Home » Forum Home » General

Topic: Color next to our player name
Replies: 47   Views: 21,024   Pages: 4   Last Post: Apr 12, 2019, 12:25 AM by: watsu

Search Forum

Back to Topic List Topics: [ Previous | Next ]
Replies: 47   Views: 21,024   Pages: 4   [ Previous | 1 2 3 4 | Next ]
watsu

Posts: 975
Registered: Dec 16, 2001
Home page
Re: Color next to our player name
Posted: Apr 10, 2019, 8:26 AM

Nice one, Karl Get within 100 points of my rating (which shouldn't take you long) and we can play. I don't mind losing to you. I just minded the idea of losing to you when you were rated 300-400 points below me, which we both know is inaccurate.

I don't want to play non tournament rules games. If you take one of my unrated invites, play tourney

karlw

Posts: 915
Registered: Mar 7, 2006
From: Eugene, Oregon
Age: 31
Re: Color next to our player name
Posted: Apr 10, 2019, 8:31 AM

What are you talking about, my friend? We just established that TB ratings are inflated, and I'm sure mine is no exception

It's no use going back to yesterday, because I was a different person then.
haijinx

Posts: 64
Registered: Jan 20, 2019
From: Salem Oregon
Age: 48
Home page
Re: Color next to our player name
Posted: Apr 10, 2019, 2:02 PM

some quoted material from me and watsu:

--

"The discarding of ratable games, which sort of violates a core principle of these systems, is assuredly a key factor in this."

Ratable games have been discarded in live games here for over a decade now, yet as you put it "Live Pente ratings seem Elo-typical".

--


First, Live Pente went years with a solid base before they were discarded, right? It was well-established. It's also been dormant for some time.

TB Pente never had such a foundation.

I also said I did not investigate this issue, just that from your description of how the ratings were and how they changed when data started being discarded you describe something more Elo-typical.

Second, I was not talking about individual games and ratings. I'm talking about systems.

There's 2700+ FIDE chess players. Some of them could conceivably get there fast by playing highly rated people now. That's because the system has built-up a huge number of players after a huge number of games.

Inflation happens in the system...but it took literally millions of rated games to get the system that high from initialization.

TB Pente has reached that 2700+ point after what 300,000 games? It just crossed 675,000 recently. That level requires an orders of magnitude difference in number of games than we see here. Millions and millions of games. Having anyone rated 2700 with so few games in the system is an extreme outlier. That's huge inflation and points to an error in the system.

Discarding data is that error.

The other sites you mention, though I don't know them, don't seem to have this issue or this reputation.

more quotes, from watsu

--

I'm not and have not been saying that TB ratings aren't currently hyperinflated here (they are) or that the no penalty for draws isn't primarily responsible for that hyper inflation (it is).

--


Great!

--

But, anyone who wants to can be a TB expert or master at Pente, for any given value of the words "expert" and "master". The database breaks TB that much.

--


I sort of disagree here. Players still have to prosecute the wins and even past greats have multiple errors in their play. People often pick the obvious, slower, and riskier choices.

In any event, I've played turn-based chess and correspondence and there's huge amounts of database work in both. Both players have access to similar resources in this day and age, and in the end it's still a contest of personal creativity.

watsu

Posts: 975
Registered: Dec 16, 2001
Home page
Re: Color next to our player name
Posted: Apr 10, 2019, 2:58 PM

"I sort of disagree here. Players still have to prosecute the wins and even past greats have multiple errors in their play. People often pick the obvious, slower, and riskier choices."

And yet... the obvious, slower and riskier choices still do win and the ratings still do go up on the basis of winning - just as much as they would for finding faster, less obvious or less risky options. The Elo system will never help identify experts or masters on the basis of their creative play, it's simply a win/lose up/down system. The database is more than adequate for finding sure P1 wins (if one searches well). Anyone with a mind to can achieve my current TB rating using the database in relatively short order, unlike in chess (or live games against master Pente players) which takes a lot more skill, study and creativity to achieve. Alright, I'm done

I don't want to play non tournament rules games. If you take one of my unrated invites, play tourney
haijinx

Posts: 64
Registered: Jan 20, 2019
From: Salem Oregon
Age: 48
Home page
Re: Color next to our player name
Posted: Apr 10, 2019, 3:08 PM

Well, my point is that there are so few games in the database, most openings are not adequately explored. I've found several dismissed lines that are just as much of a forced a win with good play, statistically against high rated opponents, and were just overlooked.

