2013 - AAG 40 League

[Jennifer]

If you lose points if you play and lose to inferiorly rated players, then who would take a challenge from someone who lost a few games and won nothing?  Only concern.

That’s what I mean by, I am not sure how we would go about implementing a weighted system.  Unlike unofficial tabulations, I have to deal with the players complaining the system is not fair.  :cry:

Don’t worry about it, no one will get shafted this year.  The only rule changes I’m likely to implement are ones that keep players playing, like upping the number of games you can have with an opponent.  We can discuss new rules in October for the next season.

[gamerman01]

Right
Well, why not consider Darth’s 2013 AA50 system?
Too bad it’s not getting tested, so we don’t really know how well it works, but it seems OK.

FWIW I have no problem with simple win% so long as the minimum # of games is sufficiently high (like a lot more than 4 in this year’s G40)

The only issue I see here is somebody going 4-0 or 5-0 or 8-0 and then sitting on that.  Talk about incentive to not play games…  :-P

I trust Jenn when she says no one will get shafted this year

[boldfresh]

@Gamerman01:

Right
Well, why not consider Darth’s 2013 AA50 system?
Too bad it’s not getting tested, so we don’t really know how well it works, but it seems OK.

FWIW I have no problem with simple win% so long as the minimum # of games is sufficiently high (like a lot more than 4 in this year’s G40)

The only issue I see here is somebody going 4-0 or 5-0 or 8-0 and then sitting on that.  Talk about incentive to not play games…  :-P

I trust Jenn when she says no one will get shafted this year

i will no doubt be shafted, per usual.

[gamerman01]

@Boldfresh:

i will no doubt be shafted, per usual.

Well don’t flirt with the border line!

[wheatbeer]

On a related note, with Gamerman’s subjective system,
Tier 1 has gone 39-0 against tier 3.
Tier 1 has gone 39-8 against tier 2.
Tier 2 has gone 48-5 against tier 3 (and the five tier 3 victories include some forfeits).

That pretty impressive predictive power.

[wheatbeer]

• That’s

[Jennifer]

@Gamerman01:

Right
Well, why not consider Darth’s 2013 AA50 system?
Too bad it’s not getting tested, so we don’t really know how well it works, but it seems OK.

FWIW I have no problem with simple win% so long as the minimum # of games is sufficiently high (like a lot more than 4 in this year’s G40)

The only issue I see here is somebody going 4-0 or 5-0 or 8-0 and then sitting on that.  Talk about incentive to not play games…  :-P

I trust Jenn when she says no one will get shafted this year

Sounds like an issue where we need to raise the limit on the number of games.

Admittedly, we had 35 games last YEAR.  4 games out of 35 total left us with only two people qualified to play at all in the last game of the year, and that was just barely if I remember right.

We have many more games now.  245 by last count on Gamerman’s spreadsheet alone and we’re only halfway done.  (I only tallied wins, and I may have miscounted, I’ve been in the hospital a LOT this year, like most of January, February and March.)

So if 4 games means you participated in about 11% of games, to get that same percentage now (assuming the league ended today) you’d need 27 games to qualify.

Assuming we get as many games again as we have now, perhaps a total of 500 games, would a 50 game minimum to be in the championships be too high?  Probably, what about 25 game minimum?  We have 6 players with over 14 games now (14 or more to 17) just counting WINS, not counting losses.

I’m good with changing it if everyone else is.

[gamerman01]

At this point here in May, it seems to me 20 game minimum might be appropriate, ±5

[Major Major]

my vote is 30 games played.  I also think that currently active games as well as finished games should count towards being able to go best of 5…

[allweneedislove]

the limit should not be too high as it might discourage new players from joining the league.
i would suggest 10 games.

i also really like infrastructure’s idea of adding percentage points for games played. this gives incentive to keep playing more games.

having an attainable low threshold while creating incentive to play more games even with a perfect record makes for the most robust league.

i would suggest adding more than a single percent per game played. i would think somewhere between 1 to 6 per game played.

