How To Get Along With KGSRating Math

    Keywords: Online Go

If you read this, then you are probably unhappy right now on KGS, because your rank has not moved up - although you played 99 games in a row and won all but a single one. You wonder, why you are treated that badly by the KGS ranking system. You have become stronger and you know it, yet KGS denies you the wished-for promotion. Probably you have been visiting different pages where the rating system is analyzed thoroughly with a lot of formulas and tables. If you have no affinity for maths then you probably quickly clicked those pages away.

Actually it is not the fault of the rating math that your rank is stuck but a misunderstanding about how to use it in the right way. Before you give up on KGS, turn your rank off and start lamenting over the ranking system in chats or kibitzes, read on - this simple explanation here might help you to evade petrification and get a fluid, accurate rank!

First of all, easily said, KGS takes the last few months, lumps all your rated games together and counts, how many of them you have won. If you won 50 %, you remain on your rank. If you won about 2/3 of your games, you promote. If you lose 2/3 of your games, you demote. It is actually quite simple and also reasonable. But there are some traps which can lead to your rank not moving up or down any longer. The main mistake lies in the uncontrolled playing of rated games.

Imagine the following situation: Go player Bryan has a few days off. In those days he really plays a lot of go since he wants to get stronger. He plays 10 games every day, for five days. So after 5 days he has 50 rated games in the list. And his average was about 50/50. So the five holidays pass by and Bryan has no time any longer to play that much. Since heís ecstatic about go, he manages to skip in at least one rated game per day.

For some reason - sometimes it helps to not only concentrate on go but to also do other things - he starts to play better and even win his games. We all know those times: You get on a streak and win and win and win (until it stops). So Bryan plays a game a day for a while and he loses not a single one. He even plays on like this for weeks, but his rank just would not move. After a while he gets frustrated, looks for help and hopefully finds this page.

The problem is, that Bryan had 50 rated games in his list, 25 of them won, 25 lost, which KGS counts in all the time, while Bryan plays more games. After 50 work days, Bryan has accumulated 100 games altogether. 50 of those games have been played on 5 days, 50 more in 50 days. KGS does not care about when which game has been played (it cares a little, but these are details you can do well without for our purposes), it just takes these 100 games and looks how many of them Bryan did actually win. If Bryan for some not-dreamt-of reason manages to do a 50-game-winning streak, he will finally have a count of 75 wins to 25 losses - he will only have won 3/4 of his games - statistically.

Bryanís mistake was to not play regularly, but to play in a chaotic way, many many games in just a few days, and then very little games in many days. Like this, KGS rating math does not work properly. The cure is simple: All Bryan would have to do is stop playing as many rated games as possible in a short time, but just play the same amount of rated games every day.

For example, if Bryan just plays one rated game each day no matter if it is a working day or not, the statistics will always be balanced and the rank will not be in disorder. I am not making this up - just try it out. Create a new account right now and make it an ironclad rule: Play one rated game every day, no less, no more. If up to now you really felt bad and annoyed about the KGS rating math you might experience a pleasant surprise.

To tell the truth: If you want to push up your rank in only one day out of vanity because you want to show off, or because you dream of strength you haven't acquired yet, better wander off to another server. Your rank on KGS is supposed to grow slowish, since the meaning of a rankings system is not to show how strong you are today but how strong you are generally, over some weeks or months.

If you use the system I just described and start to play better - that is if you win 2/3 of your games for a while - your rank will gradually go up. It will take about a month until you get promoted, but not longer. If you donít get promoted, just take a look at your game list, examine your wins and losses and you will find out that your feeling of being stronger maybe just was a bit too subjective and your average is below the desired 2/3 winning percentage.

If you are really getting stronger, and if you prove that by winning more games, and if you play your rated games in the right way, fitting the KGS rating math, then you will definitely get your promotion within a few weeks. And since you waited for a while to get there, KGS will reward you: In case you play really bad for a week or two, it will not immediately cost your rank. It will take a month of constant sucking to kick you down, so you have time to recover from your temporal weakness and preserve your rank.

The key to not suffering under the KGS rating math is to play regulated. It need not be exactly one game a day, you should just stop yourself from heaping up large amounts of rated games like in the example above unless you are sure you can play that much all the time. If you treat the KGS rating math properly like this over a while, you might well come to like it best of all possible rating systems.

- HonFu


RobertJasiek: There is another escape: Don't play for some 3 months, then win a few games. However, neither fits the reality that some players like me do play many games on some days and no or only few games on other days. Instead of forcing players to adapt to the rating system, the rating system ought to adapt equally well to all playing per time distributions. The rating system is supposed to be a model of the players and not vice versa.

HonFu: If you play a bunch of games on one day and only a few on another day, and if you just naturally go on like this, then the rating system will work just fine. It does not really make a difference! The problem of rank petrification only occurs when you play really a lot for a while, then suddenly start to play very little.

If for example you can't stand to have lost a game and want at least one win before you log off and therefore play one blitz game after another, like 20, 30, then you will not only lose one after another, but also get your rank stuck. As long as OVERALL there is some kind of balance in the number of your games, KGS rating math will work just as it is supposed to be.

Making a 3-month-break is just a way of evading the rating system. You will have a flexible rank for a short while, but on the other hand no reliable rank at all. Turning off the rank would be about the same. Of course from an idealist perspective, a rating system should serve the players, not the other way round, but any statistic system needs data, that is regulated somehow. All that is asked for by KGS rating math to give you a fitting rank is that you play games without going on a blitz frenzy. I think that is not too hard on most of the players.

Anant1111: I have been playing so many rated games and takes forever too level up! On the other-hand my brother plays occasionally and he is on a higher level than me!!!

HonFu: When you start a new account, the rating is very fluid, because there is not any statistical data yet. So when someone starts his account after you, his rank will change more quickly. After 6 months he will be just as "stuck" as you are. But, as described above, it is not really stuck, since your rank WILL move up steadily if you go on winning. It will only get stuck if you play a million games in a few days, then change back to a normal playing rhythm.

Luis Sousa: Am I missing something here? Since when will winning 66% of your games in a month rank you up and winning 75% won't?!?!?! OscarBear?: The winning percentage is not completely relevant. If you are a strong 2-dan who has a winning percentage of 80% against a weak 1-dan, that could be entirely expected and not cause much of a shift at all. In contrast, a weak 2-dan beating a strong 1-dan 80% of the time, yes that would be different.

Luis Sousa: I was just pointing out the flaw of his example!

How To Get Along With KGSRating Math last edited by on November 21, 2014 - 19:26
RecentChanges · StartingPoints · About
Edit page ·Search · Related · Page info · Latest diff
[Welcome to Sensei's Library!]
Search position
Page history
Latest page diff
Partner sites:
Go Teaching Ladder
Login / Prefs
Sensei's Library