The curious case of 1200: The Expert's rating

Sort:
blueemu
CrusaderKing1 wrote:
blueemu wrote:

If you were curious about my method of beating Engine Monkeys, it relies on the fact that cheaters rarely cheat in only one game. If they cheat at all, it is typically in several games. So as soon as I realize that I'm playing against an Engine Monkey, I immediately go into damage control mode and try to slow down the tempo and draw the game out as long as humanly possible. This gives the Cheat Detection team as much time as possible to catch the cheater in one of his OTHER games, which results in me winning on time when the cheater gets banned.

It actually works the majority of the time.

How do you know you're playing against an engine monkey? I feel like its too arbitrary to simply say "they play too good". So, what's your analysis kind sir?

Intuition, based on more than 50 years of playing chess (my first chess tournament was in 1968) and on nearly 50 years of working with computers (my first job working on a room-sized mainframe computer was in 1974, while my last was as a Computer Systems Analyst for the military).

Engines don't "think" the way people do. People think by "chunking" the position and then manipulating those conceptual chunks. Engines "think" (if that word is at all applicable) by exhaustively calculating every possible move and every possible reply to every possible move, ad infinitum. The two methods feel different... at least to me.

If you feel that I'm wrong, then get back to me after you also have fifty years experience in both fields, and we'll discuss it.

CrusaderKing1
CrusaderKing1 wrote:
blueemu wrote:
CrusaderKing1 wrote:
blueemu wrote:

If you were curious about my method of beating Engine Monkeys, it relies on the fact that cheaters rarely cheat in only one game. If they cheat at all, it is typically in several games. So as soon as I realize that I'm playing against an Engine Monkey, I immediately go into damage control mode and try to slow down the tempo and draw the game out as long as humanly possible. This gives the Cheat Detection team as much time as possible to catch the cheater in one of his OTHER games, which results in me winning on time when the cheater gets banned.

It actually works the majority of the time.

How do you know you're playing against an engine monkey? I feel like its too arbitrary to simply say "they play too good". So, what's your analysis kind sir?

Intuition, based on more than 50 years of playing chess (my first chess tournament was in 1968) and on nearly 50 years of working with computers (my first job working on a room-sized mainframe computer was in 1974), while my last was as a Computer Systems Analyst for the military.

Engines don't "think" the way people do. People think by "chunking" the position and then manipulating those conceptual chunks. Engines "think" (if that word is at all applicable) by exhaustively calculating every possible move and every possible reply to every possible move, ad infinitum. The two methods feel different... at least to me.

If you feel that I'm wrong, then get back to me after you also have fifty years experience in both fields, and we'll discuss it.

What would be an example of what you see in engine moves vs. human moves? Whenever I check the analysis of my past games, they seem to want fairly reasonable moves seem and human to me. 

 

blueemu
CrusaderKing1 wrote:

What would be an example of what you see in engine moves vs. human moves? Whenever I check the analysis of my past games, they seem fairly reasonable and human to me. 

In general: Counter-intuitive moves that make no conceptual sense but are supported solely by tactics. People use logic and concepts to winnow the available legal options down to a small number of candidate moves, and only THEN do they start to analyze. Engines don't.

But this thread is the wrong place to discuss engine use. Try the Cheating Forum.

Cheating Forum - Chess Club - Chess.com

CrusaderKing1
blueemu wrote:
CrusaderKing1 wrote:

What would be an example of what you see in engine moves vs. human moves? Whenever I check the analysis of my past games, they seem fairly reasonable and human to me. 

In general: Counter-intuitive moves that make no conceptual sense but are supported solely by tactics. People use logic and concepts to winnow the available legal options down to a small number of candidate moves, and only THEN do they start to analyze. Engines don't.

But this thread is the wrong place to discuss engine use. Try the Cheating Forum.

Cheating Forum - Chess Club - Chess.com

Hmm, ok I will take your word for it.

CrusaderKing1
Tonya_Harding wrote:

@Crusaderking1

The other day, as someone asked for help in order to improve, I played a game with them, unrated, and spent quite some time at each move, explaining the ins and outs of the situation.

