New Rating System

[crump7]

Having played 69 games since the reset and had 4 losses, and currently sitting near the top of a master league, I have realised the current league system gives a basic indicator of your performance but leaves no real comparison of how your skill ranks against other players, and leaves me personally a little disatisfied.

I believe we would be better off using a system such as chess elo to rate everyone instead of league points, there would be no need for leagues of 100, which are arbitary and a bit meaningless, there would just be the elo rating.

There should still be divisions, fluffy, master etc., with defined promotion thresholds for example an elo of 2500 or greater for Super Titan, so everyone would have something to aim for as their rating increased.

I for one would rather see my true rating for example I would rather appear at #54 in the Master division rather than #2 in a master league of unknown number of leagues of 100.

Of course if a system such as this was implemented we would not want to lose all our progress and start from scratch, so a preliminary elo could be granted based on current data held.

Any thoughts?

[Kamikaze28]

During the beta, I brought up a similar point and Alex (the Alex) made some very convincing points about why there are no viewable skill rankings:

1. Outwitters does not use ELO, the values used are anything but human-readable and knowing them doesn't get you anywhere.
2. The whole point of divisions is to avoid a monolithic leaderboard. Being ranked #35,811 of 156,425 just doesn't mean anything to anybody and the lower 95% just won't care about it anyway.
3. Show a skill rating to people and once it's high enough they are motivated not to play anymore, because they might lose and fall down the ranks.
4. Show them a score they can maximize instead and they'll go to the ends of the earth to do so.
   

To quote from the discussion:

onealexleft Wrote:A lot of the design decisions in the league play are geared to making players play *more* and worry less about outcomes.

[crump7]

(07-16-2012 11:54 PM)Kamikaze28 Wrote:  During the beta, I brought up a similar point and Alex (the Alex) made some very convincing points about why there are no viewable skill rankings:

To quote from the discussion:

onealexleft Wrote:A lot of the design decisions in the league play are geared to making players play *more* and worry less about outcomes.

I understand that the current league design philosophy doesn't include what I have suggested, I just wonder if other players are happy as it stands or would like something along the lines I suggest?

Some good points made so lets look at them:

Quote:1.Outwitters does not use ELO, the values used are anything but human-readable and knowing them doesn't get you anywhere.

Two possibilities here, the unreadable numbers could be converted to readable, or an elo type system could replace them.

Quote:2.The whole point of divisions is to avoid a monolithic leaderboard. Being ranked #35,811 of 156,425 just doesn't mean anything to anybody and the lower 95% just won't care about it anyway.

I must admit it is my preference to know where I stand, it may not be anyone elses, and to me being #5,811 of 15,142 in say the master division would be better than #30 in a random league of 100, or 71 in a different league in that same division. Basically your league position just depends on which league in that division you have been assigned.

Quote:3.Show a skill rating to people and once it's high enough they are motivated not to play anymore, because they might lose and fall down the ranks.

I would be very surprised if there was much of that, however you could easily overcome that by an inactivity penalty if no games are played, so they would descend the rankings.

Quote:4.Show them a score they can maximize instead and they'll go to the ends of the earth to do so.

I don't quite understand this comment, can you elaborate?

[Kamikaze28]

(07-17-2012 12:36 AM)crump7 Wrote:  

(07-16-2012 11:54 PM)Kamikaze28 Wrote:  During the beta, I brought up a similar point and Alex (the Alex) made some very convincing points about why there are no viewable skill rankings:

To quote from the discussion:

onealexleft Wrote:A lot of the design decisions in the league play are geared to making players play *more* and worry less about outcomes.

I understand that the current league design philosophy doesn't include what I have suggested, I just wonder if other players are happy as it stands or would like something along the lines I suggest?

Fair enough. I for one am okay with the current system. We're talking about a mobile game after all, not an eSport competition.

(07-17-2012 12:36 AM)crump7 Wrote:  Some good points made so lets look at them:

Quote:1.Outwitters does not use ELO, the values used are anything but human-readable and knowing them doesn't get you anywhere.

Two possibilities here, the unreadable numbers could be converted to readable, or an elo type system could replace them.

Either way, it's a high development effort for potentially little gain.

(07-17-2012 12:36 AM)crump7 Wrote:  

Quote:3.Show a skill rating to people and once it's high enough they are motivated not to play anymore, because they might lose and fall down the ranks.

I would be very surprised if there was much of that, however you could easily overcome that by an inactivity penalty if no games are played, so they would descend the rankings.

I find it hard to explain such a complicated system in a mobile game environment. People started playing Outwitters and didn't even know what Wit Spaces were for - even though the very-much-in-your-face Tutorial told them.

