I like Karlovich even though I can't really argue against most of the criticism against her.

She's cute and I think her occasional arrogance suits her.

Also, she gives answers like these when interviewed:

Q: You are an attractive woman. Obviously, many people appreciate it more than your talent and only see your beauty, not your skills. How do you deal with it in your work?

A: Beauty will save the world.

Here's the real question people want to know about her:

Is she single?

And ready to mingle?

Sincurly,

McLovin AKA The Sexy Hamburger

Keti and Sopiko would be better and I think they are both more attractive (fwiw). Foina would be better. The only one who I am not sure about is Anna. She overpresses almost everything she is saying and comes off as very annoying to me sometimes. Although I liked her a lot when it was her + Sopiko.

karlovich is the best person at her job in the entire world

FACTS

Just had a play around with lichess's analytics tools. I play entirely 3 0 Blitz. Current rating is 1816. Games ending in resignation or checkmate I win 71% of the time. Games ending in flagfall I lose 72% of the time heh. Any other blitz players want to dig out their stats? I'm curious because it's not obvious what is optimal here.

Someone wrote an open-source version of AlphaZero called Leela. You can donate CPU time to help it learn, or you can play against it at lczero.org.

Judging by its game against the guy from ChessNetwork, it has a lot to learn:



White to play here. Leela (Black) evaluates its position at 56.9%, where 100% is a forced mate and 50% is a draw. But even to my amateur eye, this position is obviously a disaster for Black. White has a space advantage, the c-file on lockdown, his bishop is vastly better than Black's, Black's rooks are useless, especially the awful e8 one, and the pawn on c6 is a huge liability. I was like "how does it evaluate this positively, isn't that way off the mark?" so I put the position into Stockfish and it has +2.5.

Maybe Leela just needs more time and computing power, but her learning seems to be levelling off and not having learnt about basic things like file control, piece mobility, weaknesses etc suggests to me that there's a problem with the neural network architecture.

I got something ridiculous like 1975 on that test. I'm definitely not even close to that in reality.

They put out a new version today, and the net size is set to be increased within the next few days.

However keep in mind Leela has currently played only around 4 million games (AlphaZero had over 40 million). It really does have quite a while to go. This spreadsheet shows Leela's progression compared to AZ, and it seems to be following it pretty closely. AZ also had a long stretch of performance flattening off, followed by a big jump, right around where Leela is at now.

It's hard to predict exactly how high Leela will go, but it will be months, if not a year, before it starts to approach top engines.

haha, "Based on your move choices, our estimate of your Elo rating is 1790, with a 95% confidence interval of [1659...1921]."

Since you are definitely more than 200 points better than me I guess it underrated you :P

But seriously one flaw is that time didn't count, I tanked on a ton of the problems, and in real life I would just hang something.

If we ever played a sober, normal time constraint game, I seriously doubt we're that far apart.

Not to worry guys, my rating is 2299. Lol

I did all but 10 problems before bed while I was lying in bed. Then the other ten just now. I'm going to go out on a limb and just say it's very inflated. It takes quite a longer amount of time than I would have assumed. But maybe we can get YKW to do it and post a 2500 or 2600 rating.

hi guys

yes I'd agree that it's inflated, probably by around 250-300 points

You're the best Bowlgar, never change

I got 2258 so way over inflated by the looks of it

This. That was great.

Leela's really upped her game since that last post.

Here's her scoring 10-1 against 2400-rated IM Lasse Lovik.

https://lichess.org/KpOd9pN6
https://lichess.org/ZAYlXf8V
https://lichess.org/wtwvvw0e
https://lichess.org/J4H8GVJK
https://lichess.org/fC04bbly
https://lichess.org/0AvtMYAI
https://lichess.org/LU2ZRymT
https://lichess.org/BJnLvDY8
https://lichess.org/23QeURfi
https://lichess.org/nTGzgTMt
https://lichess.org/3Bl93u7n

The last move of the first game already confuses the heck out of me. leela made a move that mates in 4, where a mate in 3 was possible. so leela is tactically weak?

Probably terminates search when a forced mate is found. There's no need to search further at that point and computing resources are limited.

I have a problem with Fritz GUI/Komodo 10 (Windows 10-64bit).

After a recent reinstall it became a huge memory hog, regularly using 12 out of 16 GB RAM. While the max hash table size is 12 GB I set it to 8GB. The task manager still shows the program using 12 GB. With Firefox and VLC player running total memory usage is 95%+ and sometimes spikes to 100% causing freezes. I need to Alt+F4 the program to get the computer back to being responsive.
To keep this from happening while multi-tasking I have to kill some background processes currently not needed (eg. cloud services). Is there anything besides the hash table size that I can adjust?

On the other hand why is CPU usage only at ~50% despite all 4 cores being assigned to Komodo?

Wait what? In the US Championships Armageddon game time controls are 5+2 vs 4+2 with draw odds for black?

Never seen or heard of an Armageddon with increment (kind of poorly named if it has one, imo. Better would be Eternal War or something).

This has got to be MASSIVELY +EV for black, no?

Black will never lose on time, and has draw odds, lol.