For whatever reason, I simply cannot seem to win against Korean players. A year ago I was 7k on Kgs and 9k on Tygem. Now I am 3k on KGS and still 9k on Tygem. Something about the playing style on the Korean servers gives me serious trouble. Has anyone else experienced this problem? I am considering playing exclusively on tygem and wbaduk until I break through whatever I suck at, but it is daunting to face so much losing. I feel like I did in my first months playing go.

Who has similar or contrasting experiences they would like to share?

If you are losing constantly on that server, why isn't your rank going down?

On KGS your rank is a statistical prediction of how well you'll perform against other players. Many other ranking systems do not pretend to make any sort of predictions, so the ranks there are a great jumbled mess, which some people really like. As I understand it, Tygem is one of those servers, which is why the "why don't ranks on Tygem make any sense" thread comes up every few months.

My rank probably doesn't go down because I rarely play on these servers. Maybe 2 or 3 games every several months. Anyway it is kind of disheartening when you have gotten used to thinking of yourself as a 3Kyu and get beaten by 20 points by a 10K.

I disagree. The model for the rank is simply different on Tygem than KGS - it's not random, with the exception that you are allowed to provide an initial rank up to some point. As you play games, the effect of the initial rank you selected decreases.

The Tygem model is simpler, and much like you might play at a local club - win a large percentage of the recent games you've played and you rank up. Lose a large percentage and you go down.

Though the KGS system uses a maximum-likelihood based approach:
1.) A maximum likelihood approach is not necessarily better than the simpler Tygem approach, because it requires you to model the importance of your historical games. This varies widely from scenario to scenario (eg. have I been studying over the last 6 months, or have I just been on vacation).
2.) It's not purely "statistically based". For one, there are anchors that "tie down" the ranking system, and also there are variant factors in the model (http://www.gokgs.com/help/rmath.html), such as the constant k that changes.

Because the math looks cool in the documented KGS approach, people assume that it's a "statistical approach" and therefore more sophisticated than more basic systems. In reality, the KGS model is simply one particular model (which is not even fully documented), which includes several assumptions such as these ranks and how importance the history of your games is (eg. 180 days for game history to be deleted is arbitrary).

The Tygem ranking system is not perfect. Sometimes you do end up playing stronger or weaker players. But it's neither fair nor accurate to act like it's a "great jumbled mess" when really, it's simply a different model of ranks with its own assumptions, just like KGS has. It's simply a different model.

It probably adds to the confusion in that the style of Tygem players, overall, is different than that of KGS players. Style has several dimensions, whereas a numerical rank is only a single dimension. Because of this, there exists confusion when people see Friend X is stronger on Tygem than KGS and Friend Y is weaker on Tygem than KGS. From the perspective of someone that plays mostly on Tygem, the same phenomenon can occur - the ranks on KGS can seem like a "great jumbled mess" (albeit a jumbled mess in which people's ranks are less volatile).

_________________
be immersed

I think the different rank systems just are just systematically shifted. If you look at http://senseis.xmp.net/?RankWorldwideComparison things might become more clear.

My KGS rank is e. g. around 5-6 k, while I'm around 11k on wbaduk. 11k is exactly in mid of the rank interval 14-8k given for wbaduk in the rank comparison table, if the KGS rank is 5k. Also with your 3k on KGS and 10-5k on wbaduk you still fit into the usual range.

All in all - nothing to worry about, it just feels a bit unusual.

It's possible that this hypothesis has merit, but given that there are cases such as yours where the KGS rank is stronger than the wbaduk rank and also cases we've seen in other threads where it's the other way around, I feel that it's difficult to capture the nuances of various strengths in go (composed of multiple dimensions: fighting, opening, positional judgment, etc.) in a single number/rank.

As a result, when you have two populations that have, by and large, a different set of strengths, it makes sense that you'll have cases where player A beats B, B beats C, and C beats A on a consistent basis, which lends to the hypothesis that it is difficult to directly compare the two ranking systems.

Still, I agree with your concluding point: it's nothing to worry about

_________________
be immersed

Kirby, the KGS approach is statistical in that a rank is a prediction about a future W/L rate, and each new game is taken as a piece of information to improve the prediction. You're correct that many different statistical models are possible. Remi Coulom, you may remember, did a comparison of multiple models. The KGS system performed nearly as well as the best system (WHR) in 1/3 the time.

However, the Tygem approach is not statistical in that it makes no predictions. So people ask, "why am I losing to a 10k", and the answer is simple; the system does not make any predictions about how a 10k will do against a 9k. It is not based on statistical inference at all.

As you say, the Tygem system is very similar to club play. When we play at our local clubs, a disconnect between the handicap two players use and the win rate is easy to grasp, because we know the social and political context behind it. He's too modest, he's too arrogant, he's trying to challenge himself, he's been improving very quickly, he's going through a rough divorce... it's a non-statistical system, but we also know a lot about what explains the jumbling, so no one makes posts about why they can't beat the 10 kyu players at, say, the Seattle Go Club.

As I said, some people like the fact that Tygem does not use a statistical ranking system. You may be one of them. But the fact that people assume Tygem ranks are predictions leads them to make incorrect assumptions about the informational content of the ranks.

Doctor, it hurts when I do this...

_________________
- Brady
Want to see videos of low-dan mistakes and what to learn from them? Brady's Blunders

I agree there is a learning curve when playing tygem but 9k to 3k sounds wrong to me. I would start an account at 5kyu and try there maybe get past someof the sandbaggers.

