Open Side Menu Go to the Top
Register
*** Chess Low Content Thread *** *** Chess Low Content Thread ***

11-18-2017 , 05:13 AM
Quote:
Originally Posted by Mayo
This article is so awful that I am legitimately upset after reading it. This dude is a charlatan, and I cannot believe that the WSJ gave him this fawning feature article.

I might be the worst chess player in this thread,* and I would crush this dude. His chess is abhorrent. The article talks about how bad 12. Qf3 is. It even correctly points out that it's not bad because it loses a piece to a tactic, but it's bad because it doesn't accomplish anything even if that tactic isn't there. That discussion was good, though it made it sound like Magnus's 12. ... Qh4 was some very subtle GM move. Even I would find that, seeing the threat of Qxh2#, which is the key threat in the tactic Magnus uses to win the bishop. Anyone who plays chess would see that. Anyone who bothered to try would be able to calculate that simple tactic.

But the article doesn't even mention that after that tactic, the dude just sits back and lets Magnus capture a rook, then walks his king and bishop directly into a fork for no reason whatsoever. In short, he doesn't just lose because he blunders, he blunders while he's just moving pieces around aimlessly, revealing his utter cluelessness.

Basically, I'm pissed at the hubris necessary to even attempt this stunt. Who thinks they can be world class at anything in 30 days? I'm actually offended by the disrespect there. And I am stunned that the damn Wall Street Journal gave it any coverage, let alone the full feature fluffer treatment.

The dude's algorithm idea is truly awful. He's basically trying to invent the kind of chess program that was used in the 50s by Alan Turing, memorize some of its results, and then use that to beat the GOAT. This plan is asinine, and I would laugh in the face of anyone who presented it to me. The fact that he couldn't even pull off this dumbass plan makes me think he's completely full of crap in general.

The other "extreme learning" things he has done are not at all impressive, and some of them aren't even learning.

Basically, this dude seems like a fraud, and the WSJ treated him like a genius. And his chess is truly awful.

*~1300 rating on chess.com


Mayo you are the ****ing man and everyone should know it.

I agree 100% with what you said.
*** Chess Low Content Thread *** Quote
11-18-2017 , 01:28 PM
Hi, im looking at how to win consistently with a outside passed pawn. I add a video just to show what the board is like. I know you are supposed to eat villains pawns when his king commits to your a-pawn. But i cant win consistently, sometimes it just ends up in a draw or even a loss. This should be winning every time right?

So i wonder if there is any principles that you can always follow that leads to a win. For example dont allow certain pawn structures to form, or trade off certain pawns before you come in with the king etc.

*** Chess Low Content Thread *** Quote
11-18-2017 , 01:41 PM
Lol you know polgar was all "wtf why are you wasting my time with these stupid questions"
*** Chess Low Content Thread *** Quote
11-18-2017 , 07:41 PM
Best thing to do is post some of the games where you fail to win, and see what people say about it, and maybe you will find some common error that you are making.

But when it comes to K+P endings, instead of trying to learn principles, try to calculate it out completely.
*** Chess Low Content Thread *** Quote
11-20-2017 , 03:02 AM
http://www.chessgames.com/perl/chessgame?gid=1891903

Beautiful win by Nakamura in the GP
*** Chess Low Content Thread *** Quote
11-20-2017 , 06:42 PM
I'd be surprised if Max Deutsch could beat a 1300 with another month's training. You could probably implant a chip in his brain with a full opening db and he'd still lose to a 1500
*** Chess Low Content Thread *** Quote
11-22-2017 , 04:45 AM
Quote:
Originally Posted by Army Eye
http://www.chessgames.com/perl/chessgame?gid=1891903

Beautiful win by Nakamura in the GP


Sick game. Def worth playing through.
*** Chess Low Content Thread *** Quote
11-22-2017 , 09:00 PM
I'm not going to dignify him with an actual link to his blog (but if you really want to look at the original, I'm sure you know how to look for it yourself), but here's a post by that fraud who claimed to be able to beat Magnus not after 30 days of chess study, but after 34 hours.

For some inexplicable reason 34 hours just didn't quite cut it. 500 hours and you should be fine, though.

Quote:
Like with all of my past challenges, today I decided to tally up the total amount of time I spent on this final chess challenge.

Since this challenge was effectively 50% longer than previous challenges, it’s no surprise that I spent a bit longer on it. In particular, over the past six weeks, I committed 34 hours to the pursuit of defeating Magnus.

It turns out that 34 hours isn’t quite enough, but, knowing what I know now, I don’t think it’s too far off.

I’d estimate that it would take between 200–500 hours to become a human chess computer capable of defeating the world champion (assuming that an algorithmic approach at this level of gameplay is possible… the verdict is still out).

