More on A Curious Case Study in KGS Ranks

[RobertJasiek]

Quotation reference:
viewtopic.php?p=161534#p161534

If it takes less effort to reach a rank, there would be more players with that rank.

No. If more players increase, they will find tougher competition on the next higher rank, and so they will also decrease again, on average.

If it takes less effort, it simply means that also the players so far affected by too great rating stability can improve to the next rank to seen if they can hold their new level.

With the situation being as it is, it certainly does not take a "superhuman effort" to reach 5d. There are many players who are 5d on KGS, they reached it fair and square, and I have hard time believing that they are all X-Men.

You still do not understand: different players are NOT treated equally by the KGS rating system, but frequently playing players are given much tougher opposition by the KGS rating system to improve.

according to this particular rating system, you are not yet strong enough to reach 5d on KGS, pure and simple.

Wrong.

It is easy for me to win ca. 70% (and create lots of 8 ~ 12 wins in a row series) as a KGS 4d if

- I play almost exclusively for the purpose of fighting against the KGS rating system,
- I play only even games with komi,
- I never play 10s games,
- I never play when tired or under stress of work.

Study my KGS games and you can notice that

- I play MUCH weaker in them than in real games (also because the rating system creates such a great frustration),
- the distribution of scores incl. the resigned games assessed by positional judgement favours myself.

I would demonstrate it immediately if only I had the necessary successive weeks for overcoming the mentioned points AND the necessary time during the successive weeks for playing several games per day. Since currently I do not have them, every attempt to reach a higher rank would fail (I might win 65%, but this is not enough), and so long I prefer to play also for fun and when tired, and you see something like 50%.

best to switch to a server

It is much easier than that: I could switch my KGS account, but (so far) I prefer to see the system changed rather than to make the system even worse by producing more noise, which KGS then fights by keeping the rating system even stabler.

***

Mef, ez4u: looking at rating graphs is fun, but you overlook that my rating graph is VERY heavily influenced by the rating system causing very great frustration. If you really want to assess my KGS playing, study my games and compare them to my real world games.

[daal]

You play when you are tired or otherwise stressed? I thought I was the only one! My KGS opponents always play at their best.

[RobertJasiek]

Rather there are players playing only at their best and others playing in various conditions. Especially for the latter, if they are frequent players, the rating system is too tough, because it does not forgive fun games because it puts them into the evaluation of the player's history. This combined with the great stability of the system for frequently playing players makes it too tough.

As a frequently playing player wishing to improve a rank, playing only at one's best is necessary, IMX.

Contrarily, infrequent players can play in every condition, and if they want to improve a rank, they play seriously just for a very short time (such as three days in a row). They need not wait three or more months until such would be successful. They are hardly affected by their KGS rating history. For them, KGS forgives their fun games.

E.g., if I created 30 accounts and rotated using them after each game, KGS would forgive me my fun games. With only 1 account, KGS does not forgive and weighs my history too heavily.

[HermanHiddema]

Rather there are players playing only at their best and others playing in various conditions.

And players playing only at their worst, of course (e.g. those that usually play only in the evenings after a long tiring day at work).

So perhaps more generally, it would be useful to say that different players have different playing strength stabilities.

Ideally, a rating system detects how stable a player generally is, and adjust the volatility of their rating accordingly. That would serve all players.

[leichtloeslich]

Ideally, a rating system detects how stable a player generally is, and adjust the volatility of their rating accordingly. That would serve all players.

Glicko-2 (an improved ELO) has precisely that: (first paragraph from http://www.glicko.net/glicko/glicko2.pdf)

Every player in the Glicko-2 system has a rating, r, a rating deviation, RD, and a rating volatility, \sigma.
The volatility measure indicates the degree of expected fluctuation in a player's rating. The volatility measure is high when a player has erratic performances (e.g., when the player has had exceptionally strong results after a period of stability), and the volatility measure is low when the player performs at a consistent level.

The RD basically means we're 95% confident that the true (unknown) rating of the player is in the interval
[r-2*RD, r+2*RD].

[RBerenguel]

Ideally, a rating system detects how stable a player generally is, and adjust the volatility of their rating accordingly. That would serve all players.

