With allegskill...

[Icky]

The real problem with using a points-based system is that an awful lot of the activities that win games (scouting, probing, nanning) aren't really reflected, and I'm not sure if they are even tracked at all by allegiance or trackable by a ranking system.

The real game-changing players who are so effective are game-changing not because of what they can do on their own but because of how they organize/motivate the rest of the team. When someone like Weed or xPeenix joins and organizes miner d, miner o runs, gets people to the right sector at the right time to intercept cons, directs everyone to defended from a real TP2 drop and not the feint, etc, there is not and probably never will be a way to track that.

[Raveen]

I agree that in any individual game individual actions and performaces could have a significant effect and not be represented by AS. Fortunately AS has a database featuring thousands of games. If someone performs that role in game after game then they'll be winning more often than someone who doesn't. That's the theory and the fact that someone like Weed is ranked highly suggests that it works.

There is a problem with point based systems, beyond capturing the needed data. Let's take nanning as an example. Let's say you assign points for nanning a certain amount of HP back onto a damaged miner. That cannot tell the difference between saving it from attack, and wasting time nanning a miner that's about to dock and isn't under attack. Or kills, there's a world of difference between chasing down a newb in a scout to pod them and taking out a bomb run, but there's no real way to get the stats system to see the different contexts.

[Icky]

There possibly is, though. It may involve a code change, but is it possible to record the AS rank of who you podded? It may even just need to be a running total for a player for one game, and it could do something with the numbers like averaging or something (killing Weed once would count the same as 4 (5)s). Regardless I think (and it sounds like you agree) it would be an exercise in futility.

I think AS is about as accurate as we are going to get, since in a game as complex and dynamic as this one collecting accurate and relevant statistical data on all the effects a player has on a team each game they play would require a database with more storage capacity than can fit on this planet.

[MrChaos]

QUOTE (HSharp @ Jun 18 2009, 09:12 AM) I'm not saying AS is crap, or that there is a better automated system out there. I'm just saying that it's still better suited towards smaller games then bigger ones.

Ahhh you'll get a correct rank faster the smaller the game but you will get just as accurate a rank for any size game

QUOTE Also a big point to make AS more accurate would be for random games. Random balanced games would be nice but barring commanders if no player can choose what team they join it would most likely decrease stack and make AS more accurate as good and bad players would be more identifiable in results.[/quote]
Randomizing the game will not greatly effect the system accuracy just the time to reach it. I do agree that auto balancing the teams will give more even teams and should improve game play

QUOTE I am sure that AS just like (H)ELO also suffers from the fact that only the winning team gets a rank increase, thus Weed could fly for Weylin and still just get rank decrease while I fly for aarmstrong and get rank increase regardless that Weed podded me 50 times while piloting a bomber used to kill ops and miners while he assault mines the formerly enemy sectors but Weylin only buys bombers while aarm gets Missile Frigs and unfortunately not even Weed can take on 20 missile frigs although he does pod most of em.[/quote]
This is incorrect. Your being sure is your gut speaking ... *sigh* I'm not trying to be rude to you, honest it's just... I'm not really the ninny hammer I act like here and under no circumstances would I even acknowledge this effort if there weren't real, fair, honest, and exhaustive checks on the system's accuracy. This is just how I'm built and Baker too for that matter.,

QUOTE However a system which can judge on individual merit of a player could work, currently the only system that does this is the points screen at the end of the game so if a system was to work it would have to require a change in how points are gained. Perhaps a system which combines both a points based system and AS might lead to interesting results so a player who has many points yet loses vs a stacked team (Weed) can still increase rank, but yes needs a change in how points are gained so not to exploit and ensure that even Bard can get points while helping his team to victory.[/quote]
Please read the long ass response to Pico. The current points at game end is probably one of the worst possible ways I can think of to rank players in the community... even SpideyRank is more accurate IMHO.

That being said I can see why people like points for things they do, it's why games have them ffs. However to say that by arbitrarily assigning points for in game actions and rig up some ranking system will give accurate results is just not correct.

