In my opinion, skill should be based on a combination of win/losses and personal performance, and this balance should change depending on the game type.
Purely basing the system off of wins and losses discourages me from hopping into 4v4 matches with random people, because I may perform very well and still lose, or I may have a horrible performance and win. Both cases happen fairly frequently and thus their affect on my skill rating should be mitigated.
I believe including personal performance statistics can help mitigate this affect. Some argue that using personal performance statistics can reduce the incentive for teamwork, and I ask you to look at how players are tracked in sports. A player's team's effectiveness is not included in their personal statistics. We judge NBA, NFL, MLB, etc. players based on their individual performance. Great individual performance is a good indicator of winning ability, but not necessarily the other way around. And when we look for another Halo 4 players skill rating, we are usually looking for an indicator of their individual performance anyway.
The old True Skill system (based strictly on win-loss, derived from the ELO system) is great for ranking one-on-one games, like chess. But Microsoft has been using it as an all-purpose ranking system for all of Microsoft Studios games. I think headshots, assists, melee kills, and other skill kills (accuracy and distractions too) are more important and much more indicative of skill than a simple win or loss.
Now that being said, K/D ratios make sense to be a part of the skill rating in slayer matches, and assists should be highly emphasized (maybe even 1.5 times the weight of a kill, this would drastically increase teamwork and reduce the desire to "steal kills"). But for capture the flag, defensive and offensive plays should be the basis. For other objective game types, K/D ratio or spread should be only a marginal factor in skill rating.
In my opinion, as was partly the case in Halo: Reach, your score in a match should reflect your performance in that game type (slayer, oddball, ctf, etc.), and should tie in directly with your publicly visible skill rating, so that you know how your skill rating is calculated.
ADDENDUM: I think it would also help if when looking at a players statistics you could view their Slayer K/D ratio separate from other game types wherein kills are not the objective. This would reduce the desire for players to jump into CTF games just to get easy kills on those putting in genuine effort to play CTF. With the reduced anxiety about dying too often, players would be able to fully enjoy the great objective based games Halo 4 offers.