While this is considerably more time than the 34 hours I spent, it’s completely dwarfed by the tens of thousands of hours that Magnus has spent playing chess.

Of course, this estimate only matters if I can actually demonstrate the result.

For now, I’m going to take a little break from my chess preparations, but, if inspiration strikes, I’m may proceed forward.

500 hours really isn’t so crazy. In fact, it’s only about three months of a standard 9-to-5 job.

I suspect that I’ll be circling back some time in the future, putting in these 500 hours, and taking down Magnus (in what will still likely be a very lengthy game).

Until then, Magnus can continue enjoying his spot at the top…
Spoiler:
Quote:
For now, I’m going to take a little break from my chess preparations
Why do I have the strong feeling this is the last thing we're going to hear from you on the topic for ever?
*** Chess Low Content Thread *** Quote
11-22-2017 , 09:13 PM
I mean.... any chance that's tongue in cheek?
*** Chess Low Content Thread *** Quote
11-23-2017 , 07:13 PM
Quote:
Originally Posted by ChrisV
I mean.... any chance that's tongue in cheek?
Given his ego, no
*** Chess Low Content Thread *** Quote
11-23-2017 , 07:43 PM
What a ****ing douchenozzle loser.
*** Chess Low Content Thread *** Quote
11-23-2017 , 08:50 PM
At this point he is an insult to those of us who spend a good chunk of our lives on the game of chess. It's too bad Magnus did not tell him "**** off, first earn the right to play me".
*** Chess Low Content Thread *** Quote
11-23-2017 , 10:28 PM
What exactly is "earning" the right to play Magnus tho?

Making NM?
Making IM?
Making GM?

Fwiw I agree with you. *** Chess Low Content Thread ***
*** Chess Low Content Thread *** Quote
11-23-2017 , 10:56 PM
I guess getting good enough to be invited to the same tournaments as Magnus? Although even making NM is not very much in danger of happening, so it's pretty moot.
*** Chess Low Content Thread *** Quote
11-23-2017 , 10:58 PM
Quote:
Originally Posted by TimM
I guess getting good enough to be invited to the same tournaments as Magnus? Although even making NM is not very much in danger of happening, so it's pretty moot.


500 hours to beat Magnus though!

Think of the product he could sell.
*** Chess Low Content Thread *** Quote
11-23-2017 , 11:13 PM
Is he really saying he can memorize a game tree with trillions of decisions?
*** Chess Low Content Thread *** Quote
11-23-2017 , 11:15 PM
Quote:
I’d estimate that it would take between 200–500 hours to become a human chess computer capable of defeating the world champion (assuming that an algorithmic approach at this level of gameplay is possible… the verdict is still out).
*** Chess Low Content Thread *** Quote
11-24-2017 , 12:26 AM
Quote:
Originally Posted by Tuma
Is he really saying he can memorize a game tree with trillions of decisions?


He is a genius entrepreneur so anything is possible.
*** Chess Low Content Thread *** Quote
11-24-2017 , 01:43 AM
I'm clicking through all his posts on the subject for lols. Dude is completely clueless. He's trying to play a game he has no idea how to play with machine learning but has little idea how that works either. Some choice excerpts:

Quote:
Yesterday, on the plane, I had a really interesting idea: Could a room full of completely amateur chess players defeat the world’s best chess player after less than one hour of collective training?

I think the answer is yes, and, in fact, they could use a distributed version of my human algorithmic approach (i.e. Max Chess) to do so.
Quote:
Before starting this month’s challenge, I was a bit concerned about studying chess. In particular, I imagined that learning chess was just memorizing predetermined sequences of moves, and was worried that, by simply converting chess to memorized patterns, the game would lose some of its interest to me.
Quote:
I’m not sure how many chess positions I’ll need in order to train a decently functional model, but I’m guessing somewhere between 25,000–50,000 should be enough
Uh huh. The obvious happened and after training on 7,000 positions he overtrained the **** out of it and then didn't even realize that is the problem:

Quote:
Eight hours later, after 10,000 iterations through the 7,000 chess positions (3,000 chess positions have processed since then), the model’s accuracy on the training data leveled off at around 99%.

At first, I didn’t realize that this accuracy was based on the training data, so I thought this was unbelievably good. In other words, I thought this accuracy number represented how the model performed on data it had never seen before (while in reality the model was optimizing itself around this exact data).

When I tested the model on genuine test data (i.e. data the model truly had never seen before), it was only able to correctly evaluate the position about 70% of the time, which is not great.
It later turns out, hilariously, that 70% of the evaluations in his database of positions are "good", which means the computer can achieve 70% just by guessing "good move" on any candidate move.

