If you've not heard of the Elo ranking system before, it's probably worth a quick read of it's Wikipedia article. It will give you a much better understanding.
The K-factor of a rank controls the scope of change between a win and a loss. In other words, it's how confident the system is that the final result is a fair representation of the players skill. The larger the K-factor the more results it will use to determine a rank where as a smaller K-factor will resist the urge to change a rank too soon.
Occasionally players will enter the system, compete in a few games with experienced players and lose. They will then not play another game in a long time. Effectively, this dumps extra 'points' into the ranking system that are then gained by experienced players. The problem we have here is that experienced players will gain too much and it will be extremely hard for new players to catch up.
To counter act this inflation problem we use a much smaller, fixed K-factor for new players until they have competed in a minimum number of games. This simply limits the rate at which new players will lose points and prevents a large dump of points into the system until the player has gained experience.
Each player has a Rank title assigned to them. This title has absolutely no effect on the Elo system nor is it effected by a players position. Instead, rank titles are determined primarily on amount of games and the players win ratio.