A few of the members of our gaming have had the probably-not-unique idea of statting ourselves as if we were player characters in a campaign. Our problem is trying to find a basis of comparison for each stat so we can properly assign our scores. Like, what would a STR of 10 indicate? Or a 10 in any other ability, since we're not messing with race.
The online survey What Kind of D&D Character Would You Be? calculates your ability scores based on answers to questions.
The trouble with basing character stats on the members of a D&D group is that it tends toward high Intelligence and lower physical attributes, so you often end up with an entire party of wizards. You can still do this: read Complete Arcane's section on how to run a balanced game when all players are arcane spellcasters.
Alternatively, you can use the following benchmarks to measure your attributes. It may not be entirely scientific, but for the purposes of a D&D game it should be sufficient.
Instead of looking at what exactly a strength of 10 indicates (like being able to lift a certain amount of weight or so), try looking at it as a place within the population at large. A score of 10 in any stat means you're just about average; a score higher than 16 means you're in the top few percent of the population.
There are two approaches we could take here. One is to look at the distribution of ability scores in D&D and compare it to the distribution of that ability in the population at large. This is approach I'm taking here. The other approach is to use the correlating factors published in various different D&D sources. This is the approach many other people have taken (such as the fun quiz made by Kevin Haw).
Assuming you're using 3d6 for each stat, here's a rough guide for you:
To get these numbers, I'm looking at the results of rolling multiple 6-sided dice and adding them together. As you add more and more dice, the result gets closer to what's called a normal distribution. It's a curve that's well-populated in the middle and sparsely-populated on both ends. In other words, there are lots of people who are typical, some people who are either high or low, a few people who are very high or very low, and a few extreme examples at both ends.
So let's take a look at intelligence and see what these numbers would mean. IQ is calibrated so the average is 100 and the standard deviation is 15. This lets us correlate D&D stats with IQ scores, as well as some common terms (found here):
Unfortunately, not all stats can be quantified so easily. Information on intelligence is easy to look up; info on charisma or wisdom, less so. Hopefully you can use the ideas here to estimate the scores you're looking for.
A first level human barbarian can out-run Usain Bolt’s fastest sprint.
Continuously for several minutes without getting so much as Fatigued.
While wearing a chain shirt, wielding a greataxe, and carrying nearly 100 lbs. of assorted other gear.
Dungeons and Dragons 3.5 cannot model actual people. Even 1st-level human characters using the standard ability array (11, 11, 11, 10, 10, 10) and restricted to non-magic NPC classes cause problems. Maybe if you eliminated Feats, as well. Anything beyond that is patently unrealistic. Dungeons and Dragons is not the system for that.
If you merely want to play characters inspired by who you are, then it’s easy: just imagine that you were growing up in Eberron or Faerûn or Greyhawk or wherever, and think about what sorts of things you would pursue. Not sure that’ll be too interesting, though, if we’re being honest: for me that would probably mean holing myself up in an arcane university and never leaving.
Here is my suggestion:
Talk with each other and figure out who would be best suited for each role in the party. Who seems like they would be a Wizard, who would be the fighter? And so on. Then stat the characters as normal DnD characters. You make small adjustments like: "Timmy would definitely be a rogue, but he would be almost as strong as he is quick. He's also not the nicest guy all the time so I'll dump-stat his CHA."
Don't try to get your "actual stats" down on paper. First of all, your characters will not be very successful statistically. If you "realistically" try to stat yourself, you are not going to be getting many, if any, stats over a 14 and you run the risk of having to tell your buddy that he really only has about an 8 in Wisdom (citing recent parties and youtube videos he was in). This can lead to some sad gameplay and hurt feelings.
Remember, in the end it's a game that is meant to be fun. If you want to play as yourselves, the most important thing to bring into the game is your personalities and your friendship. That will make the game fun.
I haven't played 3.5, so this is from 4e, but as far as I know the stats are pretty similar, but I have no knowledge on how stats are determined in 3.5.
What I would do is determine which player would have the highest of each stat. This person would get a 18 for that stat (I would avoid giving more than 1 or 2 to a single player). I would then determine which player would have the lowest for each stat, and they get 8's in those stats (again, I would avoid giving more than 1 or 2 to a single player). I would then rank the remaining 4 stats from highest to lowest for each player, and assign them (from highest to lowest): 14,11,10,10. I would allow players to modify these middle numbers up or down 1 point, but for each one they move up they should move another down one, and hopefully keep them in the same order. None of these should be less than their weakest stat.
What this does is it allow players to show off both their strengths and their weaknesses without having to try and quantify what each number represents in real life. I would really strive to try and avoid ties because this would make each character's skill set unique, and this would affect which class their character would best fit.
It's probably quite difficult to find now, and doesn't quite match your requirements, but...
The first edition of Timelords, by BTRC (and, according to the Wiki entry, the 2nd ed too) had the base assumption that players would play characters based on themselves. It had reasonably detailed rules for generating each of the stats used in the game, including skills. It was a system that used d20s (though long before the d20 system was trademarked), so it may well still work for you with a little tweaking. The only stat missing would be a Wisdom analog.
Note that the 3rd ed of the game appears to have done away with this idea, so likely won't be much use to you.
I suggest the following method:
Everyone should stat himself (or herself) and all other PCs. The DM will take the average for each stat, and can adjust it by +/- 2 max (per characteristic).
I stat myself as
the others 3 players think I am actually:
My resulting stats (averaged):
The DM can adjust each stat of up to 2 points upward or downwards.
Rationale: it's not a matter of what 10 "means" in real world. First of all it's too abstract, and also you lack any real terms of comparison (is the average D&D citizen a person who tills the fields all day, so it's much more strong than average modern human? is the average D&D person illiterate and therefore has less intellectual resources at his/her disposal?)
If you are perceived as "strong" or "dexterous" by other players, they know the game enough to "stat" you. And the DM has the final decision.
It works best if you have the initial statting done without communicating with other players (to avoid the risk to change your statting depending on how favourably the others treat you - this may be even subconscious, so it's better to do this "secretely").