There's still gold in them there hills...

watsu

Posts: 975
Registered: Dec 16, 2001
Home page
Re: Color next to our player name
Posted: Apr 10, 2019, 3:27 PM

Well, my point is, those openings don't need to be. A sure win is a sure win. Guess I wasn't quite done 😊

In one of my recent wins at another site, I beat a player as P1 who was 83-0 before I beat him. I used an uncommon second move against his first move and by my 6th move the game was out of the database. He resigned the position on the 8th move. Had he not resigned, I would have won on move 13. On a DB move line, I probably would have won on move 12, but there's only one game in the DB to base the assessment of move 6 on, so don't quote me on that. So, I got rating points for winning on a line which is not currently in the DB. What does that prove about my ability to gain rating points in general? Absolutely nuthin' - because I could have followed the database move for move and achieved the same rating result.


Message was edited by: watsu at Apr 10, 2019 3:59 PM

I don't want to play non tournament rules games. If you take one of my unrated invites, play tourney
haijinx

Posts: 64
Registered: Jan 20, 2019
From: Salem Oregon
Age: 48
Home page
Re: Color next to our player name
Posted: Apr 10, 2019, 6:01 PM

So, to make sure I understand...

You used your superior ability to create a new and improved winning line that was not in the database of opening practice.

This happened because your basic understanding of opening principles outpaced that of your opponent, thus you could unleash your creativity on him. He relied on an essentially rote understanding of practice (DB), you won on theory.

Independent of the database.

This really makes my point, no?

That at the higher levels, the database can become a crutch...one with an uneven base at that...and that crutch can be kicked out from under you by someone with a deeper understanding of the game. Like karlw for example.

Furthermore, the database is rife with errors. I routinely re-evaluate everything the DB tells me about a position.

And that's just human errors...then there's shifts, where the position would be the same as another if it were centered on a different white stone.

Beyond that, it's common for both shifts and other translations of the board to have different common lines and winning percentages in those lines. That's not human error so much as human nature.

Different winning percentages, for realistically the same position.

When you combine these factors with the broken rating system, it's not so easy to always find the best lines in the DB...even if they were really the best lines in play.

And, as you've proven, better understanding of theory can and often does still win out.

watsu

Posts: 975
Registered: Dec 16, 2001
Home page
Re: Color next to our player name
Posted: Apr 10, 2019, 6:09 PM

No, P1 wins out in master level TB Pente. Period. As best I can tell from a quick look at the one game in the DB for move 6, I might well have won a move more quickly with the move in the DB. There are usually at least 2 or 3 sure win second move options for P1 no matter what P2 does. Why do I need to be creative as P1 or know more than one of them? Why did I decide to be creative on my 6th move in that particular game? Who knows. My opponent doesn't have access to the DB and was playing as P2. He lost the game in 13 moves due to the P1 advantage. He beat me as P1 in roughly the same number of moves as I beat him, also using a nonstandard second move, one which was actually more nonstandard than my second.

I don't want to play non tournament rules games. If you take one of my unrated invites, play tourney
haijinx

Posts: 64
Registered: Jan 20, 2019
From: Salem Oregon
Age: 48
Home page
Re: Color next to our player name
Posted: Apr 10, 2019, 7:13 PM

I think this is spiraling into a wider argument.

Let's fall back to the agreed upon position...

There is rapid inflation in TB Pente ratings system.

This has led to inflated and inaccurate ratings.

A major contributing cause of this is the failure to rate the majority of games (IOW split sets). Elo is designed as a continuous rating system and this makes it not work right.

Any necessary correction will cause ratings contraction and this will affect some players more than others, both in terms of ratings and in psychological effect.

However, I don't see how reluctance to play could get worse and hoarders will hoard in any system. I prefer karlw's outlook on ratings FWIW. I'd also add that ratings are always bogged down in your history. Improving players are by definition underrated. Either you're taking from someone or someone's taking from you. Best to not be attached to the number and just work the game.

I'd also like to suggest we can agree that:

Non-live testing of alternate methods using the first quarter 2019 dataset is an unobtrusive way to compare those systems.

A group of people to look at the results and use their aesthetic judgment as to which is resulting ratings list is best. Elo goes on about the ultimate judge being the players who assess the system's fairness and how well it matches their own subjective judgements of relative playing strengths.

watsu

Posts: 975
Registered: Dec 16, 2001
Home page
Re: Color next to our player name
Posted: Apr 10, 2019, 7:25 PM

Well, I remember how reluctant and selective master level live game players were to play sets against lower rated players when a split penalized them - circa 2003. Quite frankly, I can't say I blamed them at all. It's even worse with TB games and then the DB is the final nail in the coffin. Honestly, it doesn't make that much difference to me. TB Pente is broken in favor of P1, which is why I rarely play it anymore, regardless of whether or not we have a continuous rating system which has a smaller ratings spread. Maybe more people will realize this when ratings contract again and then start playing fairer variants, but more likely not.

