What can be done to avoid and punish sandbagging?

We all know that in online playing there is always ### people who founds fun to sandbag or cheat.

I played a game against one a couple of days ago, and following him since then and I reported 5 times. He get a warning and several of his games were deleted, and rank “restored”, but it started to do the same to lower it.

He play blitz games. Hundreds of them every day. When he ranks up, he starts losing on time on 9x9 or 13x13 to keep his visible average rank low.

Things get worse when you think about that, and understand doesn’t matters he is banned, will create a new account and do the same, harming our experience in the site. For this reason punishing is not the solution after all. This thread is to find possible mitigations to avoid this kind of situations. We know is almost impossible to avoid, but we can try to make it difficult for them.

Site has cheat detection but maybe sandbagging detection is needed too. In the same way correspondence games are annulled when several games loose on time to protect the ranks, maybe on live games something similar can be achieved. If the system detects several games lost in time under a number of movements treeshold, locks ranked games with humans and let him play against bots (more later). And this system ideally must be automatic to avoid driving crazy to mods.

Ban and new account. This is the most difficult part. New accounts should be easy to be created, but at the same time a sandbagger or cheater don’t cares about the account or ban and just creates a new one because he just find fun to punish. IP ban does nothing, can harm a legitimate player and we live in the dinamic ip era, then … Maybe a possible solution is to “tag” this kind of players, and face them against fake accounts ran by bots.

EDITED:

In reality this has nothing to do with sandbaggers as is, is more related to an specific situation of an specific sandbagger. Anyway, I think shouldn’t be the way and if you want to discuss it use this thread provided by GreenAsJade

Average rank. Matchmaking works based on last 15 games or so AFAIR, but when you face an opponent on a game, you should see his rank for this board. IMHO should show the rank under 19x19, or 9x9 and so on.

If system “knows” this player is 7K strong, makes no sense to show it as 11K. This are 4 stones difference.

Would like to know your opinions about this and try to make OGS even a better place.

There are a couple of topics to respond to here.

First topic.

Sandbagging does somewhat drive mods mad, but it is perhaps the least troublesome of all the problems we try to manage.

Probably the most “maddening” aspect of it is that people tend to report something as “sandbagging” when they mean “they can’t believe this lower ranked person beat them”. They have no evidence of rank manipulation other than this :pleading_face: These waste tons of time.

I guess if we fully automated sandbagging detection, then we’d remove manual reporting, and this would solve that problem :squinting_face_with_tongue:

However, aside from that, to your points:

  • We already have means other than IP to resist returning users who’ve been suspended
  • We have automated cheat-detection that will be deployed against sandbagging (and others) once we finish tuning it for AI cheat detection
  • In the shorter-term sandbagging will soon be handled by Community Moderation, which will improve out ability to respond to sandbagging reports in a timely fashion (this has been a problem).

Second topic

this topic doesn’t seem relevant to avoiding and punishing sandbaggers. It’s a choice about the rank basis for matchmaking.

This comes up from time to time - the most recent example is here .

If you wanted to tackle that topic, you’d do well to review the past discussions first.

2 Likes

Well, you are right about that

1 Like

Don’t get me wrong, but for me it’s the most annoying thing about OGS. It made me delete my OGS-account a while ago and coming back it does not feel good that it’s still that noticeable.

my 2025:


(no handicap games)

Most users are not sandbaggers: I have more losses than wins vs stronger and more wins than losses vs weaker.

I just try to win with equal effort against users of ANY rank. When game already started I see no reason to worry if opponent is 25k or AlphaGo. I can look at opponent profile after I properly finished game myself.

3 Likes

No one said anything oposing that.

That’s nice. I wish it would work for me. I know that’s partially a me problem and my fear of feeling stupid. For me it just does not feel nice to to play a game I don’t stand a chance of winning. The fact that we have a ranking system for low level player makes me think I’m not completely alone with my problems.

I think there’s a difference between playing a normal game with someone who is significantly stronger than you and playing someone who is deliberately playing lower ranked players just for the thrill, and I would say sandbagging refers to the latter.

The important point in my opinion is the deception. I play against players I have no chance of beating, but I want to do it willingly.

I play games against ddk. The result is done before the first stone is played. But I don’t rank my account down to get these games. I post a challenge with my regular account and if someone is willing to play under these conditions, he/she can accept.