Glicko-2 (an improved ELO) has precisely that: (first paragraph from http://www.glicko.net/glicko/glicko2.pdf)

Every player in the Glicko-2 system has a rating, r, a rating deviation, RD, and a rating volatility, \sigma.
The volatility measure indicates the degree of expected fluctuation in a player's rating. The volatility measure is high when a player has erratic performances (e.g., when the player has had exceptionally strong results after a period of stability), and the volatility measure is low when the player performs at a consistent level.

The RD basically means we're 95% confident that the true (unknown) rating of the player is in the interval
[r-2*RD, r+2*RD].

Interesting!

[pwaldron]

It is easy for me to win ca. 70% (and create lots of 8 ~ 12 wins in a row series) as a KGS 4d if

- I play almost exclusively for the purpose of fighting against the KGS rating system,
- I play only even games with komi,
- I never play 10s games,
- I never play when tired or under stress of work.

It is easy for you to win 70% of your games because your rank is that of a stronger-than-average 4d. If you only play even games against other 4d players then on average you are at an advantage.

[RobertJasiek]

As 4d, I play against 3d+.

[Mef]

Mef, ez4u: looking at rating graphs is fun, but you overlook that my rating graph is VERY heavily influenced by the rating system causing very great frustration. If you really want to assess my KGS playing, study my games and compare them to my real world games.

Robert, you seem to be confused. We are not looking at your rating graph. We are looking at your actual games, and your actual win rates. Your claims of long strings of high win rates on kgs simply don't exist as you say they do either in time or in games played.

How well you perform in real life games should have as much weight on your kgs rating as how well you play backgammon or how well you cook a ribeye.

[skydyr]

Rather there are players playing only at their best and others playing in various conditions. Especially for the latter, if they are frequent players, the rating system is too tough, because it does not forgive fun games because it puts them into the evaluation of the player's history. This combined with the great stability of the system for frequently playing players makes it too tough.

If you are concerned that your fun games are dragging down your rank, have you considered using free games instead of rated games for that purpose? Or does the decision that a game is fun and not serious come after the game has started?

[Bantari]

Well, I am in a writing mood, so let me take you apart on the points you are trying to make. I highlight the points I am trying to make, to make it easier for you to concentrate on the important stuff and not on supporting verbiage.

If it takes less effort to reach a rank, there would be more players with that rank.

No. If more players increase, they will find tougher competition on the next higher rank, and so they will also decrease again, on average.

No.
If it is easier to go up ranks, more people will go up ranks.
If more people go up the ranks, there will be more people in the higher ranks.
If there are more people in the higher ranks, the system will become more top-heavy.
There is no way around it.

You still do not understand: different players are NOT treated equally by the KGS rating system, but frequently playing players are given much tougher opposition by the KGS rating system to improve.

Rating does not do the pairing. There might be a quirk of the automatch feature which behaves as you say, I have no clue. No clue if this is even what you mean. But rating itself does *not* take care of who you play. It is either your choice or the automatch. This really makes me doubt you understand what rating is.

Looking at your game history: you are a correct 4d.
Start beating 5d players regularly (or just play them more and beat them), stop losing to 3d players (important!), and you will get to 5d. That's the trick.
I don't see how you can expect the system to make you 5d when you lose so often to 3d players. Not to mention 4d players.

according to this particular rating system, you are not yet strong enough to reach 5d on KGS, pure and simple.

Wrong.

There are plenty of 5d players on KGS, not to mention higher ranks.
When somebody plays strong enough to reach 5d, they reach 5d, or there would be very few 5d players.
I have myself seen players who started as lower ranks and made it to 5d and beyond. If you play strong enough, you get there, trust me.

It is easy for me to win ca. 70% (and create lots of 8 ~ 12 wins in a row series) as a KGS 4d if

- I play almost exclusively for the purpose of fighting against the KGS rating system,
- I play only even games with komi,
- I never play 10s games,
- I never play when tired or under stress of work.

Study my KGS games and you can notice that

- I play MUCH weaker in them than in real games (also because the rating system creates such a great frustration),
- the distribution of scores incl. the resigned games assessed by positional judgement favours myself.

What does this have to do with anything?
I play much weaker online as well, but it would never ever in a million years occur to me to blame it on the rating system. Your opponents also play tired, or drunk, or whatever. How many of your wins can be contributed to such external factors? You seriously assume you are the only one who loses games because of stress or too fast timer?