(07-17-2012 12:36 AM)crump7 Wrote:  

Quote:4.Show them a score they can maximize instead and they'll go to the ends of the earth to do so.

I don't quite understand this comment, can you elaborate?

When players read "rating" and grasp the concept of it going up and down depending on how well they play, they will feel compelled to get it as high as possible and then stay there by not playing for a while.
Points on the other hand have this "get as much as possible"-character to them and players will feel motivated to do just that. In other words, players will always feel motivated to keep playing to the best of their abilities.

[crump7]

Just to shorten the length of the post here's my summary of implementing an elo type system, with suggestions to deal with most of the cons in italics.

Pro's
Each player would know his skill rating compared to others.
It may give the game improved longevity for serious players, and even attract others.

Con's
Development costs against perceived gain.
Against design philosophy of worrying less about outcomes and may discourage new or more casual players. (Perhaps keep the league system as is for all but Super-Titans and possibly Masters).
Abuse of ranking by some by not playing games (An inactivity penalty will stop this).

[Harti]

No matter what quality your arguments have, crump, ELO is probably never going to happen to Outwitters. Alex and Adam decided that way back in 2010, even before they began to program anything more than prototypes. It's pretty much set in stone and has gone through a couple of iteration lifecycles, each time evolving to a better system. The whole system itself appears to be very (!) clever and precise right now (but that could be just me having played ~230 ranked games after launch), and it will keep getting EVEN better, the more information it can get.

I don't think they will ever ditch these years of work (to improve stability and to reduce point inflation) to replace it with the obviously flawed and exploitable ELO system that every other uncreative developer uses.

Your points are good but I'd like to remind you of what Kamikaze28 (or originally OneAlexLeft) said:
You're not meant to be able to compare yourself with the others.

Hell yes, I'd be interested if the system deems me the best player currently which is unlikely because me being #1 in Black Stick Division of Super-Titan League doesn't mean I'm better than #1s from other divisions or even #2 or #3 from my own. But I really don't feel a persistent urge to know that because I'll find out in a couple of weeks. If I'm still leading, chances are good.

Knowing your currently hidden rating would either mean that you're going to be disappointed or it would mean that you're going to check what changed everytime you win/lose a game and find an exploit to get yourself on top. Neither is really cool.

[crump7]

Fair comment, if OML have a better system than elo that they have invested a lot of time in then of course it would be madness to change, and if it can't be adapted easily to give a skill rating then so be it, there's not much point in further discussion.

[MarkH]

I get that they don't claim to have a better system than ELO but they think they have a system that is different in order to accomplish that the player a) actually plays more (to get leaque points), b) still has some kind of competetion within his leaque and c) can get promotion/demotion based on a non visible skill ranking.

I like the system quite a lot. The thing that needs work IMO is the unvisible skill ranking system/leaque placement as some promotions/demotions seem somewhat strange.

[Kamikaze28]

(07-17-2012 02:26 AM)MarkH Wrote:  I get that they don't claim to have a better system than ELO but they think they have a system that is different in order to accomplish that the player a) actually plays more (to get leaque points), b) still has some kind of competetion within his leaque and c) can get promotion/demotion based on a non visible skill ranking.

I like the system quite a lot. The thing that needs work IMO is the unvisible skill ranking system/leaque placement as some promotions/demotions seem somewhat strange.

1. (jokingly) by using the non-existent word 'leaque' you have immediately lost the argument

2. (seriously) People seem to expect these ranking systems to work perfectly from day 1, which is just preposterous. Firstly, there are programmatic mistakes which could not have been found, tested and fixed in a beta with at most 100 people (iOS restrictions) - which are now actively hunted by Alex, thus the unexpected league shifts. And secondly, these systems need to be trained for a while before they function as intended.
You start with, say, 10.000 players and you don't know the skill of any of them. Now they all play 5 placement matches randomly against each other and based on the results of these games you are supposed to place them in leagues. This initial placement is a rough guess at best and can only be refined with time and more data points (i. e. games).
Even the oh so highly revered ELO system would suffer the same fate if it started without any prior knowledge. Just because you could see your rating fluctuate wildly in the first days does not mean that the matchmaking is any better.

[ArtNJ]

(07-17-2012 03:05 AM)Kamikaze28 Wrote:  Even the oh so highly revered ELO system would suffer the same fate if it started without any prior knowledge. Just because you could see your rating fluctuate wildly in the first days does not mean that the matchmaking is any better.

That just isnt true. Twenty games is a ton under the ELO system. Nor can ELO be meaningfully manipulated. However, your point that ELO is never coming to Outwitters is apparently 100% correct, and wins the ELO argument, so the real issue is whether OML needs to admit that a lot more people are a lot unhappier and more puzzled than they anticipated, and provide more guidance on how the current system works and what you can expect.