Official forums for BZFlag.org

Ok, this will probably be abit hard for me to get acrossed, but bear with me.

When I read peoples suggestions and problems toward the teamscores (including my own activity factor) it seems like people have different expectations for what teamscore means.

I see two different purposes:

Teamscores is secondary when matching
This is what we got now, as I see it. Teamscores isnt the main goal when matching. Having fun is. Teamscores is just to make it abit more interesting, but isnt really required. This is how I play. I dont care less if we lose a match, because teamscore is secondary.

This is why ELO fits this purpose very well. All teams can play whenever possible, no planning (tournament-like) is required. And it handles that strong match again weaker team. And that same teams play against eachother time after time again.

It fails when players want their team to be the best. First of all. Teamscore is just a skill-estimation. As every team consist of many players they can create a number of combinations of 2-3-4 players to match other teams. Each combination would vary (a lot) in skill. I think that is why teamscores fluctuate so much for teams.

So teamscore is basically some random number within the skillwise highest and lowest combination of player. This is bad if you want to see you team improve and "win it all". Because if team X and Y plays against eachother their "random" teamscore decides how much is earned/lost for that match.

Teamscore is the mean to "win it all"
This is more competetive approach. Players play to be the best, to be the top ranking team. Here activity (and therefore oppurtinuty to play them) is important.

Activity can be measured, but we still faces the same problem with "random teamscores" and hence random rewards for wining a match. So activity factor dosnt really fix anything.

Resetting scores wont do it either. Well it will, in the short periode were all teams are (nearly) even in score.

An alternative is to use an similar approach as i did with the ladder. Instead of rating every team then begin to rate every player (might make them more active too?). And let general teamscore be based on the average (or similar) of the players. When two teams match eachother then the teamscore is the sum of the players participating. This would probably make teamscores abit less fluctuating, beause a team strongest combination of players would give them a high score while they weakest players would give them a lower score. Then activity would need to be expicit handled - and everything is complicated.

But still, if allowing teams to "win it all", is the purpose, then ELO simply not feasible. Then take the full step and structure matches give fixed points for loss/draw/wins. Let team themselves worry about who will match when and reset scores every year. This is easy to understand, but eliminates the "hey lets play a mach"-matches. And activity is shown directly in the point (if you lose 1 point, draw 2, win 3) - inactivy is rewarded with zero points.

So my point is. What do we want with teamscore? I've outlined two extremes. The solution is probably somewhat in-between. But where? Im not sure what direction people feels but the suggestions I've seen seems to suggest that people want both. Which is hard to get

The ELO ranking in chess describes what is the skill level of a player, meaning how hard it is to beat that player. It doesn't rate the players performance.

The matches that determin the ELO rating of a player are usually those in the context of tournaments (championships) and competitions.

Competitions means everybody plays everybody once (or twice), tournament means as you progress and win a match in a tournament, you enter the next round and at the end you match the best of the other half of the tournament tree.

Both models we don't have at all for the ducati league, thus i agree completely that

ELO simply not feasible

--

A question. You described both interests:

1. Teamscores is secondary when matching
2. Teamscore is the mean to "win it all"

In what way would the first group Teamscores is secondary when matching suffer if using a competative model as described in
Teamscore is the mean to "win it all".

Without reading your post (dum me) I posted the following post: http://my.bzflag.org/bb/viewtopic.php?t=7983.
In that we can keep an ELO rating to describe the average skill level and a points system for the team ladder.

One thing to keep in mind about comparing bzflag to chess is that in chess, when you play a rated game, you're matched up with players of similar ratings. In other words, there's no way that a 1300 rated player is matched up against a 2200 rated player unless it's a really small tournament. Usually, a few dozen people show up, and you're matched up in groups of either 4 or 6. Obviously, there is no matching up of team skill levels in bzflag. It's just an ad hoc play a match as you can sort of system.

Maybe a different rating system would work better? Perhaps just a raw wins, losses, ties rating?

mistake wrote:A question. You described both interests:

1. Teamscores is secondary when matching
2. Teamscore is the mean to "win it all"

In what way would the first group Teamscores is secondary when matching suffer if using a competative model as described in
Teamscore is the mean to "win it all".

To be hornest i cant remember what i belived was conflicting between the two groups