PBW League Idea...

[tesco samoa]

a win would be victory... which is decided between the players in the game... or just a standard Last man standing

[geoschmo]

I don't think we need to compare game settings to determine the value of a win. A win is a win. Everybody in that particular game is playing by the same rules, so beating them all is no more or less impressive than someone winning a game played with different settings.

Geoschmo

[Slaughtermeyer]

I used to play rated games of Age of Empires on the Microsoft Gaming Zone, which uses a scoring system similar to that described by LazarusLong42. It seems to work reasonably well both for head-to-head and multiplayer games, including team games. For details you may go to: http://zone.msn.com/conquerors/Ratingsfaq.asp

[Gozra]

Well a multiplayer win is diffrent than in a head to head game. When you don't have to use diplomacy who ever is better at expansion and production wins. In a multiplayer game that ablity is muted. I am in a game now where the lead player outproduces everyone but they are losing because everyone has ganged up on him. So I still think it takes diffrent skills to win at diffrent types of games. Not all games are created equal.
I would go for a published ranking system. Based on wins and participation. 3 point for a 'Win' 2 for second place and 1 point just for being in the game. and -5 for dropping out. All these points would be divided by the amount of games played. And Ranked players could advertise games and you could join based on your ranking. New folks could still play with experienced folks and still walk away with something.

[ August 30, 2002, 19:56: Message edited by: Gozra ]

[LazarusLong42]

OK, because no one asked about it... my suggestion for a ranking/ladder system, in full fledged mathematical glory. Math-impaired people may wish to stop here.

OK. For any given game, assume you have N players, each of whom has a rating, R1, R2,... RN. These players will "donate"--virtually--a portion of their Ratings to a pool which will go to the winner of the game. The total pool of points would be equal to 50*N.

The "donation" for player i will be equal to Di = (50 * N * (Ri^2))/(R1^2 + R2^2 + ... + RN^2). This donation is equal to the total amount the player will *lose* if they fail to win the game. If the player wins, however, they gain (50 * N) - Di points -- or the total of the donations of the other players.

Let's take two examples. A two-player game is simple. The total point pool is 100 points. Now, let's take two players with rating 1000 and 1200 respectively.

Player 1 donates 50 * 2 * 1000000 / (1000000 + 1440000) = 41 points.
Player 2 donates 59 points via the same math.

So, if player 1 wins, he gains 59 points and player 2 loses 59 pts.
Conversely, if player 2 wins, she gains 41 points and player 1 loses 41 pts.

In other words, if the stronger player wins, he gains fewer points than the weaker player does if he wins. In a two-player game that ends in an agreed draw, each player loses their inital donation, then takes back 50 points. So, player 2 would lose 9 points to player 1 if this game had ended in a draw.

Now... because I'm a masochist, let's take the case of a *5* player game. Total pool is 250 pts. I'm not going to show the calculations, but I'm going to give you a quick table:

code:

---

Player  Rating  Donation

1       1300    80

2       1200    68

3       1000    48

4       800     30

5       700     23

---

Which ever player wins gains the sum of the other players' donations; each player who lost loses his donation. Of course... that's assuming a Last-man-standing game.

In some games, team games for instance, players will know going in that they plan to cooperate. In some games, mainly large games, players will agree to a cooperative win rather than fighting a final battle. Both of these are parts of the SE experience. However, they muddle a Ratings system of this sort. I'm not going to write out the math here because without being able to show real equations they'd become hopelessly illegible.

Suffice to say: if two people win, take the other three players and add their donations together. The two players each take a share of those points proportional to both their initial donation, and to the total number of points they would have earned had they won. So if, for instance, players 1 and 3 in the above example share the win, player 1 earns 71 points and player 2 earns 51 points.

I'll leave the math on that as an exercise for the reader <EG>

Comments? Flames?

Eric

[Rollo]

Thanks Eric,
now I got a headache .
Nice system (I guess... ).

I got another idea for a system. If you win, you get one point. If you lose, you don't.

Rollo

[Zanthis]

Good post Laz. Like the system. I've always favored point swapping ranking systems, as opposed to those that generate points. However, the system will still suffer some inflation as new players come, lose some points to better players and then leave. The players that stick around will slowly amass large numbers of points (although, given the speed of a SE4 game, that might take many years).

Of course, seasons would eliminate that problem. Or, instead of just resetting everyone each season, you could rescale them all to be within say a 1000 to 2000 point range.

If inflation is too bad, adding a "cost" for each game of a few points may handle that. So if the loser's donation is 40 points, the winner might only get 37 leaving 3 points to be removed from the system.

[geoschmo]

Ok Eric, I am assuming this could all be done automatically? Cause otherwise this seems incredibally complicated to me.

Also, how does it work for new players joining the ranking system later on. What ranking do they start with?

Geoschmo

[LazarusLong42]

Originally posted by geoschmo:
Ok Eric, I am assuming this could all be done automatically? Cause otherwise this seems incredibally complicated to me.

Also, how does it work for new players joining the ranking system later on. What ranking do they start with?

Geoschmo

Yeah, it could be done automatically. It wouldn't be too much work, except for the Last case (>2 players and >1 winner). Though I'm thinking that case may be set up in too complex a fashion and should be treated as a draw.

New players, to start, would start with 1000 as their rating. Later on, new players would start at the average rating of all active players. Whomever wishes to run the service would have to define how long after the Last played game a player becomes "inactive". This doesn't counter the inflation Zanthis speaks of, but it keeps the average points per player stable, thus making things fair for new incoming players.

I suspect that even +/-5% inflation in this sort of point system would take a good couple of years to show up, simply because of the speed (or lack thereof) of most SE4 games. Excluding, of course, some of the games in the Uni Cup Tourney, the shortest of which Lasted under a week. I would recommend not artificially deflating the system unless the inflation truly got out of hand.

Only one other matter would need to be settled: Say player A and player B start a game today. Two weeks from now, A and C start a game. The first game ends a month later, and two months later the second game ends.

What is A's starting rating for the second game?? I recommend it be considered the same as his starting rating for the first game, since at the time the second game started, the first game was not yet settled.

(This is not generally a problem in these systems because they're used with games, like Chess or Starcraft, that are finished within 15 minutes to 3 hours, rather than 3 months).

Eric

[geoschmo]

I don't know. The whole idea of this rating system just seems too contrived to me. It's just my personal opinion, but I would be happy with just a simple wins and losses. No points based on the record of who are playing. Just a win or a loss. And maybe just simply restrict you to one game of record against any one person. To keep people from playing the same people over and over.

Geoschmo