to close accounts on it forever.

BTW, a good trick I learned while writing the UCI Guess-the-Move program was to analyze the game backwards so when the engine evaluated a position it could make use of the evaluations already stored in the hash tables -- for the reason SK5 mentioned in #47.

I am not a fan of this position to be honest.

Which is to say I do not know how I would continue.

I guess I would have to think about it.