2 rating questions: beating up on 10k players, is rank loss proportionate to opponent strength

If I’m not mistaken, all ranked games against people from 1k to 10k improve your ranking, meaning that you could limit yourself to playing against people rated 10k and still earn 1-dan rating?

second question: I’m currently 6k. Do I lose the same amount of rating points whether i lose to someone 3 stones stronger, 3 stones weaker or the same rating?

No, the rating points gained or lost are proportional to the strength difference. If, however, a proper handicap is used, then the game is considered equal.


Technically it is possible, but what is it worth? If your rank is not real, then it is just a label with no meaning/value.
If you just want the label, just make a new account, log into it from another browser, play against it constantly and have it resign after 40 random moves. Within the hour you can be 1 dan.

Yeah, it can be done, but what’s the point?

Self-play in ranked games is prohibited on OGS.


Last July I won a game as 1d against a 9k, my rating went from 1962.36 to 1963.26. So if you play and win a few hundred times against 10k players, you’ll become 1d. Unless you get bad habits by playing too often against weaker players and your play deteriorates.


Please don’t promote self-play. It’s not permitted.


Actually no, the range is ±9 from your own rank, so as 6k you can play ranked games with anyone between 15k and 4d. And every win on ranked game will improve your rating (and losses will lose you some rating points), how large that improvement will be does depend on the rating difference between you and your opponent (and handicap)


It is obviously cheating and I am not promoting it, but it does exist as a loophole.
All I am saying is that even if someone was cheating like that, it is a pointless thing for them to do, so I am actually promoting against doing such a thing.


I tried playing significantly weaker opponents

It didn’t yield much difference to the current rating. Maybe it would be different if I played only ~9 stones weaker players. But maybe not. Maybe one accidental loss due to sandbagging or something would balance out tons of wins against weak players.

I mean, what are you waiting for, try it.


right, so you could play 10k players until you reached 1k and then 9k players until you reached 1d…

Heh yeah, in theory its possible xD

tho you would need to play and win a ton of games, since winning against someone 9 stones weaker in non-handicap game would increase your rating just a tiny bit, but losing a game will have lot larger negative impact on your rank


Even if you are very confident in winning, you have to ponder lose by timeout, disconnection and such.


I don’t know exactly how it is in OGS’ Glicko-2 (it has more gears and pulleys than traditional rating systems), but in rating systems that I know something about it is like this:

Over time, given the ratings of the opponents in your games, your rating will increase if you win more often that expected, and your rating will decrease if you lose more often than expected. And the expectation of the system changes as your rating changes.

The rating change from a single game result is proportional to how suprising that game result is to the rating system, given the handicap (if any) and the ratings of you and your opponent just before the game finished.
So an expected game result will not change your rating much, while a very surprising game result will change your rating more significantly.

In Elo rating systems this is the expectation of the rating system for even games, given the rating difference between the players:


Read “expected score” as win probability, where 1.0 = 100% and 0.0 = 0%
The difference between the expected score and the actual score is the amount of “surprise” for that result.


I actually think the risk of accidentally timing out, encountering a sandbagger, or even just making a huge blunder is underestimated by the ranking system.

Thus, I believe that the strategy of only playing much lower ranked players would backfire in terms of rank progression. I think it would be like for every hundred or so tiny steps forward (which might not even be enough to earn a single rank), there is the likely risk of an even larger step backward, leading to a net progression downwards over time.


Coincidentally, the European Go Federation and the French Go Federation recently changed the parameters of their ranking system, specifically for that reason, because it had been observed that the previous parameters were underestimating the probability of losing against a weaker player.

1 Like

Lol is it a loophole if it’s not allowed?

Kind of like saying “I found an interesting tax loophole the other day. Just don’t pay!”


Amazon found that loophole as well. It looks like it worked well for them.


What is allowed and what is detectable are two different things.
For example, since you mentioned taxes, it is illegal to call a plumber or a woodworker to perform repairs in your house and then pay them without them giving you a legal receipt for your money.
Yet, it is impossible (or totally ineffective and not worth the cost) to police or find such a minor transgression. Who would spend millions to make sure than noone is skippin the tax on a 50 euro repair?

Therefore performing work without issuing a receipt is illegal, but also a real loophole.

If you want a programming term for it, this is called “the ostrich algorithm”.

I’ve removed part of your post, please stop giving people advice on how to break the rules.

Self-play will get you banned, not get you to 1d.


I am not giving advice on how to break the rules. I just said that it is an actual loophole. Someone said it was not and I provided the proof. I have no problem with you deleting that part, however:

I will say that pretending that a problem does not exist, doesn’t make the problem disappear.
I am very busy at the moment, but if people insist on doing this, I will test this one of these days just for the fun of it. I will notify you all afterwards so you can ban the accounts, but fun is fun and proof is proof :slight_smile: