Write leagues system from scratch

Sorry, you must be logged in to use the shoutbox!

[MonkeyIsland]

Hello.

Personally I think TUS leagues system needs a few tweaks. But I feel many people need a brand new system to shake things up and get active again.

There are 29 days left for the current classic/TEL leagues to finish and I'm trying to run the new leagues system straight after that. So, I'm putting every idea on the table.

I'll accept league plans till 23th February so that in the next remaining days I have time to code the ideas.
If you're suggesting a new plan make sure to think about what would you do about the current players stats.

The best idea or combination of the ideas will start from new season.
This is HHC's plan, please read it (long text, but won't bore you)

You may post in this thread for discussion.
(spams will be deleted)

[franz]

so no TRL Feb 1st?  (just curious)

[franz]

HHC's plan is pretty good, as it reminds me of World of Warcraft's arena rating system -> each new season has an overall rating that carries over while the season rating starts fresh.  the season rating increases/decreases a lot until it gets back or close to the overall rating.  while the overall rating increases/decreases normally because it doesn't pay attention the season rating.

the season standings will looks fine again because it will show who is doing the best that season to try to reach their high overall rating, or even those trying to surpass their medium overall rating.  it also solves that problem of newcomers getting high easily just because they join late.. now they have to work hard because their overall rating starts low.

[Rok]

I read HHC's plan carefuly. +1 for nicely readable form in which it was presented, however, I have issues with the content. I'll try to explain it as well as I can, so bear with me, it will be a long post and I'm not much of a writer. Let's start from the beggining and set some facts on which we can base the discussion on.

Points system/rating system - the difference

This will sound like telling the obvious, but despite the numerous times it was explained on these forums, I think there are still folks who don't really understand the difference.

In closed competitions with known number of players results are usualy evaluated with points. Let's take a well known example, football world cup group stages. There are 4 teams in round-robin system, every team plays all other teams. The number of games each team will play is known before the competition starts and is same for all teams. Therefor, assigning a constant 3 points for a win is both practical and fair. When the group stage is over, we get a clear picture of who performed well and deserved the playoffs spots.

Now let's look at our beloved worms. We have a time frame for the duration of the competition (season length) and this is where the similarities with football world cup end. Players can play as many matches as they wan't or can. They can play basically whoever they want. Could points system be fair and practical in worms? Hardly.
   First we would need to limit the number of games a player can play per season. I don't think that's what anyone wants - we're not making a living from worms, we play for fun, right? Second, we can't really assign opponents or set times for matches (we do this in cups to a certain extent, but that's another story). But we have an alternative, it's the infamous rating system.

What's the difference then?
The points a player gets for a win is not a fixed amount but rather proportional to the skill of your opponent.

How does the rating formula work?
A match is worth a certain amount of points, in TUS's case 80 points. The rating of opponents is used to calcute the expected score. Let's say you play against a player with the same rating as yours (so both of you are supposed to be equaly skilled) and that winning is a result of 1 and losing is a result of 0. Your expected score in this match is 0.5 (you're expected to win half the time). Then the formula goes like this:

match worth * ( result - expected score) = rating change

So if you win:    80 * ( 1 - 0.5 ) = 40
And if you lose:  80 * ( 0 - 0.5 ) = -40

Bang! This is where the magic number 40 comes from

Another example, this time when you are playing against a much higher rated player - this time your expected score might be 0.1 (you are only 10% likely to win):

You win:    80 * (1 - 0.1) = 72
You lose:    80 * (0 - 0.1 = -8

NOTE: This is basically how TUS rating system works. It is not something MI or any other wormer came up with, but is based on a system that was developed by professor Arpad Elo and is used in chess competions since 1970. Those interested can read more about it on wikipedia: http://en.wikipedia.org/wiki/Elo_rating_system.


In the long run, your rating should reflect how skilled you are compared to others. The more games you play, the more accurate your rating is. We can see how a rating based system is more appropriate for a competition like a worms league than a points based system. Even with players playing different number of games with different opponents we can still get somehow fair picture of who is who.

HHC's proposal

One thing that bugs me the most: you match up two players with low, but equal ratings. If they would play 1-1, they'd lose points! This goes against all logic in my opinion. In no way should players lose points when they perform on par with their skill! On the other hand, players with high ratings wouldn't seem to suffer from the same problem. Yes, that would encourage more games between better players, but I think it gives higher ratings an unfair advantage. It would indeed encourage lesser ratings to look for better opponents, but at what price? Finding a tus match isn't always easy and finding a specific opponent would be even more difficult, so that really doesn't solve the problem of avoiding.

HHC mentions that his system isn't a "zero-sum" system. Hm, well that's the trick with rating, it's supposed to be "zero-sum". In other words, if you perform on par with your expected score (see previous chapter) over say 100 games, your rating will be the same as it was before those 100 games. This is how it's supposed to work. It looks like that in HHC's system, players would increase their rating even if they performed worse than expected. This will distort the picture that ratings are supposed to show.

That's my main issue with this idea. I hope I was specific enough, as HHC asked, if not, I'd like to hear it.

The "TUS world ratings" and "TUS challenger" are nice ideas and could easily be a nice addition in the future.

My suggestions

The main complaint we've been hearing, I think, is that players with high ratings don't appear high in seasonal standings table and that they have trouble getting in playoffs. So I suggest the following, based from what was discussed in other threads in the past:

- get rid of seasonal standings table
- make the standings table based on overall rating (like in ladder system)
- have only players who have played in the season appear in standings table (so inactive players are not shown)
- have a separate table with ratings of all players, including the inactive ones
- instead of seasonal standings, have a column with seasonal rating to each player name, called "seasonal
   performance" or something like that
- give playoffs spots to top 6 from active standings and the top 2 "seasonal performances"  (could also be 5-3, or 12-4
   if we ever decide to have 16 players in playoffs. For clans, make it 3-1, perhaps)
- make a nice and obvious chart with playoffs requirements, so it's clear for everybody how they can qualify (the
   minimum of games played and winning percentage must still apply, of course)
- a "seasonal performance" table could still come handy though, so people could see how much they need to qualify
   for PO that way
- Set strict and fixed deadlines for playoffs!!! (f.e. two weeks for quarterfinals, two weeks for semis, etc.)
- make announcements for playoffs on main page, raise hell and fuss for the winners, make a big deal, etc

As you can see in my opinion the system itself doesn't need much change. It's not perfect, but it works fairly well and with minor tweaks we can still improve it. There's more idead in my mind, but it's 3:30 am here and I've been boring you enough, I think. Thanks for reading!

[franz]

One thing that bugs me the most: you match up two players with low, but equal ratings. If they would play 1-1, they'd lose points! This goes against all logic in my opinion. In no way should players lose points when they perform on par with their skill!

That's not against all logic.. it's just reality.  If their overall rating is below the season start rating of 1000, they deserve to lose points until they get close to their overall rating, and only then will this stop happening.  but that's ok, they can now start trying to improve their overall+season rating together as they should.

If you still feel this is too harsh on players with overall rating under 1000, then there's always the alternative of making everyone's season rating start at zero.  As long as you make it impossible to have negative season rating (below zero), then everyone gets the advantage of boosting upward to their overall rating.  it's a trade-off though: to make it less harsh on those with overall rating below 1000 and losing points down to their overall rating, you make the climb for highly overall rated players much higher.

both ways are fine in my opinion.. but the first one seems easier to do.

[Ray]

I will try to once again collect my thoughts, especially about how the profiles should look like, work. I'll try to make up my mind about that.

[Komito]

That was interesting to read HHC's Plan.

Nice to see my idea actually had a little use.

I like the Global Rank thing he proposed, but HHC, did you just throw a bunch of Random people into that table as an example of different ways to get points, or did you actually research it so it is accurate to that amount of players being in that position?

Oh, you put Random as Green for the BnG, while f4st actually has the most points

However, I see a problem with this, in my opinion it's concrete, TuS Classic being added together with TEL and TRL(when it comes), which would mean people who excel at Elite or Roper can basically get 2 sets of very high points for the same thing, which is unfair meaning it's easier for them to appear "Globally superior" to others just because their best schemes are in 2-3 different leagues.

[HHC]

One thing that bugs me the most: you match up two players with low, but equal ratings. If they would play 1-1, they'd lose points! This goes against all logic in my opinion. In no way should players lose points when they perform on par with their skill!

Depends on interpretation. This situation occurs when players have a low overall ranking (900 for example) and a greater season ranking (1050 for example).
In your opinion both players should gain/lose 0 points and stay at 1050, because this is the expected result.
In my opinion both players' real skill level is 900. When they play a draw against each other they 'confirm' that they are equally good, namely '900'-good and not the 1050 that happens to be their season rating. Thus, it doesn't seem illogical to me, that both players drop a little in season rating (to resemble more closely their real skill-level of 900).

As I wrote in my essay, this is how NNN works.

HHC mentions that his system isn't a "zero-sum" system. Hm, well that's the trick with rating, it's supposed to be "zero-sum". In other words, if you perform on par with your expected score (see previous chapter) over say 100 games, your rating will be the same as it was before those 100 games. This is how it's supposed to work. It looks like that in HHC's system, players would increase their rating even if they performed worse than expected.

Hmmm.. this is not true. When players perform as usual, their season rating will soon approach the level of their overall rating and stabilize at that level (and not increase or decrease from there).

The essence of zero-sum is that the sum of points gained by winner and loser is zero. This is not a necessity IMO.
The part that follows on 'in other words' applies to a continuous ELO system (which is merely an example of a zero-sum system) whereby Random starts with 1648 and ends with 1648. In my system he starts with 1000 at the start of the season, and, if he performs like usual.. his season rating will approach 1648 after 100 games. That seems fair? There's no way he will reach 1648 if he doesn't do as well as his overall rating suggests.


With regards to your suggestions Rok.. it is exactly the system that we use right now (but with a slightly tighter control).. it doesn't fix the things people complain about and IMO is not the way forward.

Franz: they deserve to lose points until they get close to their overall rating, and only then will this stop happening.  but that's ok, they can now start trying to improve their overall+season rating together as they should.

..now they have to work hard because their overall rating starts low.

Thanks for the thumbs up Franz. But there's one thing I like to point out. You seem to suggest 'newbies' have to work on their overall rating in order to get into playoffs. This is not the case.
Newbies and veterans start the season alike. The overall rating they have does not influence the amount of points they get for beating someone in this season. They both get an equal amount of points for beating someone with a rank of say 1500.




However, your response made me come across a potential 'bug'.

Let's suppose Random (1300) plays Noob (900) at the start of the season and the match ends 1-1. Normally this would leave both players 1000 in a normal system. But in my system Random will lose points and Noob will have points added to his score, giving the impression that Noob ownz Random when in reality they should be ranked equally for playoffs (as they are 1-1, equally good). (I guess this is what Rok is getting at??).

This is indeed a bit of a fluke. I don't think it will influence people's scores in the long run, but it does indeed sound weird, wrong even.

Komito:I like the Global Rank thing he proposed, but HHC, did you just throw a bunch of Random people into that table as an example of different ways to get points, or did you actually research it so it is accurate to that amount of players being in that position?

There's just one Random peep in there

I picked the top players in each league and added some random names (Mods, myself, CF members). It's just to give an impression.


About TEL and TRL: I chose to have both of them added to the score. Roper and Elite have their own leagues because they are the most popular and widely viewed as the most important schemes in Worms.
The Freeleague schemes only count for 0.5, I didn't think it was too much of a sin to count Elite and Rope for 2 (as being more important than skill in WxW, shopper, T17 or BnG...).
But it's just an example of how to count such a score. (I'd make Normal scheme count for 1 as well).

[Komito]

Cheers mate, I like your proposal for the main League btw, i'm willing to try anything either way

Anyway, now that I actually noticed your post lol.

About TEL and TRL: I chose to have both of them added to the score. Roper and Elite have their own leagues because they are the most popular and widely viewed as the most important schemes in Worms.
The Freeleague schemes only count for 0.5, I didn't think it was too much of a sin to count Elite and Rope for 2 (as being more important than skill in WxW, shopper, T17 or BnG...).
But it's just an example of how to count such a score. (I'd make Normal scheme count for 1 as well).

I really don't agree with your methods of calculation here, there cannot be any doubt that this is a biased view, even if the majority do enjoy Elite and Roper more than other schemes, every single scheme played should have the same contribution to such a thing called "Global Rank".

As for saying Elite and Roper being more important than skill than WxW, shopper, T17 or BnG, I really really REALLY disagree here lol.

To me, the 2 most skillful schemes are TTRR and BnG, mainly because of the lack of luck, and I am more impressed with schemes that fully concentrate on reactions and reflex/good timing, wouldn't matter to me if I was the best Elite player and Roper in the world, the fact people DO lose Elites/Ropers quite often due to bad luck with some turns, like weird bounces with animals and placement of mines etc in Elite, and crates in Ropers, is far too frustrating for me to be enjoy it even if I was the most perfect player ever for those schemes.

^^ That is just my opinion, it isn't a statement and I am not trying to put it down as a fact, and neither of us are right or wrong.

But I really don't think YOU can judge what schemes are more important than others, if you want to have a Global Rank, There shouldn't be 2 Ropers 2 Elites, and Free league schemes should all count as 1 each not 0.5.

The reason why I am against this, it's totally unfair for players who don't like to play Ropers and Elites as much as others who do, this gives Ropers and Elite players a VIP status and a clear advantage.

For example, Mablak could just own the hell out of Elite/TEL/Roper/TRL and the TTRR challenges and be dominant, and for example Chelsea could dominate 10-17 schemes in Free League and not even get close to Mablak, and if Chelsea was dominant in 16 different schemes, and Mablak only 3-4, I would obviously have to say Chelsea was a MUCH better all-round player.

In a way it's like saying "If you wanna be the best on the Global Rank, you gotta play Ropers and Elites, the things we think are more important".

In my opinion, If you master all the schemes in Free league, you are better than someone who only mastered 2-3 schemes in Classic, and using the same scheme twice for TRL/TEL.

I hope I didn't offend anyone, I just can't believe you are trying to justify this method for Global Rank.

The idea, without a doubt, is amazing mate, I back you up 100% on that, but the system on how it works, needs a total makeover, in my opinion.

Ok, i'm finished editing now.

[HHC]

Let's agree to disagree then.

To me, owning a scheme like 'Holy war' (which only 10 people or so play more than once a month) is much much easier than being the best in Elite or Roper. When Mab ownz the TTRR challenges, the Rope league ánd the Elite league. Doesn't that mean he's the best around? I could care less if he sucks at Moleshopper, to me, just being the best Roper and Default player overall is waaay enough to be considered the world's #1.


But, as far as I'm concerned, this is a topic for later on, if it does get added at all.

[Komito]

T17 and BnG has been around just as long as Roper/Elite, in fact BnG has been around longer than Ropers if you consider previous games in the francise.

TTRR/WxW and shopper have been around for just as long as the updated Elite scheme that people use now compared to the previous.

Hysteria is actually more popular in leagues than Roper, Elite is only more popular as a League scheme, not as a funner.

And Bungee Race is actually a very hard scheme to get VERY good at.

Battle Race is harder to win than Ropers/Elites, and takes more patience and time.

Big RR/Tower takes skill.

Forts takes skill.

SSR/WfW/Intermediate/Golf all take alot of skill to get really good at.

All these schemes take just as much skill and practise to get good at than Elite/Roper, the only difference is people don't find them as popular, a schemes popularity does not make it more or less important when comparing them with skill, and who knows, the Global Rank would possibly make some players focus more on the Free League and make it more active so they can have their claim to fame...

1 person who can win Elite/Roper and TTRR doesn't make him best all-round, it just makes him the best at those 3 schemes.

With a system like this would be like being back at school, having your different groups of people, your popular kids "Ropers/Eliters" your geeks "Free League players" your inbetweeners who are cool to the geeks but geeky to the popular kids but they have their moments, "Other TuS Classic schemes".

Sorry, but I ain't playing no scheme I don't like enough just to impress the popular kids and get on some biased Global Rank, i'll stay put in the BnGlee club lol.

