Diversification vs All-In
My record on PredictIt over the past two years has been pretty solid (30%+ returns over two years, it’s hard to calculate exactly because I’ve added money over time) and I’m going to start adding $100/month because if you’re good at something, you should do more of it. This has lead to a breakdown of my previous method, which was extreme diversification, and I’m trying to figure out how to respond because I’m not comfortable with going all in on a few markets.
My previous method was basically:
Step 1: Find a market that didn’t make sense or seemed grossly mis-valued.
Step 2: Invest $20.
Step 3: If I still have money, find a new mis-valued market.
Unfortunately, I think I’ve found all the crazy markets. I’m currently invested in 14 of PredictIt’s 228 markets, or ~5%. And the proportion of markets on PredictIt that are both grossly mis-valued and that I have any reasonable knowledge of (for example, the next European leader to lose power might be a mis-valued market, how would I know?) is probably about…5%. But if I’ve got an extra $50 to invest, do I spread that money out over all 14 markets or should I focus on a few?
Here’s an example:
I just picked up the ‘Clean sweep for Dems in 2020’ and I’m trying to liquidate it for more shares in the ‘Will Biden drop out before 11/1’ market. I think both of these are profitable markets, I just want more in Biden. But this makes me uncomfortable because I’m decreasing my diversification, which increases my risk of a blowout.
Let me be clear, I think both of these markets are profitable (see footnote 1), I just think that the Biden one is more profitable and I’m not sure how to distribute the money between them. Should I got 50%-50%, 70%-30%, 100%-0%? I’ve got fourteen markets, should my new formula be to go 30% in my best market and even split among the remaining 13 markets? If I go all in on Biden not dropping out and he has a heart attack on October 1st, I could get wiped out, so I definitely want some diversification, but how much?
Or to phrase another way, if you know nothing about investing, you buy an index fund. If you’re the best investor in the world, you’re Warren Buffet and 100% in Berkshire Hathaway. But what if you know a little bit? How do you modulate between these extremes? If you’re 80% sure you’ve got a winner, how much of your investment portfolio should you put in it? Should I be making more active investment decisions? Usually, when people invest on their own, they buy radical stuff with high upside, like Tesla, but Charlie Munger is heavily into Costco and that business is going to be around forever. Why not shift some money into a Costco or a Microsoft, something that’s going to be around a long time and be profitable?
If I have $20 in 14 markets and the Biden market is the best, it feels foolish to both put everything in the Biden market and put the exact same amount in the Biden market as in all the others but I have no clear guidelines on where in that grey middle to land. Should I be $70 in Biden and $15 in everything else? $210 in Biden and $5 in everything else? How would I rationally make this decision and should it inform my investment decisions?
The Biden market is the kind of market I always invest in, one where I have no idea why it exists. Biden’s already won the nomination and it would be extremely costly for the entire Democratic party if he dropped out now. I would put the odds at 1-100,000. I have no idea why they’re currently at more than 1-10 and I’m guessing it’s some media conspiracy theory. These are the markets I try to invest it. Sometimes they blow up in my face (for example, I put money on Steve King winning reelection because why would you bet against an incumbent?) but in general they pay out.
The Clean Sweep market is just a Bayesian bet, ie people are generally bad at predicting the odds of multiple events all occurring. For example, if you think there’s a 70% chance of Dems taking the Presidency, a 70% chance of them taking the Senate, and a 95% chance of them keeping the house, that’s a 46% of them all occurring. And yeah, these all covariate with each other, but to what extent is a big question. Basically, if I see a market structured around a common fallacy, I gamble that most people will make that fallacy.