Is 30k > or < 25k? 
Depends on the coder i guess… personally, i would equate 25k as -25 on the backend, and as such -30 would resolve as less than -25
…that or, as here, use something like glicko where 600 is obviously < 1000 lol
What if whenever someone registers for the site they have to play 3 9x9 games against the cpu (auto adjusts difficulty after each game) or solve like 10 tsumego, would that give a better estimate of someone’s ranking?
Nobody would want to join our site. The idea is to make it as easy as possible for people new to Go to learn about the game and get connected with another human who hopefully gives them a good experience. Accurate ranking comes later.
2 Likes
I made this suggestion as both someone new to this site and the game of Go itself. I’ve been playing for 3 weeks I’m no match for the players I’ve had to play because I can’t play anyone that’d be closer to my actual level because I outrank them according to the site. I would have preferred to be challenged by a cpu initially to get a better indication of my rank than to lose to actual people so handily; it would’ve been less disheartening.
Playing a cpu would also provide a way to incorporate an automatic chat welcome message for a quick rundown of the rules, how to navigate the site, and suggestions for new players. You’d be able to provide a link to whichever group(s) would be most helpful to the lowest ranking players and “true” beginners.
5 Likes
Playing a cpu would also provide the chance to give an option to view a quick tutorial over object of the game, how to capture pieces, and examples of ideal shape formations. The more user friendly the site and initial assistance upon sign-up the more people would want to log on and continue playing. The only problem I can see would be players considering this an unfair advantage against them for whatever reason, but if the point of Go is to play against as many opponents as possible to better yourself in return why not try to make the best conditions to provide the greatest opponents you can?
2 Likes
I have made a feature request to optionally exclude [?] from accepting open game invites. This idea would solve the problem of higher level players being matched with absolute beginners.
Granted, it does not solve the problem of beginners with inflated rank; but half a solution is better than no solution I think.
Credit to Smog whose idea I have stolen 
That is the opposite of welcoming new players.
6 Likes
I think there’s a valid case for not being forced to play in the “new player welcome lottery”.
Someone who doesn’t feel like playing a welcome game is just going to cancel out anyhow, which is also the opposite of welcoming new players.
But ironically, all these “other” suggestions are moot (and pointless) if humble rank goes ahead.
1 Like
I’m sure we’ll hear a lot of new and refreshing complaints from 25k players being randomly mopped up by newcomers. However from all the solution which do not involve custom initial placements, humble rank is definitely the most pragmatic.
Btw. that should be actually quite accurately measurable. I completely forgot that the best way to engineer is to measure once and then cut twice, or vice versa…
- Take all the newcomers, from last 2 years and check their ranks after 5 or 10 games, or both
- Find the median, average. Plot the distribution. Analyse. Do the science. Find truth between the facts.
- Discuss and make the choice
- Implement and fine tune
So far we’ve done number 3. 
If this happened at my work I would get fired.
1 Like
Did you read my suggestion as well as Alex’, flovo’s and anoek’s posts about that and just forget about them or is there a joke I’m not getting?
None of the posts I remember where suggesting actually measuring the rank distribution amongst newcomers (please forgive me if I have missed something, I do indeed skip over some of your posts @smurph ). But since you are so sure that there are many examples, can you link at least one where such experiment was considered ?
I could have sworn that just the other day @S_Alexander posted pretty much exactly this graph, though darned if I can find it now!
1 Like
It’s goddamn hard to keep track of all the info. But probably we want to see this one by flovo.
Doesn’t look like image can show up in a mention like this so I copy it:
3 Likes
That’s the one I was remembering You’re right - lots of good info flowing!
1 Like
That’s a nice one! Thank you. @flovo would you be able to make separate distribution graphs for the ‘after 15 games’ category? That would be greatly informative especially if you could include percentile breakdown.
Meanwhile, it looks like the inflow of newcomers is centred around ~15k mark. That does confirm, 21k players wouldn’t be very happy playing the newcomers. But then if we apply a uniform correction for standard deviation across server (~2k iirc) that would give the initial ‘slightly less humble rank’ value of 19k, which is close enough to the average.
Still an average 19k player will have not more than 20% chance of winning games with new-comers (more due to distribution rather than skill), but it wouldn’t be as hopeless compared as same players distribution thrown at an average 21k.
I think that hopeless is fine for maybe one game anyhow. A string of hopeless, with hopelessly better players is the real problem.
Not only will it be “not as hopeless” from sheer “gap at the first game” but also less hopeless games to get correct rank, and likely more willing nearer to beginner players at first encounter (this is speculation of course).
GaJ
1 Like
I don’t want to go too much into psychological disputes here, as we all know psychology is not exactly a strict science. But being a 21k struggling to improve I wouldn’t be too happy if almost every ‘?’ game would mean an unpleasant beating, often by a player that’s still not very familiar with the server.
Right. The original problem this thread talks about is that every game for a newcomer is an unpleasant beating.
Now we have whittled it down (assuming Humble Rank) to “a significant number of games a beginner plays against [?] might be an unpleasant beating” so that is a big improvement.
How many games that a struggling 21k plays will in fact be against [?] in this new system?
IE is this even a problem?
I’m fairly new to Go and OGS, as a whole, but - and maybe this has already been brought up - would it be possible to have multiple bots at varying levels of play [maybe 25k up to 1d or higher, if needed. Also, with an ability for human interjection] for a provisionally ranked player, but - that until a rank can be determined - are only allowed to play? I can understand how this could be a hindrance to inviting new players,as well. Just looking for pros and cons. If this has already been discussed - again - feel free to move or remove my comment, as necessary.
EDIT: I agree with the title of the topic… I got my GF into go. The fact we couldn’t play each other, even though I believe she is potentially a stronger player than myself, has had a negative impact on her play. I honestly feel she is stronger IRL, but she let’s the "He’s a v. I’m a "
2 Likes