I was wondering if the reason why OGS ranks in the 1k/1d category are close to the EGF/AGA, but overrated in lower ranks, could be due to the over-reliance of lower ranked players on the analysis features.
Maybe disabling analysis in ranked games would make the lower ranks in OGS more inline with the EGF/AGA ranks.
I personally would support this, as I create custom challenges with no analysis. But sometimes I’m not too fussed about the exact time settings, and just want to use the automatch features. Unfortunately, I always get matched with people using the analysis. I don’t use it for live games, but I suspect most of my opponents don’t do the same.
Do you mean before the recent rating update, or after where it was chosen to line up the OGS 1k/1d with EGF and AGA specifically?
Honestly I think the OGS ranks in comparison to EGF are still a bit of a mess in rated tournaments.
On the one hand, some people just don’t play here, (similarly for KGS in the corona cup) and that makes the comparisons silly. On the other, some people just don’t have an EGF rating and sign up for one with OGS ratings, or some people have gotten stronger in the last year and without tournament play their OGS rank might go up x kyu (and then get catapulted around by the ratings update) while their EGF one lags behind.
I think it’s too early to be conclusive on whether OGS overrates players compared to EGF.
Anecdotally, I was around 15k before the update, and I jumped to 9k/10k.
When I played in person against EGF rated players, they would put me around 15k (based on the results of my handicapped games).
EDIT:
I agree, the dust still has to settle.
But intuitively I would think that something like an analysis tool would make a much larger difference in “boosting” a lower ranked player, than a dan player. Yet another variable to worry about when adjusting OGS’s ranking system. I am not familiar with other servers, but if they do not have analysis enabled by default, then they don’t have to worry about this variable.
What about before or after the recent EGF rank changes? They also bumped their ranks up (more recently than us) and obviously individual experiences will vary but the expected values were forecast to be quite similar once both updates went through.
This was what people would rank me as around the end of 2019 / beginning of 2020 (before the clubs closed).
And before people ask: no, unfortunately I don’t think I got 5 stones stronger between then and the new OGS ranks.
In a handicap game you still appeal to the players sole memory and ability to read, that is not the case with analysis.
I get your point but it is not the way I like to play. I don’t see why analysis should be part of what gives you rank on a go server. It is also not used as a handicap, as both sides can use it in their games.
For me it means that I don’t accept a number of games. It is an inconvenience if I auto-match.
I don’t think it gives someone rank
your neural net is your true rank, number of playouts is just how hard you try in current game.
With analysis you can explore moves that was afraid to explore without it, many people will be too lazy to play additional unranked games for it.
There are bad sides of analysis, but also there are a lot of good sides, so you can’t be sure that it would be better for everyone without it.
I listed a lot of good sides in my posts above.
so it seems its better to make auto-match without analysis
But anyone should be able to create custom ranked game with it.
As there are players which use analysis in their ranked games it contributes to the rank as shown on OGS.
I don’t know if that produces somewhat uncertain ranks - for ex. a player who reaches 5k by relying on analysis, but loses at that rank as soon as it is turned off?
I haven’t read all the posts above. I don’t doubt that analysis can be a valuable tool for new players to explore the game.
The question of whether analysis makes a player (or most players) stronger or weaker is separate from the question of whether analysis should be allowed in ranked games. It’s my opinion that analysis makes a player stronger on average, if they practice using it. But even if it made no difference on average, I wouldn’t personally allow analysis in ranked games because it seems like such a large departure from non-analysis Go that the two things are different games.
I would be happy if at least automatch defaulted to non-analysis.
I want to straighten what I wrote in one of my posts above: I don’t consider the players using analysis on OGS as cheaters, as this feature is available to both players.
if someone will stop use analysis after always using it, I’m sure he will not drop more than 1 rank, so still high winrate vs 5k, but rank drops sometimes for many other reasons, so quality of rank system will not be significantly lower
I don’t wish to copy paste same things again, I explained more in-depth somewhere above