[nerquen]

@Gamerman01:

At this point here in May, it seems to me 20 game minimum might be appropriate, ±5

I think 20 games is just too high. If you assume players get as many games in May - October as they did till now, then the 20 games minimum would disqualifies 1/2 of the Tier1!!!

For me even 10 games minimum is hard to get now. Joining mid season and with a rate of ~ a game per month I was hoping to get about 6 games by October.

As for the next year I would vote for ELO ranking system (http://en.wikipedia.org/wiki/Elo_rating_system). It works just great in an online ladder I participate here
https://docs.google.com/spreadsheet/ccc?key=0AtLN1dNc0SyedExSbzMtZjg1WDdLTi0zNy1HMy1wb2c#gid=0
The rules are explained on the right part of the sheet. The trick is that the point gained/lost in a game are tabulated to integers to keep the calculations simple.

ELO is a well tested system and gives more accurate estimate of players’ relative strength than % of games won. It is still easy to implement. It also does not require any artificial threshold on the minimum number of games played as one cannot get artificially high ELO ranking with just couple games played. It is true though that ELO is known to converge slowly and might require 30+ games to have accurate estimate of one’s strength but I am not aware of any other ranking system with faster convergence that is still simple to implement. And in practice even after couple games the ranking is not that inaccurate.

@ Gamerman: Don’t you want to convert your excel sheet to a google doc (as the one linked above)? It would just be more friendly and faster for players to check the update. And I believe it will also be easier for you as you don’t have to upload a file after every update. At last it does not require any installed software to view your rankings.

[Major Major]

@nerquen:

@Gamerman01:

At this point here in May, it seems to me 20 game minimum might be appropriate, ±5

I think 20 games is just too high. If you assume players get as many games in May - October as they did till now, then the 20 games minimum would disqualifies 1/2 of the Tier1!!!

For me even 10 games minimum is hard to get now. Joining mid season and with a rate of ~ a game per month I was hoping to get about 6 games by October.

As for the next year I would vote for ELO ranking system (http://en.wikipedia.org/wiki/Elo_rating_system). It works just great in an online ladder I participate here
https://docs.google.com/spreadsheet/ccc?key=0AtLN1dNc0SyedExSbzMtZjg1WDdLTi0zNy1HMy1wb2c#gid=0
The rules are explained on the right part of the sheet. The trick is that the point gained/lost in a game are tabulated to integers to keep the calculations simple.

ELO is a well tested system and gives more accurate estimate of players’ relative strength than % of games won. It is still easy to implement. It also does not require any artificial threshold on the minimum number of games played as one cannot get artificially high ELO ranking with just couple games played. It is true though that ELO is known to converge slowly and might require 30+ games to have accurate estimate of one’s strength but I am not aware of any other ranking system with faster convergence that is still simple to implement. And in practice even after couple games the ranking is not that inaccurate.

@ Gamerman: Don’t you want to convert your excel sheet to a google doc (as the one linked above)? It would just be more friendly and faster for players to check the update. And I believe it will also be easier for you as you don’t have to upload a file after every update. At last it does not require any installed software to view your rankings.

The flaws with ELO in this game is that again it doesn’t reward playing more games to qualify for a playoffs.  There needs to be a modifier to encourage play.  ELO also doesn’t work as well in a game that is impacted so much by luck.  If I was a top rated player I would not risk getting diced against the bottom player and then moving below the median.  ELO was designed for a large number of games with little/ no luck.  It is a great idea though, but a modification needs to be made. 
A simple mod (although it sounds complicated…) where you gain a percentage of your opponents points if you win but they lose a smaller %.

I.e; Lets assume a 20% win 5% loss scenario.
Infrastructure has 100 points Stalingradski has 500 points.
Infrastructure somehow wins the match.
Infrastructure has 200 points, Stalingradski has 475 points.
Stalingradski easily wins the match.
Infrastructure now has 95 points, Stalingradski has 520 points

Put this into an automatically adjusting spreadsheet and viola.