Then happened Black's move 7, and I immediatly knew I was not playing any 939 as profile displayed. I still replied 8.b3 out of inertia.

They told me, they were following opening database (which would be okay for a true game), on which, I resigned the game, furious I've wasted my time on someone who was not providing any work for the game, that was no true game, but was supposed to be a Chess lesson.

Here the game:

 

So, yes, one can pretty much see, someone who's rated a few hundred rating points below does play moves that should be waaaaay above their understanding of Chess.

Ok thanks. Just out of curiosity, why is that move so good?

GMongo
blueemu wrote:
Tonya_Harding wrote:

Ratings are surely some inflated on Chess.com...

You feel that my 2751 rating was inflated? 

... well... maybe a little.

 

Maybe because it shows 2351 not 2751?  

blueemu
GMongo wrote:
blueemu wrote:
Tonya_Harding wrote:

Ratings are surely some inflated on Chess.com...

You feel that my 2751 rating was inflated? 

... well... maybe a little.

 

Maybe because it shows 2351 not 2751?  

It does now, yes.

For the full story, check post #21 of this thread:

The "Daily rating" rebalancing change has been reverted... - Chess Forums - Chess.com

GMongo
blueemu wrote:
 

It does now, yes.

 

Wow that's really strange because even when I go to your home page is shows 2351.  When i view your profile it still shows just 2351.  When I click on your latest game it shows 2351.  Something is weird with chess.com then.  Wonder if someone else can look and confirm what they see.

blueemu
GMongo wrote:
blueemu wrote:
 

It does now, yes.

 

Wow that's really strange because even when I go to your home page is shows 2351.  When i view your profile it still shows just 2351.  When I click on your latest game it shows 2351.  Something is weird with chess.com then.

Check my edit, above.

GMongo
blueemu wrote:

Check my edit, above.

So if I read that correctly you were inaccurately rated 2751 and then it was corrected to the more accurate 2351, which is still great.

blueemu
GMongo wrote:
blueemu wrote:

Check my edit, above.

So if I read that correctly you were inaccurately rated 2751 and then it was corrected to the more accurate 2351, which is still great.

Not exactly.

I'll try again to explain it.

Ratings are statistical. They rely on underlying math, and on statistical distributions.

The key relationship that is supposed to tie ratings to reality is that a 200-point difference in ratings is supposed to correspond to a 3-to-1 distribution in actual results. If I am rated 200 points higher than Peter Patzer and we play a match, then in order to break even in rating points (no gains, no losses) I would have to end up with a 75% score while he would need a 25% score. 

Any difference in results (either more or less than 75%-to-25% for a 200-rating point gap) would result in one of us gaining points at the other player's expense.

But all this is dependant on the underlying math and on the statistical distribution of ratings.

The profile of the Daily Chess rating distribution was not similar to the profiles of the rating distribution for Bullet, Blitz and Rapid. With a distorted distribution of ratings, the probability... amounting to a certainty, in this case... is that some players are losing points when they have actually performed above mathematical expectation and should be gaining points, or vice-versa.

Chess.com decided to impliment a graduated adjustment to the Daily ratings, with the amount of the rating shift proportional to (the player's rating minus 1200). Players at 1400 strength would receive 50 points, while players at my strength would receive 400 points, and players intermediate between those extremes would receive an intermediate amount.

This would adjust the profile of the Daily rating curve to roughly match the shape of the Bullet, Blitz and Rapid curves... and now the rating formula would have similar results across all four chess time controls.

But there was a considerable outcry when the changes were introduced... a great wailing and gnashing of teeth... and Chess.com reverted the changes (and back down to 2351 I went) and they decided to study the matter further before taking any action.

Marie-AnneLiz
Tonya_Harding a écrit :
CrusaderKing1 a écrit :

Ok thanks. Just out of curiosity, why is that move so good?

That's beside the point. I'm not here to share all my openings secrets with everybody. Learn Chess strategy, and you'll find out.

The point is only, it's beyond any possible understanding or even intuition of someone as low rated as my "pupil" was.

thumbup.png

CrusaderKing1
little_guinea_pig wrote:
CrusaderKing1 wrote:
Tonya_Harding wrote:

@Crusaderking1

The other day, as someone asked for help in order to improve, I played a game with them, unrated, and spent quite some time at each move, explaining the ins and outs of the situation.

