I quite like the Iphone mode for the site - except one little issue - when trying to see the "latest kibitzing" for a famous player, one has to scroll past sometimes a tonne of bio information. Maybe in Iphone mode, the bio stuff should be somehow optional.

The <32nd Reykjavik Open (2017)> ends tomorrow.

It's a 10-round Swiss tournament taking place in the Harpa Concert Hall in the Icelandic capital from 19-27 April 2017.

The event gathers together 266 players, including 2700 players Giri, Andreikin and Jobava. The top prize is €5,000. The time control is 90 minutes for 40 moves then 30 minutes to the end of the game, with a 30-second increment from move 1. Draws offers are not allowed until move 30. There are no rest days but players can take one or two half-point byes in the first seven rounds.

<WannaBe: On my Android phone, I hold it upside-down, and it scrolls to the latest kibitz'n!! =))> Now that's a little bit of serendipity! On my iPhone all I get is an upside down chess board that quickly rights itself.

<zanz> <what do you mean 30-ply vs. 35-ply vs. 105-ply? (E.g. most searches are fixed time, seeing how deep the engine gets, or fixed depth, allowing the engine as much time as needed.)>

We only use fixed time searches (in fear of somebody asking for an unrealistic depth for a certain position) — but then we record the number of seconds processing, and also take note of the depth that it explored. This way we can reprocess positions only if the new reprocessing is longer/deeper than the previous attempts.

Not sure what you're saying here.

You know the processing, so it's only dependent on when you want to devote more processing time for a reprocess, right?

(I'm not understanding something here, I think.)

<i.e. the number of possible moves in the various ply>

This is confusing, a ply being just a 1/2 move after all.

And then this: < but in that one aberrant case where a perpetual was forced, >

The engine normally recognizes a perpetual long before the 50-move rule and drops the eval from some small advantage to 0.00 accordingly.

But maybe not in this case? OK, let me run SF8 on my local machine from the position to see if I can catch a hint as to what you're suggesting.

It seems to me that you're saying the eval was finite until you ran long enough to get to 100-ply (aka 50-move limit)?

<zanz> <This way we can reprocess positions only if the new reprocessing is longer/deeper than the previous attempts.> <Not sure what you're saying here.>

What I'm saying is that we keep track of the analysis of various positions, and when we're done we record the analysis itself, but also the number of seconds it took to get it, and the depth (in ply) that we achieved. That way if we ever want to reprocess a position, we'll only do it if we do it for more seconds, in hope of getting a more thorough analysis.

For example, today's Opening of the Day, the Richter Attack:

click for larger view

The Opening Explorer will show that we've done a ridiculous 1.5 hour analysis of it to achieve a depth of 33 ply.

If anybody wants to delve deeper into that position, it should be 2 or more hours. There's no point in doing a 30 second evaluation if we already have a 90 minute eval.

About 100 of the most popular opening positions have all been analyzed to depths of 30+ ply, at 30 minutes each.

The "biggest think" was on the starting position: it's been analyzed for 6 hours, just for fun. It turns out Stockfish agrees with Fischer: the best move is 1.e4. You might be surprised at what it thinks the next best moves are.

<It seems to me that you're saying the eval was finite until you ran long enough to get to 100-ply (aka 50-move limit)? That's a crazy depth to run to for any opening study, wouldn't you agree?>

It instantly realizes it's a draw (just like any human being would) but if you tell it to analyze it for a long amount of time it will. It won't learn anything new, it will just grind its gears for the prescribed time and continue maintaining it's a draw all the while. I'm not really positive if the 50-move limit came into play there, but the depth suggests that.

Opening Explorer is one example where the move number is off with Black to play.

Averbakh vs Kotov, 1953 (kibitz #100)

How is it that <CG> can actually play this game?

What movelist does <CG> use?

There could even be voting to select the game of choice.

https://www.chess.com/forum/view/co...

<The "biggest think" was on the starting position: it's been analyzed for 6 hours, just for fun. It turns out Stockfish agrees with Fischer: the best move is 1.e4. You might be surprised at what it thinks the next best moves are.>

I'm pretty sure I already know for Black... it's 1...e3.

The French, best by test.

Would it be the Winawer specifically?

The KGA has been "solved" - are you aware of this?

http://en.chessbase.com/post/rajlic...

< ... The main line of the King's Gambit, 1.e4 e5 2.f4 exf4 3.Nf3, is indeed winning for Black. Moreover, the only winning move is 3... d6!, just as Fischer claimed. For instance the more popular 3...g5 allows White to draw after 4.h4! In fact, Fischer's main line holds up incredibly well: 3...d6! 4.Bc4 h6! 5.d4 g5! (an exclam denotes any move which gives a better theoretical result than every alternative), although some side-variations from his article do have inaccuracies.

Q- Genius or luck?

A- Probably mostly luck. Naturally some of his lines are not accurate: they weave in and out of draws. But the main conclusion is correct.

Q- So is the King's Gambit really busted?

A- No, just if White plays 3.Nf3. Incidentally 3.Bc4 loses as well to 3...Nf6! (incredibly every other move allows White to draw). But this is where the fun begins. It turns out that the weird looking 3.Be2! leads to a draw. In fact we found that 3.Be2! is the only move that avoids a white loss.>

But then again, there's this pov on the matter:

GRENKE Chess Classic (2017) (kibitz #262)

Which is an Ionescu-like existential question- just what is the role of Carbon in this Silicon-based universe?

* * * * *

Actually, here's my main point:

<There's no point in doing a 30 second evaluation if we already have a 90 minute eval.>

Well, that's not exactly true. A couple of reasons come to mind

1) It's nice to have this info locally, i.e. on my own PC, for easy playback.

2) You don't say if you record the leading candidate moves at intermediate depths. This is potentially very-interesting to a human, as our visualization depth is limited. It's also interesting to know just how far a GM had to see to spot a "best"-move, if one emerges within the carbon-horizon.

In the best case there will always be situations where the analysis numbers don't match the line you're looking at, if some optional exchange of moves took place. Just for example Opening Explorer includes the optional 6. Nh4 Bc8 7. Nf3 if Black wants to tacitly offer a draw.

All of this causes me to think that perhaps the analysis should all be normalized as 1, 2, 3, etc. and a move-shift can occur when you click on the analysis window.

Anyhow thanks for pointing that out, it's exactly what I've been on the lookout for but haven't been able to find examples until you showed me those.

All of this causes me to think that perhaps the analysis should all be normalized as 1, 2, 3, etc. and a move-shift can occur when you click on the analysis window.>

The exciting part will be when premium members can request an analysis on the fly. They'll be told to come back later when the analysis is present, and other members can share in the fruits of the search as well. That way some obscure lines that are in vogue can be thoroughly evaluated by member request.

I am pretty sure that the move numbering problem is solved, for the most part. (There is still the possibility of transpositional move-shifts until I decide to implement the offset-method.)

Unless the old analysis had other problems beyond bad move numbers, wouldn't it have been simpler to keep the old lines and evaluations and implement the move-shift?

<I am pretty sure that the move numbering problem is solved, for the most part. (There is still the possibility of transpositional move-shifts until I decide to implement the offset-method.)

If you find an exception point it out please.>

Isn't that the point, all the same?

At some point analyzing too deep for the "best" move isn't really helpful.

Suppose the engine determined the French was the only opening offering Black a draw against 1.e4.

Would you advise people to stop playing the Sicilian or Spanish?