Yep.



White is down a pawn here but has great compensation -- he controls d5, Black's bishop is quite bad, and Black's structure kind of sucks. 17. Qxb4 regaining the pawn would have been anti-positional for at least a few reasons.



The theme of superior knight against bishop persists for quite a while. The White pawns serve more as restrictions against Black's bishop than as any sort of overt king attackers. And then So ends up blundering.

Carlsen had more than full compensation for the pawn for three reasons:
1. He got some positional advantage (+0.5)
2. He is Carlsen (+0.4)
3. His opponent is So (+ 0.33)

Reasons 2 and 3 would have been valid even if Magnus hadn't sacrificed the pawn

Still, the suboptimality of 17. Qxb4 is beyond my understanding as a club level player I mean, I'm aware of d6-d5-d4, but I'd think I'd still be able to defend that position vs a similarly weak player.

It's bizarre that the rating analysis gives Levon only a 17% chance of winning the event; seeing his current tactical brilliancies, I think it's at least 30%.

I think it also has to do with opening the b file

Yeah, opening the b-file, activating Black's bishop, giving him space in the center for an attack on the queenside...all with tempo.

I wonder what someone's rating would have to be to have converted that game. Like, would a 2200 or something have converted magnus' position more often than not against So.?

I agree it seems a little low at first, but of course the model doesn't adjust for form or his "current tactical brilliancies" in any way beyond using his improved live rating. He's still 2776, while Carlsen is 2854, and they're tied for first with four rounds left (plenty of time for the rating edge to matter). Back of the napkin sanity check might say that if Aronian's odds of winning are X, Carlsen's should be at least 2X. Then there's Topalov just half a point back, also with a big rating edge over Aronian, and with a pretty easy remaining schedule as I've said before. Makes sense that those factors could overcome the rating edge and give him roughly X odds as well. Right there that gives Aronian a cap at 25%, and that's if you thought noone else had any chance at all which is obviously silly because with four rounds left all sorts of crazy stuff can still happen. Give some equity to the long shots, and Aronian drops to maybe 22%.

Then we run the numbers and see that the model gives Carlsen more than a 2:1 edge, and Aronian shakes out at 17% instead. Reasonable enough if you think 2776 is a good estimate for him. If you wanted to give him credit for being in better form than that his odds would increase. I reran the sim with everything the same except Aronian, whose rating I bumped to 2820, and those 44 extra points improved his odds of winning to 27%, right around your estimate of 30. I can get him all the way up to 30% by putting his rating at 2832, just two points above his published peak.

So you're not really questioning the model, you're just saying that it feels like Aronian is genuinely playing at 2830 strength in this event (not just getting lucky so far) and that you expect him to continue doing so over the last four rounds. Not particularly unreasonable

The only element in the model that I feel like questioning is the draw probability, which the ratings don't predict (they do only the expected score in a game - p(win)+0.5*p(draw) points); e.g. if Aronian or Topalov remains tied with or half a point behind the leader and starts 'gambling' (making positional sacrifices or just creating imbalances) in the last couple of rounds, the loss and win probabilities will rise, while the draw one will fall.

But OK, I was too optimistic, the 22-25% chance estimate sounds more plausible.

FML, I had to tune in late, and Nakamura is playing my favourite KID.

Going to have to watch the recording on Youtube tomorrow to see it in all its glory.

I wish they would show a bit more chess on their chess show.

Anyone watching So-Naka right now? Really interesting game.

Looks like Caruana is blowing his chances of getting back into this.

Good guy So let's Naka mate him.

Toppy spinning out of control; messy Caruana wreck; So so-so? no So oh no!

Gets mated, for the fans!

I thought it's a population tendency that top players let a mate appear on the board if it's fast and forced but resign if it's not forced and they're just at a fatal disadvantage that will take many boring moves to be converted into mate (e.g. they're a piece down and don't want to spend 20 more moves on having their pawns grabbed and their king checkmated by a promoted queen, that's too standard and not spectacular).
So, who's the scrub now?

You have to give So credit for his fighting spirit. He is playing these sharp trash-lines in order to win. If he wanted a draw he could have played the Qc2-Nimzo against Aronian, the Marshall against Carlsen and the Exchange-KID against Naka. It's very easy to play safe with white.

So took many risks and got zero rewards so far. That's the difference between games with complete information and games with incomplete information. The only variance comes from blunders.

Yeah, So seems hell bent on winning or losing most of his games, lol.

Props to Levon for playing a non-Berlin Spanish

* looks at the board again 5 minutes later * Where the heck are the queens?

OK, Aronian has proved once more that he's #NotAQueenPerson. I guess the last time when he could deviate was 10... Nxe5 instead of dxe5. White usually doesn't want to exchange queens then.

Aronian wins, Carlsen could lose. Drama.

Grischuk's position against Carlsen does look promising (KRBPPP v KRNPP).

I haven't been able to load chessgames.Com all day

yeah it's down

http://downforeveryoneorjustme.com/h...hessgames.com/

LOL enginebots

Hah. I'm watching at http://grandchesstour.com/2015-sinquefield-cup-live which, while it has some engine score, it at least blissfully free of amateur commentary.