Filter investing is great, but it also takes time and has some downsides. It used to be that simply filtering out loans that didn’t have 0 inquiries in the last 6 months yielded much higher returns. This still may be true to an extent, but filtering on any one criteria is like chasing a moving target. Lending Club and Prosper have the data just like we do and they do their best to set rates appropriately. There is no doubt that they will continue to do better over time. So what if you could invest based on machine learning which adjusts over time? Enter PeerToPeerQuant.
PeerToPeerQuant uses what is called a genetic algorithm. They define this on their website:
A genetic algorithm tries to solve a problem (who pays the most on a loan) by mimicking techniques found in nature like mutation, selection and inheritance.
I think their example describes it best:
We feed a computer program all historic loan data and create a population of initial investment strategies like:
- Invest in loans where borrowers have more than $100,000 in income and own their home
- Invest in loans smaller than $10,000
- Invest in loans where borrower’s FICO score is below 700
The program evaluates each strategy against all historic loans to determine which strategies have the highest total return. The best strategies become the “breeding population” for the next iteration.
When two strategies “breed”, attributes from each strategy are combined. If the second and third strategies (above) were to breed, we might get a strategy like “Invest in loans smaller than $10,000 where the borrower’s FICO score is below 700”. A random mutation might change that 700 FICO threshold to 750.
This process is repeated thousands of times until an optimal solution is found.
Their site is similar to what P2P-Picks was before the models were automated in both BlueVestment and Nickel Steamroller. They offer a list of notes that they recommend based on their algorithm. The loans are listed in a random order as there isn’t a ranking system. All loans that meet that criteria are ranked the same. Any additional drilling into the criteria has already been tried with the algorithm and hasn’t yield better results.
For example, the criteria might be income over $100K and owns their own home. The algorithm would have tried other criteria in addition like only “debt consolidation”. If it didn’t improve performance it isn’t included in the criteria so we do not recommend users do the same.
You use their service by clicking an ‘invest’ button which will take you directly to Lending Club’s site and add it to your cart. The invest page looks like this:
The pricing is as follows:
Free – Seeing note recommendations is always free
Free – 20 free INVEST clicks for new accounts
Free – 5 free INVEST clicks every month
$0.15 – Each INVEST click after free ones are a flat $0.15
Keep in mind that whether or not the note gets issued or eventually defaults – you are still charged the initial $0.15. They have factored this into their pricing to accommodate for this.
Currently, the system updates every 5 minutes. They can adjust this if necessary, but so far there have been notes available every day and time they have checked. Since most investors and automated tools attempt to purchase notes right at release time, I asked the folks at PeerToPeerQuant whether or not logging on at release time would potentially increase returns or give access to better notes. They stated:
So far speed does not improve the number of loans available that meet our criteria. If this changes, we are committed to doing the work needed to speed it up.
PeerToPeerQuant also created a handy tool called the “Lending Club Percentile Calculator”. With this tool, you can compare your results to other Lending Club investors by providing your returns, weighted interest rate and average age. Although their account average age is still young utilizing their model, they are in the 98 percentile vs all accounts. I look forward to seeing how their accounts track over time.
So how can you get started? Check them out at http://www.peertopeerquant.com and click the Login button on top right hand side. (Login currently requires a Google account)
The folks at PeerToPeerQuant would be happy to answer any questions in the comments section. I’d also be interested to hear your results of the Lending Club Percentile Calculator.