Player Ranking in Allegiance

takingarms1 · Post by **takingarms1** » Sat Dec 15, 2007 4:28 am

Compare test results?

jgbaxter · Post by **jgbaxter** » Sat Dec 15, 2007 4:29 am

takingarms1 wrote:QUOTE (takingarms1 @ Dec 14 2007, 09:28 PM) Compare test results?

Well if they have the AllegSkill system mostly done, they can run a side by side comparision of ratings?

peet · Post by **peet** » Sat Dec 15, 2007 11:58 am

Ranking systems can't calculate the teamwork efforts of one player, as far as I know.

It can "only" calculate events, like hitting something, capturing something, picking something something up, dropping something etc. If you know how the ranking system works, you can use that knowledge to collect points or increase your rank. These are all individual points for one player. The ranking system promotes individual behavior like kill a lot, capture a lot. It doesn't calculate the quiet team assisting players.

For example the spy who collects valuable data about the position of enemy miners get little points. The players that use that spy data to destroy enemy miners get the points and the glory of a high position at the end game stats page. Same for the players carrying fuel and ammo packs for their fuel hungry teammates ints - little points but valuable teamplayers. Doesn't matter if you're on the winning or losing team. You can have < 10 points and be a valuable teamplayer or having > 50 points wasting 50 minutes flying in circles and spend 10 minutes to kill 2 miners. You must be a strong commander to boot the selfish team whore when he's killing miners and enemies and being a complete @#$%@#...

Now about (H)Elo, it's basically the same individual system. It promotes selfish behavior, if you pick the strongest team you have more chance to win, and increase your (H)Elo. Some players lower their (H)Elo on purpose, they can squeese in at an important game with "too low" points and join the "buddies" team causing havoc.

Most Alleg players are competitive players, nothing wrong with that. It's nice to be on top of a list.

IMHO, the best ranking system is the good commander. (S)He knows the player base and the players reputation. Good commanders will balance the teams, or no game will start due to lack of an enemy team. They will create a mixed team with experts at every field.

sgt_baker · Post by **sgt_baker** » Sat Dec 15, 2007 11:59 am

spideycw wrote:QUOTE (spideycw @ Dec 15 2007, 04:23 AM) "With somthing proven to be more effective"

How do you prove if something is more effective or not?

As noted, and to differentiate the question from 'assumed to be more effective', by making a side-by-side comparison of Helo and AllegSkill.

Adam4 · Post by **Adam4** » Sat Dec 15, 2007 1:00 pm

Tigereye wrote:QUOTE (Tigereye @ Dec 14 2007, 04:36 PM) Don't listen to Malicious Wraith. He's a newbie. Look - only 4 posts /rolleyes.gif" style="vertical-align:middle" emoid=":roll:" border="0" alt="rolleyes.gif" />

/owned-anim.gif" style="vertical-align:middle" emoid=":owned:" border="0" alt="owned-anim.gif" />

--TE

Tiger /wub.gif" style="vertical-align:middle" emoid=":iluv:" border="0" alt="wub.gif" />

Zapper · Post by **Zapper** » Sat Dec 15, 2007 1:10 pm

So what actions score's points??
Cause that is what make Trueskill viable.
If you did make a addaption of Trueskill!!

Z

takingarms1 · Post by **takingarms1** » Sat Dec 15, 2007 2:15 pm

@peet - in point of fact TrueSkill purports to be precisely a system that gages the teamwork effectiveness of each player. Google TrueSkill and do some reading, also wait until the learned Professor Baker presents his most interesting data and conclusions before drawing conclusions of whether it will or will not work.

sgt_baker · Post by **sgt_baker** » Sat Dec 15, 2007 2:18 pm

Zapper wrote:QUOTE (Zapper @ Dec 15 2007, 01:10 PM) So what actions score's points??
Cause that is what make Trueskill viable.
If you did make a addaption of Trueskill!!

Z

Trueskill doesn't measure a players skill by recording specific in-game actions. In fact, it functions in almost the exact opposite way. Trueskill makes the assumption that the game outcome is the result of the cumulative effect of the unobserved in-game performances of a team, and that the game outcome contains small amounts of information regarding the skill of each player. Over time (i.e. when TS is allowed to consider a large number of games) and regardless of the specific actions being taken by any given player, Trueskill is able to determine that something the player is doing is contributing to their tendency to win or lose by accumulating the information contained in the game outcomes. In an effort to appreciate how Trueskill functions it is beneficial to give some thought to the advantages of the system's indifference to the specifics of in-game actions.

This indifference is, in fact, one of the principal reasons Trueskill is effective for generating skill ratings (and the beliefs therein) in complex games such as Allegiance. Whilst it is relatively simple to assign points-based metrics to numerous in-game actions (kills, nanning etc), there are a number of abstract performance factors which are incredibly difficult to interpret in terms of scalar values. Situational awareness, 'being in the right place', ability to predict and correctly interpret the 'flow' of the game/enemy are a few factors generally accepted to be strongly associated with highly-skilled players. How would one even begin to assign points to such attributes? Trueskill is able to measure these skill-contributing traits on an equal par with those which are easier to interpret.

To draw an analogy (if not a somewhat silly one): If someone were to introduce a new win condition into Alleg whereby making a cup of coffee half way though the game would improve the team's chances of winning, Trueskill would eventually assign higher skill ratings to those players who made their cups of coffee. It sounds bizarre, but it is one of the wonderful effects of using a system based in statistics, and avoids the thorny issue of having to explicitly define the in-game value of coffee.

I'm not convinced that this explanation is as clear as I could possible make it. One of the reasons I'm taking my sweet time in presenting the detailed workings of AllegSkill is that I'd like to make absolutely certain I'm expressing the important points in the clearest possible manner.

/smile.gif" style="vertical-align:middle" emoid=":)" border="0" alt="smile.gif" />

B

sgt_baker · Post by **sgt_baker** » Sat Dec 15, 2007 2:23 pm

P.S. Microsoft have a site where they explain Trueskill in humanly digestable terms. http://research.microsoft.com/mlp/apg/trueskill.aspx Check it out.

Sakana · Post by **Sakana** » Sat Dec 15, 2007 3:09 pm

just a thought (i know i am newb, have no experience of HELO and whatever you could flame me for):

TS seems like a great system to evaluate the overall benefit of a certain player to a team by showing a single number next to the players name, be it for his whoring, commanding probing/deprobing skills. (i'd vote for it but wasn't able to, although i can't state wether HELO is working or not))
BUT
imho it would be good and useful for the comms to be able to see some statistics of the player, like the kills:podded,win:loss ratio, in order to be able to have a team which is balanced regarding the number of people who are good at fighting and those who are otherwise useful.
this is not intended to give the commanders the opportunity to ignore newbies, just to have balanced teams and thus have more exciting games (i.e. with an relatively unpredictable outcome, assuming that stacked games are boring).
the TS system would still enable newbs to be picked because there is? a limit to how much difference between the teams (TS) scores can be.
i don't think having that statistics available in NOAT in the near future is realistic, because there are more important things to do i believe, but just to think about if it wouldn't be a useful option to implement somewhen in the future.