1 Like

1 Like

I understand your point of view and congratulations about how you enjoy the game, but this is not the point. When you play under “ranked” rules, is because you want to face against an equal player. Is not about winning or losing. I win against stronger and weaker opps than me. It’s ok to loose, nothing bad, I try to learn from every game, and BTW, when you really learn is when you loose, BUT, if I face a stronger opponent I want to know it is stronger, this is what the Rank is for.

What I don’t want, is to face an opponent “like me”, and being trampled. I play to have fun and improve, to find a balanced game to learn from it, and being trampled is not what I understand as being fun. If you are 8 stones stronger than me and wanna play, I don’t have any problem, but I want to know you are 8 stones stronger.

When you face an equal opponent, you can understand why you loosed the game, but when you face a much stronger opponent than you, you learn nothing without a review.

That’s exactly what I’ve been trying to point out when I complained about the way general average of rank is used to match up with potential opponents, instead of respective ranks in the diferent time settings / disciplines. But that’s not the issue in this topic. The issue here is whether or not you can trust somebody to be the opponent you reasonable can expect him / her / them to be.

the whole ‘thumbs up’ thing started a huge discussion but the general idea could be implemented in a useful tool, just as it is used on digital marketplaces and, in a way, on this forum. Certain markers can be created to monitor certain aspects of the behavior of players to be reflected in a score or badge or something else which reflects the trustworthiness. If a user is polite, plays out accepted games in full, hardly times out, doesn’t escape a series of games etc. that could be reflected in a 1-2-3-4-5 star rating for example. Or whatever we can agree on. That would perhaps create a barrier for potential sandbaggers or bullies.

Just a thought that came up. I don’t know. Personally I hardly ‘meet’ sandbaggers but I always take a look at my potential opponent’s game history to get an impression of what I’m up against. In the case of playing live-games that’s almost impossible, due to the risk of the challenge disappearing while checking.

It’s not something very new. Since years moderation is considering this and taking more evolved policies around all the various wrong behaviors.
You will find many topics, suggestions, debates and such by using the search in this forum, you can reopen topics if you think you may contribute with something new useful for all of us.

I think the pie chart might compare the rank of your opponent at the time the match happened to your current rank, rather than your rank at the time the game happened. So the chart might not always be reliable.

See here Issue with pie chart on user page

1 Like

https://online-go.com/player/1782909
that user is 8k now
image
first 50 ranked games were not handicap and vs ddk players only, all of them still ddk


chart shows most games are vs higher ranked


so its not true

1 Like

I appreciate that it’s very annoying.

I will say that I didn’t say it is not annoying :slight_smile: Perhaps I should have say “it does drive mods AND USERS” mad.

Nonetheless,

… this is good news. If the worst thing about OGS is not people escaping, stalling, AI cheating, harrasing in chat or with images … wow, then we are doing a good job!

Another piece of good news is that the numbers tell us what you said: we are actually nearly there!

When I said

… the reason that sandbagging is there is because it’s the last “report category” that we can see from the numbers is “not under control”.

So:

  • It’s a problem, and it’s annoying
  • We’re working on it, and have a roadmap to containment

BUT

Sandbagging has a special element to it:

many many sandbagging reports are people feeling that they were deceived simply because they lost.

This unfortunately won’t change. We can prevent people losing on purpose to a good extent. We can prevent them returning with new accounts to some extent.

But there will always be an element of distrust that we are deceived when someone with a lower rank defeats us.

Perhaps if we can show that we have a working system of solid detection, then this distrust also will lesson

Summary: we hear you, hang tight, we’re working on it, and numbers tell us it’s already improving.

3 Likes

I think one problem is that ogs adjusts ranks very very quickly. A couple of losses is enough to lose a rank, while on most other servers ranks are way more stable. I’ve seen a lot of wild rank fluctuations on users who play a lot of games, on a bad day they might drop 5 ranks and then gain 5 ranks the following day.

Some might even go as far to say that ogs’s rating system needs more stability, at least for users who have thousands of games under their belt.

1 Like

Yeah this is a good concept for this thread “what an be done to avoid (the perception of) sandbagging”

In theory one of the glicko parameter is “how sure we are about a person’s rank”, I thought that as this increases (more sure) the person’s rank changes more slowly?

Maybe we need some investigation into that…

Good point, it might not be the issue I thought it was, but there is definitely something wrong. I just manually looked through their first 50 ranked games in the table and got this:

9 wins vs. stronger opponents
24 losses vs. stronger opponents
8 wins vs. weaker opponents
9 losses vs. weaker opponents