Robert, this really sounds like excuses. The fact remains - there are plenty of people who made it from 4d to 5d and beyond on KGS, so if you are strong enough it is certainly possible. As a matter of fact, people graduate up the rating ladder all the time on KGS, when they play strong enough to reach the next level from wherever they are. This includes frequently playing people, seldom playing people, tired people, drunk people, and anything in between. Even you cannot refute this simple truth which is easily verifiable.

Since you don't seem to be able to make it, to me it clearly means that you are not strong enough. Now, this might be for many reasons, one of which might be mood you are playing or whatever other distractions, and you could surely try harder and make it. But this is not the fault of the rating system.

And this does not change the fact that - the way you play, you play as 4d, period. Maybe all these 5d players could really be 6d players if they tried harder, or maybe they would fall down to 4d if they tried less, who cares? The fact remain - the way you play, you're a 4d. Just like the way they play, they are 5d. And so on.

I would demonstrate it immediately if only I had the necessary successive weeks for overcoming the mentioned points AND the necessary time during the successive weeks for playing several games per day. Since currently I do not have them, every attempt to reach a higher rank would fail (I might win 65%, but this is not enough), and so long I prefer to play also for fun and when tired, and you see something like 50%.

This is your choice, and I see no problem with that. But again - do not blame it on the system. There are many casual players who made it from 4d to 5d, or from 5d to 6d, or whatever, on KGS, so this too must be possible when you are strong enough.

My circumstances in life right now are such that I also play at a rating a few stone below what I could accomplish. So? This is how I play, and once I start trying harder and playing stronger - my rating will go up. I mean - this is how the system works - it ranks you based on how you play. If you play like a 4d, this is where you land. When you start playing like 5d, you will eventually graduate. I am not sure why would you even want something different, it would be a mess.

It seems to me that what you are saying is that you can try hard and play like a 5d for short spurts, and therefore the system should upgrade you to 5d. This is not how it works, nor should it. Because, by definition, then you fall down again and play like 4d or even weaker, for whatever reason - tiredness, distractions, fun, whatever. Well, this apparently does not cut it, and I for one am glad it does not. Everybody can have a good weak or put extra effort to put up a short spurt - this does not make them a rank stronger. If the rating system was so sensitive, lots of people would be overrated out there, and then lots of people would be underrated as well (after a bad week), and I for one am glad it is not so.

best to switch to a server

It is much easier than that: I could switch my KGS account, but (so far) I prefer to see the system changed rather than to make the system even worse by producing more noise, which KGS then fights by keeping the rating system even stabler.

There is nothing wrong with creating two accounts, one for serious games one for casual ones. Many people have that.
I know you would rather change the world so that it makes sense to you, but you are not the only user and you will always have to compromise with what some others might want or prefer.

If you really play in different "modes" - like sometimes fun games, sometimes serious games, sometimes when you're tired or stressed, sometimes when you try very hard - this almost begs for multiple accounts. The rating system ranks you based on all your rated games, and it it not the rating system's job to cherry pick only the "proper" games from your history, the ones you like, and rank you based on those. I am not sure why you would ever expect that. This is what multiple accounts are for, or at least the non-rated option in game setup.

You are smart enough to know all that, no?

[Bantari]

As a frequently playing player wishing to improve a rank, playing only at one's best is necessary, IMX.

Depends on your rank, no?
If the system rated you as 2k, you will certainly not have to play your best to get to 1k, and even to 4d. But 4d is where your level sits as long as you play the way you play. There is absolutely nothing wrong with that. There is nothing wrong with having to play at your best when you want the system to rank you at your best.

Again - what you seem to be expecting is that you wish to play "not your best" and yet you wish the system were to rank you "as if you were playing your best". And you blame the system if it does not do so. This is a very silly expectation and not a very smart attitude neither. As far as I know, no system will cater to your whimsy like that.

[skydyr]

As a frequently playing player wishing to improve a rank, playing only at one's best is necessary, IMX.

Depends on your rank, no?
If the system rated you as 2k, you will certainly not have to play your best to get to 1k, and even to 4d. But 4d is where your level sits as long as you play the way you play. There is absolutely nothing wrong with that. There is nothing wrong with having to play at your best when you want the system to rank you at your best.

