Saturday, September 19, 2015

Peer to peer lending strategy using the Kelly Criterion.

There is very little literature on how to invest in peer-to-peer lending campaigns and I am pretty sure that very few financial bloggers who have a decent framework on peer-to-peer lending.

I am going to attempt to come up with a blueprint on how to size your bets when you are being offered a campaign on any lending platform.

This approach is based on the Kelly formula which is employed by gamblers in casinos. It is an optimization strategy which maximises long term returns. Based on what I know about the current state of lending platforms, the Moolahsense platform has enough data to facilitate this form of bet sizing.

I would leave the mathematical proof of this approach to the experts.

[ For the purposes of this article, this article only applies to lending projects and does not apply to equity crowdfunding campaigns. As I have a fairly personal bias against property crowdfunding, I do not advise that this be applied to property projects at all but some folks ]

Let's say you have a campaign and have allocated $10,000 into peer to peer lending which forms your bank-roll.

Scenario 1 : Attractive Campaign

This sample company will pay-back a total of $1060 in one year. Suppose you check the company data and find that the calculated default probability is 2.5% over a year.

The Kelly criterion which recommends that you bet proportion x of your total bank roll where

x = ( Expected net gains ) / ( Gains on successful campaign )

Start with an assumption that your position is $1000.

You can earn at $60 on a successful campaign. When the company defaults, you expect to lose everything or all of $1,000.

Expected net gains
= Earnings x Probability of success
= [Earnings x Probability of Success - Loss x Probability of Failure]
=  $60 ( 100% - 2.5% ) - $1000 x (2.5% )
= $33.50

Gains on a successful campaign
= Best case scenario
= $60

x = ( Expected net gains ) / ( Gains on successful campaign )
   = $33.50 / $60 = 55.83%

In the above example, the recommendation would be to bet about $5,000 or $6,000 into this campaign as it is fairly attractive.

Scenario 2 : Campaigns to avoid

Let's consider a campaign which returns $1,090 in six months but the website says that there is more than a 16% chance of default within a year which you estimate should turn out to be about 20% a year.

Return is $90 on a successful campaign. You lose everything in the event of a default.

As it is a six month campaign, you should be using a default rate is (100%-20%) ^ (6 months / 12 months) or 10.56% or just about 10%.

Expected net gains
= Earnings x Probability of success
= [Earnings x Probability of Success - Loss x Probability of Failure]
= $90 x 90% - $1,000 x 10%
= -$19

Once you get a negative number, you should avoid this campaign and look for something else to do with your money.

Scenario 3 : Risky campaign where you can ask for more

If you can offer more for Scenario 2, what happens ?

Suppose you can choose to offer a high rate which returns $1,120 in six months.

Return is $120 and you lose everything in the event of a default.

As it is a six month campaign, you should still be using a default rate is 1- [(100%-20%) ^ (6 months / 12 months)] or 10.56% or just about 10%.

Expected net gains
= Earnings x Probability of success
= [Earnings x Probability of Success - Loss x Probability of Failure]
= $120 x 90% - 1000 x 10%
= $8

x = $8 / $90 = 8.88% which you can round up to 10%

For this campaign, you should bet a smaller amount - no more than 10% of your total bankroll.

You should therefore bet the minimum of $1,000 of your bankroll is $10,000.

This framework should be a superior but riskier approach to something which I am currently doing, which is to bet the minimum amount of $1,000 across as many campaigns which I can get my hands on.

If you wish to follow my strategy which maximises diversification, you should still employ the Kelly formula to find out which campaigns to avoid. At the very least, you should choose the minimum amount of return such that your expected net gains are a positive number.

I think this is cutting edge stuff, so comments from seasoned traders are welcome !



7 comments:

Passive Income Builder said...

sibeh chim la!

John Smith said...

Does the Kelly formula require N to be large?

Christopher Ng Wai Chung said...

Good point. Kelly wins as N tend towards infinity.

N should be fairly large if you consider that we are at the beginning of the rise of peer to peer lending websites as an alternative to banks for SMEs.

We are looking at about 2-3 campaigns a month and the momentum is likely to increase.



John Smith said...

If we take N to be 50 then we would need about 50K and 2 years.

Would be interesting if you could write up something with your knowledge of the law to turn this into a collective investment scheme with you as the manager. You could be a venture capitalist.

Christopher Ng Wai Chung said...

Maybe less if more platforms starts to open up. Seems that SMEs really need the loans.

I doubt that my knowledge of the law in any state to do this for quite a while.

By then N > 200 for me, haha !

MoolahSense said...

Hi Christopher,

This is such a very well-thought of blog post and the use of financial modelling to calculate risk and return is definitely useful.

Debt-based crowdfunding, similar to all forms of investment, carry risks. Therefore, your article helps to educate investors, esp. those who are new to crowdlending/crowdfinancing.

MoolahSense also aims to educate its investor community on diversifying their portfolio and spreading their investment risks via info sessions and other educational materials.

To find out more about Crowd Investing on MoolahSense, logon to www.moolahsense.com.

Market Today Research said...

Nice Post! LoanMeet is a Peer To Peer P2P Lending marketplace that brings borrowers and lenders together on the common platform to get Personal and Business Loans.