Ad hoc rank adjustments

From SAGoClubs
Jump to navigation Jump to search

This is a discussion page to find a consensus on an appropriate ad hoc rank adjustment system. If no consensus can be found, we can formalise competing proposals and subject them to a vote at the next council meeting.

If you want to comment, please do. Mark your comments with two dashes and four tildes, so that we can see who said what, and indent your comments with colons at the beginning of the line.

Status Quo




Thus far, the maintainer of the ranking system has pretty much performed ad hoc rank adjustments on an ad hoc basis. Maintaining this system would be the simplest, and until evidence of abuse were to surface, or until the maintainer were to complain that it is too onerous a task, it could be retained.

Steve Kroon: My view (as "maintainer" of the ranking system) at this stage is that if SAGA requests me to make ad hoc rank changes, I am willing to do it. The major issue here is how SAGA takes decisions on whether ad hoc rank adjustments should be made. Historically, this has been done on the recommendation (viz. "perpetual grumbling") of the local clubs.

Konrad Scheffler: This is not quite correct: across-the-board adjustments have been done once or twice but there is no system in place for adjustments, especially for individual players. The reason for the proposal is that the lack of a SAGA decision making system means, by default, that decisions don't get made.

Andre Connell: Currently it's not clear at all how to do adjustments, there are no rules in place, and maybe that's the best way to do them - have one person making the decision for or against. The problem though is how do people get ranks adjusted up here if that person is in the Cape, or vice versa. Perhaps there should be one person per region in charge of ratings - the person adding in games on the system maybe?

Formal system proposed by Chris Welsh




There are three people involved:

  • The nominator - the person who proposed that an individual is given an ad hoc rank adjustment
  • The acceptor - the individual concerned who must agree with the proposal
  • The approver - a SAGA appointed individual with the authority to approve ad hoc rank adjustments

The nominator can be any SAGA member who feels that someone's rank is completely inappropriate. He/she would formally propose to the approver that the acceptor should be given a promotion (or demotion) to some specific level. Formal in this sense means a written request, either a letter or an email. Once the acceptor has also formally accepted the proposal, the approver would be required to play an unrated game against the acceptor in order to assess the validity of the claim. Once the game is complete, the approver can accept or reject the claim, or, alternatively, suggest some other suitable level (subject to the caveats outlined below). Once this is done, the approver will formally instruct a person with suitable rights on the ranking system to make the ad hoc adjustment.

In order to prevent abuse, certain safeguards need to be put in place. Firstly, as indicated above, the approver would need to be formally appointed by SAGA. The intention would be to have one such individual in each of the major regions, which currently would mean one person in Gauteng, and one in the Western Cape. It would make sense for this person to be the strongest or nearly strongest player in that region. These individuals would retain this role indefinitely, until they declined the role or SAGA felt they were no longer suitable.

Secondly, as the intention is that this occurrence should be the exception, there should be some minimum adjustment required before it can be considered. I am proposing that it should be at least 5 stones. This means that if a 17k player is proposed to be promoted to 11k, and after the game between the acceptor and the approver, the approver feels a rank of 13k is more suitable, this would not meet the 5 stone minimum, and nothing would thus be done.

There are some concerns that need to be considered. Firstly, the actual necessity of such a system. It has been argued, with some validity, that the system can cope with such inappropriately ranked players already. They merely need to play a number of games until they reach their correct level, which should be relatively swift if they are in fact as strong as has been suggested.

Let us consider a fictional example. A player, say at 17k, emigrates to a foreign country, where they vigorously pursue their go development. 18 months later they decide to return to South Africa, after reaching a level equivalent on the South African system of 6k. In order to reach their correct level, they would need to either WIN approximately 30-40 games at the "correct" handicap in order to reach their correct level, or alternatively, play games at "incorrect" ranks to advance more swiftly. Given the known problems with playing games at the incorrect rank (i.e. the extremely punitive index points deduction from the "stronger" player) the player will find it difficult to persuade opponents to play at incorrect handicaps. All in all, this would seem to be far too much trouble for everyone concerned, and you would probably find the 6k player washing their hands of the whole business and sticking to playing their go online.

I would thus argue that in rare, extreme cases, there is a strong case to be made for permitting an ad hoc adjustments.

Secondly, the issue of abuse. You would not want a system that allowed players to inappropriately recommend each other for promotions. I feel that by making the process formal (i.e. "on the record"), by insisting on the approver being appointed by SAGA, and the 5 stone minimum, you remove the possibilities for abuse (or undue tinkering).

Thirdly, the issue of players in other regions, such as Pretoria and Durban. This can be resolved either by appointing a suitable approver in the relevant region, or requiring the approval game be played online. This is of course potentially quite inconvenient, particularly for the approver, but as has been hinted at in this draft, the occurrences of these ad hoc adjustments should be extremely rare.