Again - what you seem to be expecting is that you wish to play "not your best" and yet you wish the system were to rank you "as if you were playing your best". And you blame the system if it does not do so. This is a very silly expectation and not a very smart attitude neither. As far as I know, no system will cater to your whimsy like that.

I don't think it's directly applicable to go ratings, but a golf handicap is basically the average of the top 10 of your last 20 scores, plus some massaging for particular holes with high stroke counts. What a high stroke count is, of course, depends on your handicap. As a result, your handicap isn't actually your average play, but more of a figure for determining what your score should be if you play well through a given round.

Now if only there were some sort of stable criteria against which go skill could be judged...

[moyoaji]

I don't think it's directly applicable to go ratings, but a golf handicap is basically the average of the top 10 of your last 20 scores, plus some massaging for particular holes with high stroke counts. What a high stroke count is, of course, depends on your handicap. As a result, your handicap isn't actually your average play, but more of a figure for determining what your score should be if you play well through a given round.

Now if only there were some sort of stable criteria against which go skill could be judged...

The point of that system is to remove volatility, which is also what KGS strives for in its ranks. The goal is to give you a fair handicap against opponents so the game you play is close to even. In fact, that is the goal of all ranking systems.

Honestly, someone should not be upset about having a rank lower than what they deserve. Instead, their opponents should be the ones getting upset. If I was ranked 5 kyu on the KGS that gives me a one stone handicap that I don't deserve against 4 kyus and takes away the stones from my handicap against 6 kyus.

I went on a losing streak in November where I fell from high 3 kyu back to mid 4 kyu. After that happened I started to win more games. I am again back to 3 kyu, but now I'm a bit nervous to play because I don't have the crutch of an extra stone. I do understand where Robert is coming from, because when I play at my best I can play to 3 kyu ability - maybe even 2 kyu, but I know I don't always play at my best.

That being said, it would be horribly unfair if the KGS suddenly raised me to 2 kyu just because in some of my games I play that well. Your rank should be a fair average of all of your games played recently. Would I want that just to be able to say "I'm 2 kyu on the KGS?" No, I honestly wouldn't, because it would make my KGS games less fun and make me more likely to lose.

I also would not want to have a rank inflation on KGS where everyone goes up 1 stone just to make players feel better about their ranks. The fact that you have to work hard to earn a rank on KGS is the reason I base my rank on that server. I would love to have regular AGA tournaments I could attend to get ranked there, but that is not my reality. Instead, a stable and reliable rank system like the KGS is what I want to be able to point to and say "I'm this rank."

[Bantari]

Hi again, Robert.
Sorry for posting so much, but the more I think about what you say the less sense it makes to me.

So, basically, putting aside what I said before (all true, but not focused enough), let me take two specific points here.

Point #1.

• Lets look at your various playing "modes" hypothetically. Lets say that: when you play only casual and fun games you play like 3d, when you play seriously you play like 5d, and when you play a mixture of both modes, you play like 4d. This is how it can hypothetically look when taken your history into account, and this is pretty much what you are saying as well. Now what you seem to want is a system which lets you generally play in the mix mode but ranks you as if you were constantly in the serious mode. This is not reasonable, and no system should do that. This would mislead your potential opponents as well, so its not good.

Point #2.

• Alternatively, you want a different system - one which lets you play in your mix mode, the way you do, but calls this level 5d rather than 4d - which is basically a label change. This can certainly be arranged, by applying the system you advocate for, or whichever other way - but then you will be stuck with another problem: you will then play in spurts like a 6d according to that system (because you will still have periods of "just serious" play, or at least be able to have such periods) and we will go through the same dance but with you complaining that you should be 6d instead of 5d. This too is not reasonable.

So what is left?

• The only other option is to implement a system which is so volatile that it changes your rank after every few games, which is the worst of all, imho. Because then you are almost never at your real rank. After a few wins, you get to 5d, but by then your winning streak is done and you really play like 4d again - which makes your rank inappropriate at that moment and misleading. So you lose a few and get back to 4d, but by then your losing streak is over, and you play like 5d who is rated 4d, which is also no good. I really cannot imagine why anybody would seriously want such system. Sure, it might be fun, "roulette", but seriously... I would not wish to play on a server like that. Besides, ranks would become very meaningless on such server, just like they don't mean much on Tygem.