P2P-Picks Brings Hedge Fund Style Loan Picking to Small Investors

Hedge Funds have many advantages over individual p2p investors. One of them is that they have Ph. D level statistics gurus combing through the loan history of Lending Club and Prosper in order to develop sophisticated investment models.

Enter Bryce Mason. He has a passion for statistics and p2p lending. So, he has spent hundreds of hours developing an investment model for his Lending Club investments, a model that is similar to what the large hedge funds use. And he is about to make it available for small investors.

He first told me about his work at the beginning of last year and he has been sharing the picks from his model with me since last summer. A couple of months ago he launched his website, P2P-Picks.com, in private beta and a small group of investors have been putting it through its paces.

What is P2P-Picks?

The simple graphic above demonstrates what he has done. He has analyzed the historical loan data at Lending Club focusing only on those loans that have matured. This way he can work out exactly how much each loan has returned for investors. And we are talking about quite a large sample size of loans. More than 10,000 Lending Club loans issued from 2007 through March 2010 have reached maturity.

P2P-Picks uses all the data in the historical file (and even some data that is not) in order to determine which loans on the platform will perform best. Bryce has developed a ranking system that can assign an expected return for every loan. The loans with the best expected returns (in the top 10%) are shared with investors on his site.

How much of a difference is there when investing with this model? Below is a chart that shows the projected performance. As it says this chart is based on backtesting the model on the loan history. The X-axis of this chart shows the percentage of loans included and the Y-axis shows the expected return. As you can see if you target the top 10% of loans the return can be expected to be between 12% and 13%.

How P2P-Picks.com Works

P2P Picks works off the Lending Club BrowseNotes.csv file. This is the file of all the available loans on the platform that any investor can download. Every 15 minutes P2P Picks downloads the latest file and then runs every loan through its proprietary scoring model. It then ranks each loan based on expected return and presents the top 10% of loans to its users. Then users can click on buttons to invest in $25, $50, $75 or $100 notes.

P2P Picks will not do your actual investing for you. When you click the green Go button on the site you are taken to Lending Club’s site where it will allocate the note amount selected to the particular loan.

This is how I use it. I look at my available cash to determine how many loans I need to get back to being fully invested. I then start at the top clicking on the $25 buttons of each note. Then I click the Go button which will launch a new browser window but then I quickly come back to P2P Picks again and click the next Go button.

Then when you are done you stay on Lending Club’s site and click View Order. Then Confirm the investment, assign it to a portfolio and you are done. Once you get used to it you can easily invest in 15-20 loans in less than a minute. I have been doing this, on average, once every two days since the site launched.

What is the Cost?

Right now while it is still in beta the price is free. Once it launches you will pay 0.5% of the value of the notes (50 basis points). So, a $25 note will cost $0.13, a $50 note costs $0.25 and so on. Then at the end of every month your credit card will be charged for all the picks you have made in the preceding month.

P2P-Picks is Launching in May

Almost $200,000 has been invested now by over 100 investors through P2P-Picks and the numbers are growing every day. The system is very stable and I have been using it on an almost daily basis for several months now with no problems.

I have been investing a portion of my main account with this model since August of last year. I have 325 notes totaling just under $10,000. While it is obviously too early to make a judgment call yet on how his model is performing for me I can say this. Of the 325 notes I have invested in I have had no charge-offs, there are no late loans and there are just two notes that are in grace period. My Nickel Steamroller ROI for this portfolio says it is 16.79%.

Bryce is intending to make P2P-Picks available for everyone next month. As of late April it is still in private beta but if you are interested you can join the waiting list by filling out the Contact form on his site.

I welcome a site like P2P-Picks coming online – it is one way to help level the playing field between the hedge funds and the small investors. What do you think?


  1. says

    Great writeup Peter. I too have been using P2P-Picks for a few months now and currently have approximately 120 notes between my two accounts invested in this method. I am a big fan of Bryce’s work and believe P2P-Picks has the opportunity to be successful as he provides a quick and simplistic means to invest.

    Beat me to the punch on posting a review! 😉

  2. Dan B says

    Considering that there have been somewhere close to 59,000 notes issued since August 2012, 325 notes seems like a really tiny amount for 8 months worth of (several times a week) investing within the top 10% of notes offered. Why is that? Are you doing additional filtering of the 10%?

    Also………..what is the average age of this 325 portfolio? And the 36 to 60 month balance?

    • says

      Dan, I have not added any new money to Lending Club for several months so this is all reinvestment. The 325 notes represents close to 100% of my reinvestment in my main LC account over the last few months and they are all 36-month notes. The average age of this portfolio is 3 months, so obviously still very young.

      • Dan B says

        I thought you did add $3k to that account last quarter?
        Regardless, I understand that these are mostly reinvestments. I (obviously) also understand that, for a variety of reasons, pretty much no one would attempt to invest in every note recommended. Still…………..In order for me to put future potential returns in perspective, I’m curious as to what your involvement was in the process of note selection. So, did you do additional filtering to come down to that tiny 325 number? ……………. Because that 325 number is obviously a very small percentage of the already filtered “top 10%” of notes.

        • says

          I have not added any money to Lending Club since June of 2012.

          As for my selection process with P2P Picks, I use the system pretty much as is. The only tweak I make is to exclude the A and B grade loans, of which there are not many. I realize I am not investing in the entire 10% population of loans but the 325 notes I have invested is what has been needed to keep me fully invested in my main LC account.

          • Dan B says

            Makes sense. Obviously I don’t plan on using this or any of the services discussed in this blog, but I do find it of enough interest to warrant a more detailed understanding. Thanks for clarifying.