Im going to let you in on a concern that I have with AllegSkills. It's so accurate and get's to a player's correct rank so fast it discourages players. We are not XBox Live with 10,000s of players in our game base coming and going but about 500-1000 active players so the need to wait for about 50 or so games to get a real good grip on a rank is not a big hardship. The player improving will get an adjustment in rank obviously, but at the end of the day it takes no time at all for those who have real Allegiance talent to get to their proper rank and then what.... people like to "earn" things. I think this is what people are really saying about points. I did a bunch of stuff in a game my team lost and where is my cookie?

tl;dr points blow chunks for ranking but people like to earn @#(! for time played

[guitarism]

So Baker. How much game data do you need to help with your AB algorithm balancing solution so it doesn't get borked.?

In other words, how many games should I command with AB on.

[SpaceJunk]

It would be nice if ship class usage time % was recorded.

An int pilot joining a team with many int pilots and few scout pilots should have more AS weight than otherwise.

[MrChaos]

QUOTE (SpaceJunk @ Jun 18 2009, 12:04 PM) It would be nice if ship class usage time % was recorded.

An int pilot joining a team with many int pilots and few scout pilots should have more AS weight than otherwise.

NO

[zombywoof]

QUOTE (guitarism @ Jun 18 2009, 09:33 AM) So Baker. How much game data do you need to help with your AB algorithm balancing solution so it doesn't get borked.?

In other words, how many games should I command with AB on.

Best attitude of the thread award goes to Gui.

Anyways, AS is not, nor did anyone claim that it was, perfect. It's more of a guideline, kind of like IQ. Sure you end up with your Feynmans of the AS world (richard feynman's IQ was tested at around 120 or lower, and he's considered one of the most brilliant physicists ever) but just as you can generally assume an IQ 190 is smarter than an IQ 120, you can rest assured that the 19 on your team is probably better than the 12 on the other team. There is never going to be a substitute for using your eyes and brains when it comes to figuring out how even the teams are, and AS isn't even trying to be a replacement. It's trying to be a tool to help you think, not something that thinks for you.

That last bit is very much my opinion on Allegiance Skill, and I'm not sure why there's so much drama surrounding it. Who $#@!ing cares what your rank is? Everyone who's important knows who's good and who's not. Weed wouldn't suddenly stop being able to aim if his rank was set to 5, and I wouldn't become uber just because there was a 28473 next to my name.

[Raveen]

To elaborate on MrCs post:

QUOTE (SpaceJunk @ Jun 18 2009, 06:04 PM) It would be nice if ship class usage time % was recorded.

Yes

QUOTE (SpaceJunk @ Jun 18 2009, 06:04 PM) An int pilot joining a team with many int pilots and few scout pilots should have more AS weight than otherwise.

No

[sgt_baker]

QUOTE (Icky @ Jun 18 2009, 04:43 PM) There possibly is, though. It may involve a code change, but is it possible to record the AS rank of who you podded? It may even just need to be a running total for a player for one game, and it could do something with the numbers like averaging or something (killing Weed once would count the same as 4 (5)s). Regardless I think (and it sounds like you agree) it would be an exercise in futility.

I think AS is about as accurate as we are going to get, since in a game as complex and dynamic as this one collecting accurate and relevant statistical data on all the effects a player has on a team each game they play would require a database with more storage capacity than can fit on this planet.

Something like this has already been thought about for AllegSkill v3+. Some way off, but for certain in-game activities, such as whoring, it is possible, using a technique similar to the one you've suggested, to generate a whoring rank. It's not quite as simple as it sounds, since a newbie int a heavy int podding a vet in an unshielded scout doesn't mean the newb is uber.

In other words, we've been thinking of how to provide extra useful and interesting stats for a while now. It is not, however, our top priority at this time.

QUOTE (guitarism @ Jun 18 2009, 05:33 PM) So Baker. How much game data do you need to help with your AB algorithm balancing solution so it doesn't get borked.?

In other words, how many games should I command with AB on.

It's not that we're lacking data at the moment. I have events data (join/leave/who podded who/what base was built where) for 10s of 1000s of games. The tricky part is is modelling human behaviour when trialling a post-launch AB system. Take this example: Whilst iterating through the join/leave data for any given game, we see that VoobX joined Team 1 at 10:30. In the actual game, VoobX contributed to Team 1's stack. In our trial, AB would intercept that join. What the player does after the interception is pure guesswork on our part. Does he accept AB's decision and join Team2? Does he wait until the stack has evened out and join Team 1? Does he get pissed off and leave altogether? etc etc etc

I might post some polls to determine peoples' general attitudes towards AB decisions, with which I might be able to arrive at some vague guesstimates for the aforementioned factors.

B