Originally Posted by blu.knight

The theory to this situation is the same with any situation...

It should balance itself out. Yes, it might value them high... but if it does so and puts them on the marine side, that team will be weaker and they will lose... which will bring that ratio down making it more accurate.

--

The absolute best system would be a system that managed to do the following:

- Track who is normally on one team when they win. This would be used to try to determine who works well together, and who just has enough skill that their teammates do not really matter. (I know with me there are certain people who, when we are on the same team, tend to win much more often.)
- ALIEN: Track each lifeform's frequency and skill level (skill level might include damage dealt and the average lifespan of your lifeform). This would be to see what lifeforms/roles the person seems to prefer, and what they are good at.
- MARINE: Track accuracy, track how often they are solo... Track K ratio
- MARINE: Track who often com's.
- Track who is often picked first in Captains games. This adds a "human computation" element to it... And would give an added weight to those that we know are valuable teammates... Though the rest of this should have been an indication of that already. This also would account for things like how likely someone is to sacrafice for the "greater good". Those that are big team players tend to be picked earlier... but it's very difficult to try to determine who these people are just by code.

And then, when balancing it would...

- Try to make sure there is a good fade on the alien side
- Try to make sure there is a com on the marine side
- Try to make sure there are some people who work well together on both sides
- Try to ensure even skill distribution. Give any extra skill to the marine side. (one will be commanding, so their skill in the field does not matter)

--

That was amusing to type up.
I would love to make a system to try to keep track of those that work well toegether and everything. I think it could be really interesting, part of that whole "social mapping" idea. Represented graphically people like Golem would have tons of connections but none would be very strong. People like me would have less connections, but ones to people like Radix would be stronger.

It would just be very interesting to see.

I wouldn't mind seeing someone attempt this, but I doubt the results will be very good. This kind of piecemeal approach gets incredibly complicated, meaning it's very easy to leave something out or include something redundant. In addition, figuring out the weights for each factor is virtually impossible without some sort of machine learning algorithm (which I don't believe we have the data for). Also,keep in mind that explicitly considering who might comm or lifeform will actually make the balance worse when that person doesn't comm/lifeform, and forcing them to do so isn't really an option. There are ways around this, but it involves having people type their preferred role for that game while still in the readyroom, before the command is executed. If that were done, in theory this could be incorporated into the current algorithm after we collected data for a few months.

Thane does present a good point, which is that if a player is 0.9 on marines and 0.1 on aliens, they will stay at 0.5, and this will situtation will not improve itself unless that player gets better at aliens. This is a problem we knew would be present from the beginning, and is at the top of my wish list for things to be worked on if someone with access to the server had time to test it. Unfortunately, even if we could work on it, there is no simple solution. The main problem that we run into is that as soon as you break someone's W/L into these two parts, it becomes difficult to determine the order of player assignment without repeatedly assigning players to the same team (people tend to get assigned to whichever side makes them closer to 0.5). Given how many complaints we've had about that when it's not actually happening, I'd hate to see the reaction if it were.