Then happened Black's move 7, and I immediatly knew I was not playing any 939 as profile displayed. I still replied 8.b3 out of inertia.

They told me, they were following opening database (which would be okay for a true game), on which, I resigned the game, furious I've wasted my time on someone who was not providing any work for the game, that was no true game, but was supposed to be a Chess lesson.

Here the game:

 

So, yes, one can pretty much see, someone who's rated a few hundred rating points below does play moves that should be waaaaay above their understanding of Chess.

Ok thanks. Just out of curiosity, why is that move so good?

I'm just a patzer, but I think it's good because it stops b4 and grabs space on the Queenside - ideas which no 900 player would have in mind at all when playing.

Sounds reasonable to me. I looked up a couple games where that move was played and I think you're right.

nTzT

a5 stops white from gaining more space on the queenside which would come with tempo on the bishop etc.

Using a database while playing is cheating.

CrusaderKing1
Tonya_Harding wrote:
nTzT a écrit :

a5 stops white from gaining more space on the queenside which would come with tempo on the bishop etc.

Using a database while playing is cheating.

Not in daily.

Database and chess engine are different correct?

CrusaderKing1
Tonya_Harding wrote:

Correct.

A database is like past games being played? Where are the best places to find a database?

xor_eax_eax05

 

Just no. That table is nonsensical. For everyone else, I've said it many times before: ratings on this site, especially under 2000 ELO, make no sense. There are 1000 ELO here who are stronger than 1500s (from this very same site) and, for some reason, are stuck in their bracket.

As a player who dwells around 1600-1800 elo in another site which is almost excusively dedicated to Daily time control, I have played ppl here at the 10' time control here who are 1000 and are way stronger than many club players I've beaten. I can give you loads of examples, but I'll just post one. This is a recent victory of mine over a 1980 rated player on the other site. I'm only including the moves and not the site because I dont want my post removed:

 

 

 To my surprise, guy playing white has a blog. Turns out he was a frequent poster till mid 2019... he's a 1900 +/- elo club player living in Germany.

This is another win of mine vs a 1950 player:

 

 

Ok then, let's go and see here the playing / defense strength of 10' rapid 1100-1200 ELO players here - very recent games:

 

https://www.chess.com/analysis/game/live/6442456396

https://www.chess.com/live/game/6442184685

https://www.chess.com/live/game/6442159828

etc. etc. etc. I simply have no chance vs these kind of players, they are too strong, even stronger than the 1800-1900 club players I play against in the other site. 

It's also quite funny I've recently been paired against a 1500+ in the Arena, and it was an incredibly easy game:

https://www.chess.com/live/game/6418431254

It really felt like playing a 1500 from the other site, you know, the ones who join tournaments with players several hundred points above them just for the kicks? I've faced those.

 

Now, who's going to claim a 1100 ELO rated from this site is stronger than a 1900 club player? Probably no one. Yet I play them every day here.

So, in my opinion, ratings on this site make no sense and that table is wrong. Maybe once you go past the 2000 elo barrier things start to make more sense, but I wouldn't know about that because my higher ever was 1890 or so.

Oh, and just one more thing. Many years ago when I first started playing chess frequently rather than once in a blue moon, my initial rating after the provisional period was 1200, which matched my FICS server rating ... so, according to this site's score, I've grown weaker ... but for some odd reason I gained 400-600 points on the other site. 

CrusaderKing1

Refer to xor's post. His post makes more sense. It also shows you how insane it is for you to care about ratings +/- 100 elo.

blueemu
CrusaderKing1 wrote:
Tonya_Harding wrote:

Correct.

A database is like past games being played? Where are the best places to find a database?

There are free databases online.

Chess Opening Explorer - 365Chess.com

CrusaderKing1
blueemu wrote:
CrusaderKing1 wrote:
Tonya_Harding wrote:

Correct.

A database is like past games being played? Where are the best places to find a database?

There are free databases online.

Chess Opening Explorer - 365Chess.com

Thanks!