I don't want to play non tournament rules games. If you take one of my unrated invites, play tourney
karlw

Posts: 915
Registered: Mar 7, 2006
From: Eugene, Oregon
Age: 31
Re: Color next to our player name
Posted: Apr 10, 2019, 7:35 PM

I think it's possible to reconcile some of watsu and haijinx's disagreements.

For example, watsu makes a good point about being able to achieve a rating far above your "natural" ability in turn-based play by deep study of a good database, such as the one here at pente.org (there are many times when I wish I could just remove all of my games from this database, and maybe those of a few other players :P). But can't we agree that the skills you use when analyzing those lines are skills that could translate to making you good at OTB pente? Unless you're just blindly copying the moves of a handful of top players, in which case (as haijinx pointed out) you will undoubtedly walk into a number of traps, because even top players make mistakes. To really get the most out the db, you have to be critical of every single move and every single player.

To bring in another chess analogy, when grandmasters know that a player is "booked up" (as in they have memorized openings 30 moves deep, or have a team of seconds that has done the work for them), they will intentionally try to get their opponent "out of book" early by playing an uncommon (but equally or nearly equally strong as the "book" moves) move, forcing their opponent to use only their own mind. Since pente is a much simpler game than chess, this is becoming increasingly difficult as more and more openings are rigorously analyzed, but it's still possible. I would argue that my K10 K9 N10 M8! achieved this for a long period, as did zoeyk's K10 K9 N10 N9!, and many other moves in history that were once the unexplored frontier of pente but are now nearly solved to death (I'm looking at you, wedge). This is how you separate the database copiers from the true masters.

It's no use going back to yesterday, because I was a different person then.
watsu

Posts: 975
Registered: Dec 16, 2001
Home page
Re: Color next to our player name
Posted: Apr 10, 2019, 7:51 PM

"as did zoeyk's K10 K9 N10 N9!"

Dmitri King told me he successfully played this second as P2 (or an isometric variant thereof) in Vegas in '04 and therefore was kicking himself that he lost to it as played by Richardiii in Tourney 5a round 8...
https://pente.org/gameServer/pgn.jsp?g=34194139631912

I'm not at all saying there's been nothing new, revolutionary or outside the DB box brought to the table by you and zoeyk. You guys are among the names whose games I always examine. It is indeed becoming increasingly more difficult over time to do this, however. And, what works in live games doesn't work as well in TB games, I think we can agree.

I don't want to play non tournament rules games. If you take one of my unrated invites, play tourney
madmike

Posts: 102
Registered: May 27, 2014
From: Houston
Age: 64
Re: Color next to our player name
Posted: Apr 11, 2019, 7:30 AM

The OP wanted to know about the color next to our name. Rainwolf answered that question in short order.
But it still remains the issue for those lower rated classes are only separated by a relatively few rating points.

** I think there may be a gray class but was to lazy to discover when it came to be.
1000-1399 = green..... spread of 399
1400-1699 = blue...?...spread of 299
1700-1899 = yellow.....spread of 199
1900-infinity = red ?..spread of infinity

Pente_gon is going to break 3000 in the TB setting in a few weeks. surely after generating a spread of 2100 he should have his own color.

To me is seems fairly obvious that we more rating classes, bigger rating classes, or just no rating classes.

Then we tackle the issue of how, why, and why not to tame rating inflation.

madmike

Posts: 102
Registered: May 27, 2014
From: Houston
Age: 64
Re: Color next to our player name
Posted: Apr 11, 2019, 7:39 AM

If "P1 wins out in master level TB Pente. Period." Why would two master level players even bother. They both get to be P1, and its going to be a split.

watsu

Posts: 975
Registered: Dec 16, 2001
Home page
Re: Color next to our player name
Posted: Apr 11, 2019, 10:52 AM

"Why would two master level players even bother. They both get to be P1, and its going to be a split."

SJustice and Virag = co champs in live tournament here
SJustice and Rollie Tesh = co champs in 1984 OTB
Dmitri King and Gary Barnes = co champs in 2003 OTB
^ live co champs, not TB co champs. At Its Your Turn back in the old days players would just keep making random edge of the board plays at the end of games until IYT got around to declaring them co champs. Mostly, they master the game and retire. I'm not saying they never make/made any mistakes, but the number of drawn sets vs. won sets at that level is pretty overwhelming and the oddball P2 win quickly disappears with analysis of the game. This is why we have so many variants and so few active old time master players here

I don't want to play non tournament rules games. If you take one of my unrated invites, play tourney
Replies: 47   Views: 21,024   Pages: 4   [ Previous | 1 2 3 4 | Next ]
Back to Topic List
Topics: [ Previous | Next ]


Powered by Jive Software