  3. says

    Thanks for the write-up, Peter. I’m very eager for the next 3-6 months. We’re going to have even more value for our subscribers. I’ll keep an eye out on this thread throughout the day, in case readers have questions.

  4. taiganaut says

    Meanwhile, Prosper is cutting more data out of their feed, setting back some efforts to do the same on their struggling platform. I’m not pleased with Prosper’s management right now AT ALL over their decisions on data disclosure (especially bids, thus making “blenders” impossible to find going forward).

    Worse, LendingClub STILL is not legal for investment in Oregon. What. in. the hell.

    • says

      I think the changes in the Prosper data will not affect Bryce’s model at all. I know that he is in talks with Prosper to develop a similar thing for their platform.

      And I completely agree. It is ridiculous that you cannot invest in LC if you live in Oregon.

      • taiganaut says

        I’m playing with my own software, and the ability to spot “blenders” – lenders who were now borrowing – was a quick and easy way to find loans that tend to perform a lot better than a random sample.

        Prosper is eliminating sharing of bid data, which eliminates this (very cheap, as compared to extensive data mining and classification) method of finding potentially good loans.

        I love P2P lending as an idea. Love it. I want to invest more in it than I already am. But Prosper seems to bungle a lot and have tech problems fairly regularly, and as an Oregon resident, I’m shut out of LendingClub because they haven’t found the right toady to bribe in Salem yet. :(

        • Chris says

          I feel your pain. I also am in Oregon where the only game in town is Prosper. It Makes no sense, why regulators would approve Prosper but not Lending Club. As there only 98 loans to buy on Prosper right now, I would LOVE to be able to click into Lending Club and peruse theirs.

          • taiganaut says

            It helps to know when the big investors are out vacuuming up loans on Prosper. They should limit the max percentage any one investor can buy of a given loan. I stay away from it near the end and beginning of any month… more to buy a week into the month. *sigh*

  5. agd says

    Bryce, great job! I’m just curious as to the discriminatory performance (AUC/ROC) of your default model? How does this compare against that of LC’s newer risk models? Also, for your ROI backtest, I’m guessing this is based on the in-time APR, would LC provide what the pricing would be under today’s risk models?

    Again, great job. I’m just curious as to whether the edge your method is providing is due the model being just overall stronger than LC’s or that it is identifying risk in different ways.

    • says

      Hi AGD, thanks for the compliments and questions.

      One post on the boards during our launch asked the same question about goodness of fit and AUC metrics. I have since added a couple important predictors, but you can see this post at the link below. I don’t know of any more discriminating model.

      Yes, backtests used then-current interest rates. I found this acceptable because the purpose of the backtest was to compare strategies rather than quote an expected return (dangerous to “promise”). IR at LendingClub have increased ver the last couple years.

      The P2P-Picks edge comes from using more relevant borrower information and setting risk better. As long as LendingClub is using just a handful of possible interest rate buckets, there will be arbitrage opportunities.



  6. says

    Peter, thanks for sharing for those of us who were not yet aware of Bryce and his work. I’m going to sign up for the waiting list right now!

    Question for Bryce: Peter mentions 50 basis points (above), and your website says 60 basis points, in regards to the fees. Which is it?

    I am about one paycheck away from having my first $5,000 to open a Roth IRA over at LC. I have a very small non-IRA account there now, and am totally sold on p2p at LC. I’m super-pumped to get my money working harder for me through my Roth. (I’m waiting to hit the $5,000 mark so I can save the $100 for opening account). Every penny counts for middle-income moms like me with a kid in college.

    With the big day looming, I am a bit nervous and excited about picking 200 $25 loans all at once. Until now, I’ve just hand-picked to diversify a very small number of A-F loans across the board. I was planning to just use one of Peter’s filter recommendations for that 200, but I am willing to pay someone like you a very reasonable fee to try to improve my returns and not really add significant time to the loan selection process.

    I’m not enough of a math whiz to try to predict how my ROI may be better/worse by adding your fees (vs. selecting and running my own filters at no additional cost) to my $5,000 initial LC IRA investment, but I am curious about that.

    And THANK YOU for charging per pick – I am obviously one of those very small portfolio gals!

    • says

      You’re welcome! When the site goes live for new subscribers, the fee will be 60 basis points. Those in beta got defaulted accidentally to 50 bps. This level of fee roughly shares the strategy’s alpha about 66/34 in your favor. You will be able to select 200 notes quite quickly, if you can hit all 4 magical hours at LC. Maybe 1 week or so.

      • A Choi says

        Thanks for creating such a powerful yet simple tool. I’m relatively new to P2P and have deposited money but have not actually invested as I have not yet backtested my models. This tool will make it easy to invest all my money in one go. Thanks Bryce. I’ve applied to get into the beta if you open it up. Really looking forward to using your tool.

  7. Chris E says

    This is what I don’t get. Both LC and Prosper have rocket scientists (lead by ex-JP Morgan Credit Cards and Barclaycards) developing models to assign risk/yield ratings to each loan. These companies use state of the art supercomputers and have access to incredible amounts of big data to develop their models.

    So why should I believe that any of us mortals are able to pick loans or develop models that can statistically outperform LC or Prosper’s model?

    • says

      I think you are giving the “rocket scientists” a bit too much credit. And you certainly don’t need a supercomputer to process the somewhat limited dataset that is the loan history of Lending Club and Prosper.

      For what it is worth the fact is that many of us here have been able to develop simple models that have outperformed LC and Prosper’s model over the last couple of years. As they get more sophisticated we will have to do so as well but it is difficult for them to price every loan perfectly and I think there will always be opportunities to exploit mispricing.


Leave a Reply

Your email address will not be published. Required fields are marked *

Notify me of followup comments via e-mail. You can also subscribe without commenting.