Forced Handicap in Ranked Play

I kind of get it and I kind of don’t. I just feel that “ranked” play should have the same conditions every time instead of 10 or so customizable settings that can effect the game on some level.

But again, just my opinion.


A solution to this is to create separate accounts for live and correspondence games.

1 Like

One of the people here also created this thread, which you might want to check out:


Nah just separate experiences. Multiple accounts is a terrible user experience

I’m pretty sure that whatever kind of game you play, it’s the whole average rank used and the specific are just statistics or information.
From many times this was the answer provided here on the forum.


Please don’get me wrong, I’m a very content user of OGS. I don’t consider my occasional issue a serious problem. I just wanted to mention it. I’ll try to improve my ‘live game’ rank asap to get rid of it :wink:

1 Like

I don’t think I agree with this in relation to ranked games. The system can essentially cope with working rankings for games which have a whole range of variables and I don’t see why these shouldn’t be available if people want them. However, I could agree that the automatch or default settings should be simplified so that they are always say Chinese rules handicap or something like that.

even more off topic thought

I can envisage a painful introduction when people first start and have a handicap with free placement that they know nothing about, play a stone and sit there patiently waiting for their opponent to play while opponent gets cross at beginner taking so long to place handicap stones.
There are swings and roundabouts…

1 Like

I hear you on this.

I handle this by playing on fox with the same setting and rely on fox rank to evaluate my progress. Then I am free to play whatever I like on KGS and OGS.

Those who play only 19x19 Live have more stable rank than those who play any board size and any time.
If stable rank user will use ±2 rank restriction in opponent search, he will sometimes get too easy game(vs -2) or too hard game (vs +2)
if unstable rank user will use ±1 rank restriction, they will be in same situation as stable rank user with ±2 rank restriction
so, its not that bad problem that ranks are slightly inaccurate because they calculated on different settings. Its same as if we just do more wide rank restriction.

But, if handicap will be everywhere by default, then those ±1 problems will be intensified. Two actually even strength opponents would play with handicap stones only because rank of one of them is slightly too high and rank of the other is slightly too low.

I think ranking system should serve the players and not other way around. Restricting players to please rankings is not the right way.


Not forced, but by default handicap on. I’ve banged on about this a few times in the past.


Don’t forget to factor in non-skill based game results: if you have a 1d playing a 10k even with the 1d expected to win 99.9% but each player has a 10% chance of lagging out or getting an error submitting move bug or an important phone call and losing then the further from 50% expected winrate the worse junk data you are inserting into the ranking system when the 10k wins by timeout.


The rating system is supposed to handle both even and handicap games gracefully. Even games (between players of very different strength) can be handled by default with a Glicko2 system like OGS uses.

Handicap games are treated with some extra considerations.

OGS adjusts the ratings of the players according to the handicap when calculating the new ratings after a game.

Example with totally made up numbers, to show how it works in principle:
Black has a rating of 1000 and gets 2 handicap stones against white with a rating of 1300. In an even game black’s rating would be increased by 50 points after winning, because 1000 beating 1300 happens rarely. But with the help of the handicap stones black is treated like having a rating of 1250 and a black win would increase black’s rating only by 12 points, because 1250 beating 1300 is not that surprising.

Source: Search · get_handicap_adjustment · GitHub

So there should be no need to force even games or handicap games.


These are very different options.

I could 100% support swapping the default value for the toggle.
I could never support removing the option for ranked games w/o handicap

Edit: actually, for automatch the default is already set to handicap enabled :+1:

I guess it worked :+1: well done


No, I’m pretty sure there’s just one rating that is used, like this:

(That page has a lot of my guesswork but I think it has been proven right in many examples.)

1 Like

I may have missed it, but did you clarify why you think this? What do you mean by it?

The way rating works, the correct rating adjustments are applied to players taking into account handicap, at least in theory. So what “results” does it mess with?


When players of higher rank difference play ranked without handicap.

I will say I only feel this strongly about handicap because there are just not enough players.

1 Like

… when players of higher rank difference play ranked without handicap … what effect do you feel is broken?

Generally, when this happens, neither player’s rank is affected much at all, because the much stronger player wins, and gets no rating update for beating someone of much lower rank, and the person of lower rank takes no rating hit because the opponent was so much stronger…

… as a result it’s hard to see how this messes up anything?


Absent deliberate ratings abuse, the main way I know of that it might mess up results is the way that Uberdude mentioned - the small chance of getting a massive upset that has nothing to do with skill, but rather is due to someone’s internet breaking, or some real-life emergency, or even just the cat jumping on the desk and causing a misclick whose undo is refused.

Generally Elo-style ratings models only model skill-based probabilities of game outcomes, and so assume that the probability of an upset approaches arbitrarily close to 0 as the ratings difference grows, but due to things like the above, the probability of upset doesn’t actually approach 0 in real life. Once the skill difference is so extreme that any possible upsets are predominantly non-skill-based and don’t otherwise provide genuine information about skill, such games are going to be on average harmful to the ratings system. Of course, because upsets for such reasons are still very rare, and people don’t actually tend to challenge players when the difference is too horribly lopsided (again, aside from the much bigger and real problems of things like of ratings abuse / sandbagging) this means in practice the harm on average should still be very small.


Mmmm … but one non-skill related upset isn’t “messed up results” is a very low probability event though, right?

So for an active player, it’s a blip? It would be a very aggravating blip for the recipient, but from a systems point of view, it doesn’t “mess up the results”.

It just means someone has to play a few more games to fix up their rating :wink: