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 differenceThis 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 proposalOne 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 suggestionsThe 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!
Forums
Author
Topic: Write leagues system from scratch (Read 17719 times)