Talk about splitting up the community, we should support all schemes and make them all look as important as others, not ridicule others because they don't enjoy the "popular" schemes and judge them as less skilled or important because they can't get double/triple points for certain schemes...


It's obvious where I stand on this method for Global Rank, but yeah, we definately agree to disagree lol.

[franz]

Thanks for the thumbs up Franz. But there's one thing I like to point out. You seem to suggest 'newbies' have to work on their overall rating in order to get into playoffs. This is not the case.
Newbies and veterans start the season alike. The overall rating they have does not influence the amount of points they get for beating someone in this season. They both get an equal amount of points for beating someone with a rank of say 1500.

true, a newbie's overall rating doesn't affect how many points they win.. so all they have to do is win every game  

my point is just that->it's about when they start losing games vs equally skilled newbies: that gap between season rating and low overall rating converges after enough losses (assuming season rating still starts at 1000 instead of zero). so no, I'm not trying to suggest that 'newbies' have to work on their overall rating in order to get into playoffs ->  I'm saying that their low overall rating is the only thing holding them back if they lose any games, especially vs equally skilled newbies. anyone can reach playoffs if they win all the time    If they do end up losing a few times, and converge their overall rating, well then they'll have to start working on their overall rating to get into playoffs.

However, your response made me come across a potential 'bug'.

Let's suppose Random (1300) plays Noob (900) at the start of the season and the match ends 1-1. Normally this would leave both players 1000 in a normal system. But in my system Random will lose points and Noob will have points added to his score, giving the impression that Noob ownz Random when in reality they should be ranked equally for playoffs (as they are 1-1, equally good). (I guess this is what Rok is getting at??).

This is indeed a bit of a fluke. I don't think it will influence people's scores in the long run, but it does indeed sound weird, wrong even.

The whole point of a rating system is to prove that 1-1 does not mean equally good.  There's nothing unusual about Noob (900) winning more points than Random (1300) in a 1-1.  It also doesn't mean that Noob ownz Random either... are you seriously arguing this?  Having a higher rating only means you have a higher% chance of winning vs someone.  You can't win every game.. so even if Random loses, it doesn't mean he got 'owned'.. it could have just been an unlucky game or a mistake-filled game.  I don't know why I have to even say this. I feel like you already know all this.

[ZiPpO]

"Thanks for the thumbs up Franz. But there's one thing I like to point out. You seem to suggest 'newbies' have to work on their overall rating in order to get into playoffs. This is not the case.
Newbies and veterans start the season alike. The overall rating they have does not influence the amount of points they get for beating someone in this season. They both get an equal amount of points for beating someone with a rank of say 1500."

best system rating.

overall points dont influence in season points.

The standings will be keep?

[HHC]

The whole point of a rating system is to prove that 1-1 does not mean equally good.  There's nothing unusual about Noob (900) winning more points than Random (1300) in a 1-1.  It also doesn't mean that Noob ownz Random either... are you seriously arguing this?  Having a higher rating only means you have a higher% chance of winning vs someone.  You can't win every game.. so even if Random loses, it doesn't mean he got 'owned'.. it could have just been an unlucky game or a mistake-filled game.  I don't know why I have to even say this. I feel like you already know all this.

I do. And it seems fair that Random gets fewer points (cause he won/lost against a noob) than Noob (who won/lost against an awesome player). But what I'm saying.. suppose Random and Noob don't play any games after this anymore, nor do any other players. Then the season ranking would look something like this:
1.Noob 1015
2.Everybody else 1000
3.Random 980

Based on the 2 games played: Random beating Noob and Noob beating Random, it doesn't seem fair that Noob would get the season win even though he has proven to be no better than Random (their games ended 1-1).

Of course everything changes when Random and Noob play more games, but it may be that players will find the above case weird when this happens on the first day of the season. They may be inclined to think the noobs have a benefit over veteran players, even though this is not the case.. Noob's 1-1 against an elite player is just valued more than Random's 1-1 against a noob.

So yeah, I panicked prematurely



@Zippo: the standings will show the seasonal ranking and p/o spots will be handed out according to this rating. Overall ranking isn't important anymore, but may still be shown to give players a better impression of what they are up against.

[ZiPpO]

lol...for me season ranking is more important than season ranking rofl