I personally like tygems system because it rewards playing, while on kgs sometimes you feel stuck.

I don't agree. The model is just different. If I am 1d on Tygem and I've played a number of games, having won the threshold percentage of games, the system infers that my rank should be increased. And this inference is indeed statistical, because it's based on real data - the games you've played.

Both models take, as input, result data from games. And both models adjust your rank based on results.

The only element that is not "statistical" about Tygem's method is that it allows for a user-defined starting rank, intended to speed up the process of getting to a rank where you can play people of your same strength.



This is only a single type of "statistical" rank prediction, and it's not actually a completely accurate statement - each new game is not taken to improve the prediction. Only games from the last X days, where X has been arbitrarily defined.

In the same manner, it can be argued that Tygem's method is statistical in that it takes the last Y games (again, where the number is arbitrarily selected), computes the percentage of wins/losses, and adjusts the rank accordingly to improve its prediction of your current rank. Your argument that Tygem's ranking system is not statistical is based on the idea that the system "makes no predictions". If adjusting a user's rank based on game result data is not a prediction of one's rank, then what is?

tl/dr: Tygem's ranking system does make predictions, because your rank is a function of the game data. What else do you want to define it as a system which makes predictions?

_________________
be immersed

Just so we have a definition of the term we're talking about here, here's the first excerpt from wikipedia's page on statistical inference:



WRT the Tygem rating system, the system draws conclusions about your rank (the rank that it assigns to you) from data that is subject to variation (your game results). The system could be described as a procedure used to draw conclusions from datasets (your games). The system makes an estimate of your rank based on data (the games you've played).

I really can't see how you don't see the ranking system as a form of statistical inference, albeit simple.

Bottom line: both systems model strength to suit you with a good opponent, and assumptions have been made in the design of both models.

_________________
be immersed

20 wins and you rank up. 20 losses and you rank down.

There's nothing statistical about that. That's just simple counting.

19 wins followed by 1 deliberate loss and you can sandbag forever.

If statistics were involved, then the one loss wouldn't keep you from ranking up.

It's a very limited data set but those wins/losses are still statistics.

P.s. What's your source for how the tygem ranking system works? That does sound ridiculous but I've read different things before and I can't find any official pages about it.

That sounds ridiculous because it is not used by Tygem. I think the below is the one used:

From 5d-9d (latest 20 games):

Rank Change: Winrate Required:
+2 100%
+1 75%+
-1 35%-
-2 15%-

From 1d-4d (latest 20 games):
+2 90%+
+1 70%+
-1 30%-
-2 10%-

From 18k-1d (latest 15 games):

+2 90%+
+1 65%+
-1 30%-
-2 10%-

Edit: Replaced "Wrong" with "Not used by Tygem".

Kirby, I'm not going to argue with you about what "statistical inference" means. You seem to know something about how inference works, or you wouldn't be throwing around jargon like maximum likelihood. But at the same time, you are throwing up ink in a way that has poor Splatted saying:



For those who care: any number can be a "statistic". A "statistical prediction" or "inference" is one that applies statistical techniques to data to find the best prediction. "You won 17 games! Here's a gold star!" is not a statistical technique, and the gold star is not a prediction.

This is just plain insulting and completely out of line. Neither you nor said XedOver said Tygem ranks aren't a statistical prediction, you said they aren't statistical because they don't make predictions.



My point was that there are many other ways in which something can be statistical. "You won 17 games! Here's a gold star!" is a statistical technique, it's just a very basic one.

Edit: Or rather, a system that awards a gold star for 17 won games would be statistical.

It seems like we've reached the most rewarding part to disagree on - definitions

For what it's worth, I agree with Kirby and Splatted that the Tygem system is statistical, but would accept that some people may consider sophistication as part of their definitions. I do disagree with jts's initial comment that the Tygem system does not attempt to make predictions though, as I imagine that would be implicit in any ranking system.

jts, statistical analysis doesn't need to be complicated. There are complicated techniques and models that sometimes work well, yes, but it's not a requirement.

To give an example, you can take an elementary language processing course, and one of the first techniques used to determine a maximum likelihood estimate of a word, given a set of training text, is to simply count up the words and take the average to determine the probability of seeing a given word. You can obtain a more advanced model, but it's still true that a basic measurement is a statistical method for estimating uncertainty. Various ranking systems like KGS's, Tygem's, etc., are all statistical methods for determining rank.

But the main point I want to make here is that, just because a given model is more mathematically complicated, it doesn't necessarily make it a better predictor. Overfitting in machine learning is a good example of this.

When math gets complicated, sometimes people forget that the given statistical method is still subject to its own base model and assumptions. Stats are useful, but the only thing that can fairly measure how good a model is track record.

So it comes down to, when you play on KGS or you play on Tygem, in which case do you feel you are having a better playing experience?

This varies from person to person, so it is not accurate to argue that a given model is "more statistical" or "superior". The model simply makes different assumptions, and KGS's happens to use more math.

_________________
be immersed

Sorry, I apologize. Use "statistical inference" to mean whatever you want. All numbers are statistics, and anytime anyone says anything about numbers, that's statistics too. A t-test is just when you g-give a test to a f-fellow with a stutter; and a parameter is about six anahalf feet. The fact that there are multiple threads about how unpredictable Tygem ranks are is surely part of the mystery of Korean play-style, like Kirby says, and has nothing to do with how that server determines ranks.