I'd be grateful if you could gather comment on this proposal and then submit it to SAGA for consideration

Comments by Steve Kroon

Just to reiterate my point that I feel such a system is unneccessary, since I feel (a) that such cases as outlined are rare,

This is true by definition in the proposal --Konrad Scheffler, by mail.

and (b) when they do happen, a player who wants to fix his rank is likely to enter tournaments, where he will be able to find opponents, and his rank will adjust rapidly at these tournaments.

This may well be true. These points argue that the suggested change is non-urgent, but I don't think they actually argue _against_ the suggestion. --Konrad Scheffler, by mail.
This issue is urgent. Bob is back from Korea, he may have improved a lot. Dale Morris has jumped to at least 4-kyu in his abscene from recorded club play and tournaments, he is officially ranked at 8-kyu. I fail to see why we can't deal with these kind of improvements without forcing other players to pointlessly play (sorry for split infinitive) over many weeks other players a wrong handicap for the sake of moving numbers. Btw, not all go players in need of a rank change play tournments for a variety of legitimate reasons.--Tristen

(c) I also feel that it would be more worthwhile to design a system more capable of absorbing these rapid changes than trying to design fixes for them.

I think that's easier said than done. We (Steve and I) are agreed that a complete overhaul of the system is low priority, so will have to wait until either we have more person-power available or someone does it on their own steam. In the meantime I don't see any reason not to patch the system where it can be done without significant work. --Konrad Scheffler, by mail.
I agree with Konrad--Tristen

Comments by Konrad Scheffler

Thanks for that. I like the general idea. One can question whether this is necessary and/or sufficient but I don't see how it could be a bad thing, and I don't think implementing it would require any actual work as far as the rating system is concerned.

One relatively minor issue: I don't think most players can judge the strength of someone more than 9 stones weaker than them - I for example have no idea what the difference is between 15 and 20 kyu (there are certainly exceptions to this - people like Charles Matthews are knowledgeable about the various learning stages of go, i.e. know the strengths at which various concepts are typically mastered). So I would suggest that an appropriate approver should be selected for each case, who should not usually be more than 9 stones stronger than the suggested new rank. In this case the whole process (appointing of approvers) should be overseen by a ratings officer appointed by SAGA (by default probably the membership secretary).

A detail to be debated is the minimum adjustment. Personally I'd go for 3-4 rather than 5 stones, but that's just me...

Comments by Andre Connell

I have been pro ad-hoc adjustments for a while now, but I'd be very much in favour of a system with as little work to it as possible. I propose something along the lines of: Any SAGA member can propose, to the rank administrator, an ad-hoc rank adjustment for someone less than 9 stones weaker than they are in real strength. This adjustment must have been confirmed by one other SAGA member i.e. 2 members must have played against the recipient of the adjustment, and must be in agreement on the rating adjustment needed. Once this request is sent through to the rank administrator that person can then makes a final decision in favour of or against.

Let's keep this as simple as possible.

Example: Joe Soap is 5k and a SAGA member, he thinks Fred Soap, who is 20k according to SAGA, is probably actually around 10k. He asks Paul Soap, who is 8k, to play a 2-handicap game or whatever against Fred to judge Fred's strength. After the game Paul agrees that Fred is much stronger than 20k, but probably around 13k. Joe and Paul agree to submit a rank adjustment request to the administrator that Fred's rank should be adjusted to 12k or whatever they agree to. The administrator makes a decision for or in rare cases against.

Comments by Chris Welsh

Andre's proposal does not seem in substance very different from my original proposal. I think some combination of the ideas so far might work. For instance, Joe suggests to the SAGA appointed Ratings Officer that Fred is much stronger. The ratings officer requests someone to play against Fred, based on Fred's level. (In practice, Fred might propose Paul to the Ratings Officer, and might even get Paul to play Fred in advance of sending the request). I'm firmly of the opinion that the process needs to go thru a more formal process than simply sending some requests to ratings administrators. I'm not suggesting that Rank Administrators are going to mess with the system, but I think something as potentially open to abuse and complaints needs a more rigorous protocol.

Third Way

Please describe one if you can think of one. Maybe require ad hoc rank adjustments to be made by local club committees? Maybe require them to be posted on the mailing list two weeks before taking effect?

Tristen: Gotta love the third way. Players at clubs know best if an individuals rank is way off (for example, John Leurner returns from Indochina and years of go lessons, and is really a 2-dan, JHB/PTA guys will pick that up in one evening of play). Why not allow the clubs to make recommendations for rank changes?

The suggested proceedure:

1) Club players identify a rank is off and alert club administration 2) Club administration plus the club's two strongest players agree on the new rank (how a club does this is up to it) 3) Suggest rank is sent to person in charge of ranking engine 4) Ranking engine person double checks for fraud (if the Potch Go Clubs has a sudden influx of dan players, for example). If no fraud is found, rank is changed. If fraud is suspect, query club administration. 5) SAGA Council reviews all rank changes on a quarterly basis for fraud