[Major Major]

I.e; Lets assume a 20% win 5% loss scenario.
Infrastructure has 100 points Stalingradski has 500 points.
If Infrastructure somehow wins the match.
Infrastructure has 200 points, Stalingradski has 475 points.
Instead Stalingradski easily wins the match.
Infrastructure would have 95 points, Stalingradski now 520 points

[boldfresh]

@Infrastructure:

@nerquen:

@Gamerman01:

At this point here in May, it seems to me 20 game minimum might be appropriate, ±5

I think 20 games is just too high. If you assume players get as many games in May - October as they did till now, then the 20 games minimum would disqualifies 1/2 of the Tier1!!!

For me even 10 games minimum is hard to get now. Joining mid season and with a rate of ~ a game per month I was hoping to get about 6 games by October.

As for the next year I would vote for ELO ranking system (http://en.wikipedia.org/wiki/Elo_rating_system). It works just great in an online ladder I participate here
https://docs.google.com/spreadsheet/ccc?key=0AtLN1dNc0SyedExSbzMtZjg1WDdLTi0zNy1HMy1wb2c#gid=0
The rules are explained on the right part of the sheet. The trick is that the point gained/lost in a game are tabulated to integers to keep the calculations simple.

ELO is a well tested system and gives more accurate estimate of players’ relative strength than % of games won. It is still easy to implement. It also does not require any artificial threshold on the minimum number of games played as one cannot get artificially high ELO ranking with just couple games played. It is true though that ELO is known to converge slowly and might require 30+ games to have accurate estimate of one’s strength but I am not aware of any other ranking system with faster convergence that is still simple to implement. And in practice even after couple games the ranking is not that inaccurate.

@ Gamerman: Don’t you want to convert your excel sheet to a google doc (as the one linked above)? It would just be more friendly and faster for players to check the update. And I believe it will also be easier for you as you don’t have to upload a file after every update. At last it does not require any installed software to view your rankings.

The flaws with ELO in this game is that again it doesn’t reward playing more games to qualify for a playoffs.  There needs to be a modifier to encourage play.  ELO also doesn’t work as well in a game that is impacted so much by luck.  If I was a top rated player I would not risk getting diced against the bottom player and then moving below the median.  ELO was designed for a large number of games with little/ no luck.  It is a great idea though, but a modification needs to be made. 
A simple mod (although it sounds complicated…) where you gain a percentage of your opponents points if you win but they lose a smaller %.

I.e; Lets assume a 20% win 5% loss scenario.
Infrastructure has 100 points Stalingradski has 500 points.
Infrastructure somehow wins the match.
Infrastructure has 200 points, Stalingradski has 475 points.
Stalingradski easily wins the match.
Infrastructure now has 95 points, Stalingradski has 520 points

Put this into an automatically adjusting spreadsheet and viola.

The ELO system sounds very interesting and sophisticated.  Is there an easy way to implement it?  Ie an automated spreadsheet into which you simply enter wins and losses and the rankings are calculated?

[Major Major]

it could be done with gamermans spreadsheet with several adjustments…

[allweneedislove]

i like the elo rating system. it is great at showing rank against competition.

however, showing rating and determining who qualifies for playoffs can be two different things.
if you want to generate a league with lots of games against diverse opponents elo is not the best system.

if the league wanted to get rid of annual seasons and just have a fluid ratings system to show comparative strength against competition, then elo is near perfect.

the triplea ladder used/uses elo and did a great job of allowing anyone to join the ladder at anytime and have a chance to be highly rated against veterans of many games. the big drawback is it discourages higher ranked players to play against lower ranked players.

assuming we keep the annual league/playoff system and have a goal of trying to create the most robust league then i think using gamerman’s subjective ranking system to determine playoffs works very well if you add a high minimum of games played to qualify. i know that it is not perfectly accurate but it is a very good approximate that shows lots of information and gamerman is willing to donate his time to create and maintain it! gamerman’s rating system also does a good job at promoting many games with the arbitrary tier system, a elo system does not.

if players will not tolerate a subjective system then i recommend using win percentage + bonus points per game played to determine who qualifies for the playoffs. the big problem with this method is it encourages higher skilled opponents to play noncompetitive games against lesser skilled opponents and avoid tougher matches.

if players can stomach a subjective system to determine who qualifies for playoffs, i recommend just using gamerman’s rankings to determine who gets in. this would be my preference.

