Well at least that information could be aggregated in the game info as IMHO that may be a bit too cluttering the game list.
besides the effectiveness could be something a bit too theoretical to be associated with the game information, I’m pretty sure some will argue on its estimation. I mean a komi and a handicap are facts, the effectiveness is an idea, not a fact.
I don’t talk about now, I talk about when handicap would be mentioned. to jgk
i think it would be honest to ask if komi afterall have to be in the list too so one last poll
Do we add one more column with komi?
Plus, rank is generally shown as kyu/dan level, but handicap stones and reverse komi is (as it should be) calculated directly from the Glicko-2 rating, which is more precise and thus can easily give apparent off-by-one errors
yes, but I wouldn’t add a separate column: I’d put both handicap stones (if any) and reverse komi (if any) in a single “handicap” column
Note: I use “reverse komi” in this sort of technical context to refer to the amount of komi that is added or subtracted from the ruleset’s base komi for handicap purposes. So for example, if you play a NZD game (7.0 komi for even games) with 0.0 komi, that’s -7.0 reverse komi. If you play a -14.0 reverse komi game (equivalent to giving one handicap stone with normal komi), you’d set the komi value on OGS to -7.0, because 7.0 (base komi) + -14.0 (reverse komi) = -7.0. Since the base komi is always known for a given ruleset, and all games must either leave the total komi set at default or a custom number, the missing element (reverse komi) can be trivially calculated if needed
I feel like it makes more sense to explicitly state the actual komi the game used, if listed at all, rather than stating it as a difference from even komi for that ruleset, both for immediate clarity and further because ruleset isn’t even listed on profile page.
Positive if white gets komi, negative if black gets reverse komi.
But then it’s not comparable to handicap stones. 1 handicap stone == 2*komi reverse komi. Without separating komi from reverse komi, you get 1/2 stone apparent errors
Also, I’d prefer to base it off perfect komi (6.5 most territory rulesets; 7.0 most area rulesets), so Chinese rules in an even game would show 0.5 reverse komi, but basing it on ruleset, at least on the front end, is the more conservative option
I don’t like the idea of blending the two, if I have understood you correctly. Links between stones and komi are subjective at best. The UI should be unambiguous. State how many stones, if any, black received, and if a good format can be agreed on, then separately state the actual komi used.
This feels like it would be far too confusing for most users, IMO.
I don’t think you did. They wouldn’t be blended, just in the same column
I believe it was @Maharani who showed that Katago thinks the 2*perfect-komi value for one handicap stone holds up remarkably well to fairly high handicap levels. So I’d say there’s a lot of objective evidence backing up the rationale intuition
Yes (with the caveat of stating reverse komi, not total komi), just in the same column
Then they should play rulesets which merge fair, perfect, and base komi, such as NZD. Not my fault if rulesets are written to misleadingly call a 0.5 reverse komi handicap game an “even game”
Yeah but you can’t just reject reality and substitute your own. Rulesets are what they are, usually based on centuries of national tradition. Whether you like them or even whether they still make sense post alphago, katago, et al., they are the rules being enforced and used by most players, and so we need to base our UI off of this current reality, and not off some ideal world.
NZD has a lot going for it, not least of all 7.0 komi, but most players use AGA, Japanese, or Chinese ruleset, so it would be strange to base our whole UI around only NZD. It will upset too many players.
I agree. You can’t reject the reality that perfect komi is 6.5 for territory (unless you do something to adjust this, like a “biased button” that is worth 1.0 points if picked up by White, but 0.0 points if picked up by Black, is thus worth effectively 0.5 points miai, and thus justifies reducing komi to 6.0) and 7.0 for area, the AI estimate is too confident on that for any other proposal to be reasonable. I’m not the one rejecting reality, and the creators of the NZD ruleset agree with me
And that has led to some deficiencies in them, most famously the traditional territory ones, hence why we need to encourage people to use only good rulesets like Lasker-Maas, NZD, and Tromp-Taylor
Well, it would be interesting to have OGS Japanese rules modified to be a correct implementation of Japanese rules complete with both players lose, AI to execute hypothetical play, No Result, &c., but if you’re not advocating for that, then neither of us are talking about correctly implementing territory rulesets at the very least, which are the most popular on OGS (citation needed)
Note that I initially proposed calculating reverse komi from the ruleset’s base komi. I only brought up perfect komi as my preference
I’m not advocating for basing this off of NZD, I’m saying it would be better to base it off perfect komi. Just because the creators of NZD recognized this as well, does not make NZD the grounding for that decision. We’re both appealing to the same objective observations about the degree of Black advantage in Go
I fail to see how this makes it wrong
Regrettably, we’re beginning to plumb depths beyond my available energy levels to follow, so I cannot reply with integrity to all the points you raised.
If I may, I will restrict my answer to just your last point.
This is a server for people to enjoy, I think anoek takes that rather seriously. While there are a very vocal minority of pedants here on the forums, who like things to be as ideal as possible, I think ultimately anoek cares most about what the player base as a whole will find enjoyable and usable.
At the very least, if we are going to be setting standard komi values that differ from those set by governing bodies, maybe we should create an entirely seperate “OGS Ruleset” that specifically has the rules anoek, or else the OGS community, like best (which is already a dumpster fire fight between the live players and the corr players).
I’d like that. Or perhaps just the expected winrate given the rating gap, handicap stones and komi, which would just be an alternative way to display basically the same thing (an indication of the odds).
One immediate difficulty is that (as far as I know) the server only reports ratings at the end of the game. Especially for correspondence, these can be quite different from the beginning, so the current system would not be able to give the correct “expected” win rate.
KGS implements it like this, and OGS could use “Km” as a header to save space on the profile page if needed, though “Komi” isn’t many more characters either.
(maybe “Hd” or “Handi” for the handicap column header on OGS would work, too)
This is about finished games right (history page)? It is the correct rate at the time the game finished. I think the rating system uses those ratings in its calculation of the rating updates. Why would you consider it incorrect?
Or would you advocate to use the ratings at the start of the game for rating updates?
I think it shows the rating after the update for that game. Before would be a lot better, (almost perfect for live games).
No, but for setting up a game, handicap is decided based on the ratings before the game. So those are the relevant ones for explaining the level of handicap. With the post-game ratings, the calculations will seem off.
I think it would make most sense to show the expected winrate based on the player ratings right before the game result was processed. Also when displaying the “effective handicap” or “effective rank”, it should use the player ratings right before the game result was processed.
Reflecting on all the good replies here about “effective rank” vs displaying stones and komi. Perhaps stones and komi are clearer since the stones benefit black and komi benefits white all the time.
Where with effective rank, may not be as clear since the player can play both black or white. I.e., effective rank could be seen as boosting black’s rank or weakening white’s ranks depending on which color the player took for that game.
So while it may be clearer in terms of fewer UI elements or mental calculations, it may be less clear in terms of perspective of who took the handicap?