After running into these problems, he decides that the way he's built his database of positions is no good. So this is, no joke, his solution:

Quote:
Labelling my dataset v2
An hour ago, I came up with a much better (and computationally faster) way to label my dataset. Here’s how it works:

Load a bunch of chess games.
For each game, determine if White won or lost.
For each chess position in each game, label all the positions as either +1 if White won the game and -1 if White lost the game (and 0 if it’s a tie).
Process millions of games in this way, and keep a running tally for each distinct position (i.e. If the position has never been seen before, add it to the list. If the position has already been seen, increment the total associated with the position by either +1 or -1).
Once all the games are processed, go through each of the distinct positions, and label the position as “good” if its tally is greater than or equal to zero, and “bad” if its tally is less than zero.

In this way, every chess position has a unique label. And each label should be highly accurate based on the following two thoughts: “Over millions of games, this chess position led to more wins than losses, so it’s almost certainly a good chess position” and “Over millions of games, this chess position led to more losses than wins, so it’s almost certainly a bad chess position”.
He's going to build his own Chessbase and then use its win/loss percentages from each position as a position evaluation. This where I stopped reading. There's weeks and weeks of this ****, folks.
*** Chess Low Content Thread *** Quote
11-24-2017 , 08:10 AM
rofl
*** Chess Low Content Thread *** Quote
11-24-2017 , 11:15 PM
ROFLCOPTER
*** Chess Low Content Thread *** Quote
11-25-2017 , 01:38 AM
You guys can laugh all you want but right now this guy is no doubt toiling away in his lab, on his way to solving chess and attaining a 3000+ rating

Spoiler:
*** Chess Low Content Thread *** Quote
11-25-2017 , 01:39 AM
LOLOLOL
*** Chess Low Content Thread *** Quote
11-26-2017 , 12:58 AM
THE VERDICT IS STILL OUT THOUGH
*** Chess Low Content Thread *** Quote
11-30-2017 , 02:38 AM
Lifetime bucket list goal met!!!! Just broke 1800 USCF!!! I would have to drop out of 2 monthly tourneys for it to officially show up on my profile, but my plan is to just keep playing well and do my best to win the tourneys. Once upon a time, I thought I would stop at 1800 and move on to something more financially rewarding (as opposed to financially draining like chess lol). But since 1700, I have known that I plan to keep going as long as I feel like I am learning and improving.

And yes, I have in fact learned the secret to rapid chess improvement:

1) Stop wasting time doing dumb stuff like researching chess books, stop repeatedly calculating hypothetical USCF ratings based on hypothetical scenarios, stop looking up your opponents' profiles to find some convoluted way that your opponent is "way underrated", stop planning obsessively, stop wasting time watching videos of your favorite GM playing with his cat, etc.

2) Your current chess library and software is "good enough". The time you waste perfecting your study plan could be much more efficiently spent studying and re-reading and then re-reading again what you already have. Book and board was more than adequate for monsters like Lasker, Capablanca, Alekhine, and Fischer. In my case, I just dumped the good chess books I already have on my coffee table and vowed to finish them before even considering buying more books or software. Jeremy Silman's books are good enough. Those weird Russian books from 1973 written in descriptive notation are good enough. Your chess.com study plan is good enough, etc. Get your face in the books and go. You're still going to dick around and have fun anyway, but do your best to get your ass in gear.

3) Poker is a game for pessimists. Chess is a game for optimists. Never let thoughts such as "my opponent is underrated" or "mathematically I have no chance because this guy is 450 points higher than me (untrue btw)" or "this position is hopeless, time to throw a hail mary or two and go home (pull your dick out from between your legs and find the stalemate, perpetual, fortress, intelligent complications that will save you... only give up against club players when you are down massive material or about to be mated... they will find ways to screw up!) You may suck at endgames, but find a way to believe that you can win or save a draw in game time. Maybe your opponent suck even harder at endgames or just hates them so much that he quits (quite common!)

4) Bring your poker hustle into your chess games! Watch for signs that your opponent is uncomfortable and know his weaknesses. Blitz your openings and create weird positions against "perpetual time trouble guy". Force long, slow maneuvering against "impatient attack, attack, attack even if it ruins my position guy". Sac a pawn and open up the position against "grindy, positional, endgame guy". Trade queens against "quits in the endgame guy". Learn how to psychologically break a wide variety of opponents. They will wonder how in the heck you play so bad, but get so lucky.

And an observation... chess and the chess atmosphere are absolutely awesome and refreshing compared to poker imo. Playing with kids and teenagers is fun. Playing with old guys who refuse to grow old is fun. Playing with "9-5 family man whose guilty pleasure is chess" is fun. Getting in melodramatic arguments with the tournament director is fun. I knew poker players were scum, but playing chess regularly has shown me that poker players are even scummier than I realized. Poker players can always benefit from spending more time around normal people.

Off to go study!
*** Chess Low Content Thread *** Quote

      
m