Proposal: One Rank

OGS Development
1

This post is about ranking in OGS.

I fully embrace the ELO based system. It works and has only minor problems.

What I think is more of a problem is that there are 4 different ranks for each player. This causes a lot of trouble when talking about the user interface and user experience, but also functional problems.

I know this might have been discussed all over already, but still I would like to elaborate on that and suggest alternatives.

Alternative

So let’s skip the criticism for now, because I think the problems with the speed based rankings are actually well known. Let’s look at how we can replace the speed-based ranks, while still providing meaningful information about the player’s different skills on different game configurations.

So what I suggest is: keep only the overall ranking, which is determined (as-is) by elo rating, and is a pretty accurate representation of a player’s strength. Replace the other speed-based ranks by more meaningful information. Say, looking at a 4k player (and he is 4k, not 4k-8k-5k-23k), show some additional statistics based on his last some 100 games played:

profile of sulai[13k]

  • played 90% 9x9 games recently
  • plays blitz and normal speed games mostly
  • rarely plays correspondence

So if you see a 4k player which plays a lot 9x9 and blitz, you know he is 4k in that game configuration and a little weaker in the other categories. In my opinion much better than speed based ranks, which are often not accurate and cause a lot of confusion and also functional problems.

What’s the problem with speed based ranks?

I think it’s safe to say, a good player on normal speed settings can pretty quickly gain the same level on blitz settings. So he might be the same or at most a few ranks weaker when short on time. In other words, the information provided by speed-based ranks is quite low.

Another problem is the accuracy of the rankings. As soon as players improves in the game, the speed based rankings become inaccurate. Why is this? Let me explain. Typically the ranks of the majority of players looks similar to this:

overall - 4k (most “realistic” value)
blitz - 7k (lagging behind)
normal - 6k (lagging behind)
correspondence - 23k (rarely played correspondence)

This player advances in playing skill, and plays blitz and normal games mostly. The overall ranking is much better, because each game is ranked twice: once in the overall category, and once in the respective speed category. That’s why the overall category is more accurate, while the speed categories need more time to adapt to the actual player’s strength. Conclusion: the blitz and normal games ranking in this example do not reflect the player’s actual strength in that game configuration, but they are inaccurate, let alone the correspondence rating.

The time category based rankings of many players vary a lot, and I think not because of real difference in strength on the different time settings, but because of inaccuracy caused by less games being played in one category than in the other. Even when having the same amount of games in each category, the overall ranking will still be much more accurate. For players playing only one speed category, this is even worse, because the other speed ranks just don’t change at all.

Because of the inaccurate rankings, we experience UI-related problems. Which of the 4 ranks is shown where? When in a game, the Ranking shown on the clock widgets is different than in the chat and the player’s details reveals that the real ranking is much different than his displayed rank. Confusion and lot’s of clicking to get a proper idea about a players strength.

Also, functional problems arise. For example, start an open correspondence game as 13k. A 21k joins and wins. Looking into his profile: overall ranking is 8k, correspondence 21k. Meh. Happens so often, especially in ladder games, where many players improved in overall rating, but still have a low correspondence rating. I know mods can help there, but I feel like trolling reporting so many games.

Another functional problem: which ranking to use for game matching? If we use speed-based ranking, we are using an inaccurate value. Facing a 19k, you might actually be playing a 6k. Besides playing an uneven game, it might ruin your own ranking, should you happen to be a 10k.

Also there is the aspect of simplicity. One rank. Roughly represents the player’s strength. That’s all.

Facit

I’m not totally determined that the speed based rankings need to be replaced by something more meaningful, but I see them as a weak spot on OGS, causing various problems while providing little value.

I think these problems are all well-known, and I sincerely hope my post starts an open discussion about this topic :slight_smile:

15 Likes
2

Rank is not accurate thing anyway. So better to have only 1 rank used in games.

But its not bad idea to display other ranks Only in profile page. There may be information about color rank, board size rank and speed based. Just for information.

6 Likes
4

Second this.

The four different ranks idea is novel, but I find it to be too clunky. It’s nice having stats on your individual performances in each of the different time settings, but four ranks is too much. Your rank is your rank whatever it is. One rank affected by how well you do in various games.

5

Idea for discussion:

Incorporate the speed ranks into matchmaking based on rank confidence. If a speed rank has low confidence, use more of the overall rank when matching. If a speed rank has higher confidence, use more of that rank. Since this can be done behind the scenes, it won’t be as confusing from a UI perspective. It will benefit players who may differ in play strengths per speed type.

1 Like
6

I support this proposal, like every time it comes up.
Don’t hold your breath.
Sorry for cynicism.

7

In the Chess world it is entirely normal to have different ranks for different time controls. For example, I am a ~1900 ELO Chess player. This is reflected in my standard and daily Chess ratings. However, my blitz Chess is much weaker; I oscillate between 1200 and 1500 ELO.

If I had only one Chess rating I would never play blitz chess. In the space of one hour I’d obliterate weeks of hard work on my other ratings.

The innovation that OGS has supplied is to provide an additional overall rank. I find this useful and I wish the Chess servers that I play on did the same. It means that I get the benefits of a single-rank system whilst also preserving my progress in standard and daily games.

I suggest anyone wanting a single rank mechanism merely ignores all speed based ranks. If you only look at the overall rank then bang, you have one rank - problem solved.

7 Likes
8

This topic was automatically closed 91 days after the last reply. New replies are no longer allowed.

9

This thread is outdated since Jul 2017. At this point the OGS rating system changed to glicko2. The rating system changed in many aspects. All info in this thread about Elo on OGS is obsolete.

1 Like