Semantics? While understanding other people you can't see with a few sentences leaves a lot of room for interpretation, what you said seemed to imply that you were finding notes that would enable high returns recently without jumping on the 'dump times', which with my limited experience would seem to require mostly E, F and G notes, and you said you don't use an automated service. Over the last two weeks, using frequent random look-ins, I see generally 30-50 notes, most of which are A or B grade (which wouldn't support a 15% return), very few notes past a D, and the E-G notes I've seen had very low FICO, very low income, 3 or more inquiries, several recent deliquencies, numerous public records or some other serious defect that statistically would seem to indicate a high default rate, which would likely dip you well under 15%.
I think the gist of it is that I hardly see ANY notes in a range that would sustain a high rate of return, let alone finding them without 'extreme difficulty'. I had good luck this morning in the 60 seconds some notes were available at the 9am drop time, but I'd actually term the ability to find E-G notes at all the last few weeks to be 'nearly impossible'. Of the bunch I managed to buy this morning, only one was an E, one D, and the rest C and above.
So I'm guessing you're A) putting high 3, 4 or 5 figure amounts into each note, B) buying and selling on the secondary market and getting some great discounts, C) you're accidentally buying new notes in the one minute after the drop time when decent ones are still available on a persistent basis, D) you've developed your own automated system (which I've considered, although the last time I coded I was excited about PL/1 because it was so much better than Fortran), or E) you're a statistical anomaly.
There is also another option, which is one I weigh frequently. In mutual funds, actively managed funds always trail index funds over the long haul, and the number of managed funds that do exceed their index for more than a few years is much lower than it should be, given even a random distribution. As William Bernstein noted, monkeys throwing darts at the Wall Street Journal's stock page would give better results on a random basis than we see from 'investing professionals'. Much like that, the % of LC investors with high rates of return is lower than it should be, indicating that like many investments, a person trying to employ strategies does worse than someone buying a big diversified bucket of a product and then doing nothing with it. Some LC reps have implied that to me, saying that most large investors trust LC's vetting process and just buy on grade. Given that my analysis of prior results as a predictor of future results tells me that most of the data people go on is at best a bit skimpy and was effected by situational influences unlikely to repeat, it seems there is a great analog between stock market performance and loan return performance in terms of guessing where a stock/loan is going to go.
With all that in mind, it seems the wisest choice would be to A) put your money to work, B) diversify, and C) buy every E and F note you see, as G's high default rate drag their end returns to below E's and F's, and D) don't bother looking at anything else and when something goes wrong with a note, just stand there and do nothing. But historically, that wouldn't give you 15%, but closer to 12.5 to 13.5, which is exactly where my returns are with a couple of thousand notes over 18 months. Although my results are dragged down from buying B's and C's to get a lot of capital deployed in a short time. I expect my returns to move higher as the composition of my portfolio shifts from heavy B/C/D to D/E/F.
I think its also interesting to note that high return investors tend to do better with G's when holding fewer, higher amount notes, so I'll go out on a limb and guess that your strategy is to buy all and F notes, and put a lot of $$$ into them, and you've been slightly lucky or found one or two things to filter on that have been effective the last 3 years, but may not work for the next 3.
All this is very interesting and I'm really glad to have found people willing to talk about what they're doing. My LC "experiment" has proven to be a very educational and very interesting endeavor.