I’m just kinda curious, how is the overall rank calculated? Why is it higher than all of my other ranks
overall goes up each time you win. all the other ranks only go up when you win that game type. Because this is true for everyone, everyone’s overall is higher than their game type which means beating them gives you more experience in overall than in game type, continuing the same pattern.
It seems daft.
What use is it?
Is there some reason why it makes sense for it to be that way, or we just haven’t thought of a better way?
I agree, it doesn’t make sense for a player who is 2k live, 2k correspondance, 1k blitz (and play all three very actively with good win rates on all of them) to be 2d in overall rank
As i understand it the reason is that the overall rank is calculated completely independently from the other ranks. Afaic there is no real reason behind having it. Its not used for any feature the site provides as handicap is calculated off time specific ranks. It was simply kept around when the other 3 ranks were introduced.
But ive already stated at other occasions that i dislike the multi-rank-system. Its not something that concerns me enough to deter me from playing though .
Your blitz, live and daily ratings are adjusted whenever you finish a game of the relevant type. They are adjusted according to the result and your opponent’s rating for that game-type.
Your overall rating is adjusted whenever you finish a game - regardless of what game-type it was. This rating is adjusted according to the result, just like the game-type ratings, but it is not adjusted according to your opponent’s rating for that game-type. Instead, your opponent’s overall rating is used.
This has two relevant consequences:
- Your overall rating will change faster than your game-type ratings because it gets adjusted for every game.
- There’s only a loose connection between your overall rating and game-type ratings. Their drawn from a different rating pool (the overall rating pool) so it should not be especially surprising that it is different to your game-type ratings.
Remember, your rating is just a measure of how you have performed relative to other players in the same rating pool:
- Harry has a blitz rank of 10k and Susan has a blitz rank of 8k. In a blitz game between Harry and Susan, we can predict that Susan will win.
- Jim has an overall rank of 5k. In a blitz game between Jim and Susan, however, we would have no way of predicting the result. We simply don’t have enough information - comparing Jim’s overall rank and Susan’s blitz rank makes no sense; they are from different pools.
- If we discovered Susan’s overall rank, we could compare her relative strength to Jim’s, but we wouldn’t be especially confident about predicting the result. Maybe, when playing blitz, Jim panics under time-pressure and Susan will defeat him easily? We have no way to tell.
- If we discovered Jim’s blitz rank, however, we could compare the players’ strength at the relative game-type and make a prediction with confidence.
Is that helpful? Or have i just confused things further?
I think (unlike kickaha ) that the multiple rating system makes a lot of sense. As Farraway said, just because you are good when you can spend a lot of time thinking doesn’t mean you will be good under time pressure. And just because you are good under time pressure doesn’t mean you will hold up against someone who shines when they have time to think. That justifies the rank for game type. It helps us choose opponents who are well matched in the situation we are about to play.
I’m not sure what the overall rank is useful for when it is basically a variable amount above all the others most of the time.
If it were more indicative of the others - like say average - then it would be more helpful for a general prediction like in the case of Farraway’s second bullet. If Jim’s overall rank was his average rank, and it was 5k, then at least there is some reasonable indication he might be better than Susan at blitz.
Not quite, overall rank is much more useful than any average - it allows you to see how you perform overall compared to everyone else on OGS. An average isn’t attached to a rating pool and would not afford this comparison:
- Ted only plays blitz games.
- Garry only plays correspondence games.
- Comparing their average ranks is entirely meaningless as they are from separate rating pools.
- Comparing their overall ranks is significantly less meaningless; they are from the same rating pool.
This relative performance stands across all game-types. For that reason, the overall rank is the most useful rank for comparing relative abilities.
Of course, if you want to compare relative abilities at a specific game-type, such as blitz, then it’s best to compare that specific game-type’s rank. The key is simply not to bother comparing your overall rank with game-type ranks; they are different measures. Any correlation between them is accidental.
I very much like this response
Yeah. I guess so. I just wish it scaled somehow so it doesn’t feel as ridiculous as Issho pointed out.
The thing is that the absolute value of the ranking actually means something ephemeral. I know I’m “a 20k”. Just because my overall jumped up to 19, that doesn’t make me 19. And as you approach SDK and Dan it can feel even more silly.
Well, it’s important to remember that a rank is only a valid measure of your performance relative to the rest of the pool.
So when you achieve a 1 Kyu blitz rank, you can’t actually say that you’re a 1 kyu player. Just that you are approximately two stones stronger than someone with a 3 kyu blitz rank. It won’t even translate across servers - people who play on KGS or Tygem are in different rating pools, so you can’t compare those ranks to your 1 kyu blitz rank either.
I think the disappointment stems from the idea that ranks should somehow be an absolute measure of performance. The reality, perhaps sadly, is that they are a relative measure of performance compared to others in the same pool.
I think that is only mathematically true, which misses the “human” element.
To test this, here’s an experiment: imagine how well received the “overall” would be if we simply decided to scale it up by 10 ranks. Your “overall pool number” = whatever it is now + 10 ranks. IE my overall would be 9k instead of 19k.
Any problem with that? It’s still just a pool, relative to only other numbers in the pool.
But if we did that, OGS would be laughed out of existence. Because people attach meaning to the absolute value of the number. 9k means damn good and 19k means just better than a beginner.
The only reason we “get away with” our overall being a rank or two higher than our other rankings is becase the effect is not so extreme. You can kind of accept that I get labelled with 19k when in fact in absolute terms I am 21-20k.
But I don’t think it sits easily. It’d be better if the maths produced a more coherent result.
Actually compared to other servers, OGS is apparently a little harsh. I’ve heard that we’re more or less 4 ranks lower here than on KGS so if you have a problem with feeling like a 20k when your overall says you’re 19k, if you played on KGS you might actually be more like 15/16k. Like Farraway said:[quote=“Farraway, post:11, topic:12310”]
they are a relative measure of performance compared to others in the same pool