[Major Major]

why not have two different playoffs?  if anyone doesn’t want to be in the one using gamermans ranking they can opt out.

[Jennifer]

We could have a “Playoffs” and “Junior Playoffs” using a number of games criteria to qualify to be in one or the other.

What could be done (I am not saying WILL, I said COULD) is the following:

• Players have their total number of games tallied.
  ** The top 20% of players based on number of games COULD qualify for the playoffs
  ** The next 40% of players based on number of games COULD qualify for the junior playoffs
  (60% of participants could qualify to play in the playoffs at some level.)

• Players who qualified for playoffs would have a point system of some kind to be determined.
  ** Each player tallies the number of points earned/lost against the opponents they played.
  ** The top 8 of the playoffs threshold would then be selected to play.
  ** The top 8 of the junior playoffs threshold would then be selected to play.

[gamerman01]

@nerquen:

I think 20 games is just too high. If you assume players get as many games in May - October as they did till now, then the 20 games minimum would disqualifies 1/2 of the Tier1!!!

I would point out that I said 20 ±5, so as low as 15, and I also used the shaky word “might” as in “might be appropriate”.  Whatever Jennifer decides goes.  She is very conscientious to making sure no deserving player is left out - don’t worry.

@ Gamerman: Don’t you want to convert your excel sheet to a google doc (as the one linked above)? It would just be more friendly and faster for players to check the update. And I believe it will also be easier for you as you don’t have to upload a file after every update. At last it does not require any installed software to view your rankings.

For everyone else’s benefit - I sent Nerquen a lengthy PM as I am interested in his ideas.
I’m writing now because I do have a concern about switching to a google doc.  My spreadsheet is very big because it shows the results between each set of players.  In Excel, (which is an extremely common almost universal product - who doesn’t have it?) you can zoom in and out so each player can see it how they want.  Also, it includes a lot of information in cell comments that would not be translatable to an internet page…

I just realized the biggest advantage of Excel…
Everyone is free to manipulate and edit the spreadsheet in any way they desire.  I don’t know for sure the capabilities and limitations of a web page display format, and I see the advantages you point out and am interested, but I’m starting to realize in the comparison that Excel spreadsheets have significant advantages also.

Again, thanks for the suggestions and I am still interested as I said in the PM in instructions for converting the spreadsheet to a google doc and displaying on the internet.

:-o Perhaps I will do both - best of both worlds?!

[nerquen]

@Cmdr:

We could have a “Playoffs” and “Junior Playoffs” using a number of games criteria to qualify to be in one or the other.

What could be done (I am not saying WILL, I said COULD) is the following:

• Players have their total number of games tallied.
  ** The top 20% of players based on number of games COULD qualify for the playoffs
  ** The next 40% of players based on number of games COULD qualify for the junior playoffs
  (60% of participants could qualify to play in the playoffs at some level.)

• Players who qualified for playoffs would have a point system of some kind to be determined.
  ** Each player tallies the number of points earned/lost against the opponents they played.
  ** The top 8 of the playoffs threshold would then be selected to play.
  ** The top 8 of the junior playoffs threshold would then be selected to play.

Hmmm… I am not sure I like it much. What is your motivation to separate players to three groups (frequent - top 20%, average - 40%, occasional - bottom 40%) based on the number of games played?

In the suggested system, a very strong player who is just busy with life (or joins a bit late in the season) and thus gets less games than 60% of the ladder would be out of the playoffs automatically.

I believe we want the strongest players compete against each other in the playoffs, does not matter if one plays 50 games a year or 10 games a year.