4 ms·
I worked at a prediction market startup for a while. Even completely rational prediction markets generally have a structural problem with events with likelihoo
by enjeyw 3y ago
I worked at a prediction market startup for a while.
Even completely rational prediction markets generally have a structural problem with events with likelihoods close to 1.0 or 0.0. This leads to cases where conspiracy type events have much higher likelihoods than they should, leading people to assume the market has lost all rationality when actually it hasn't.
For example, imagine if there's an event with a predicted likelihood of 1 in 50, but it should actually be 1 in 1000.
Assuming the market uses real money, then on paper someone should be able to make a roughly 2% return by betting that the event won't occur. However, if this market isn't closing for a while (for example for an election at the end of the year), then no-one is incentivised to take the bet and correct the market, because they'll get a much better risk free return just by putting their money in a term deposit and collecting the interesting. And so the market remains un-corrected. Basically prediction markets have an accuracy upper bound determined by the current risk free return.
For markets using "play" money a slightly different effect is at play. When it's not real money involved, generally the people are incentivised to try and top a leaderboard of some sort. In this case, a really effective strategy for people who currently aren't near the top of the leaderboard is to take long shot bets. If they loose, who cares they weren't gonna win anyways, but if the the bet plays off they have a lot to gain. Basically there's an asymmetry in returns that once again messes with probabilities.
- xur17 3y ago> Assuming the market uses real money, then on paper someone should be able to make a roughly 2% return by betting that the event won't occur. However, if this market isn't closing for a while (for example for an election at the end of the year), then no-one is incentivised to take the bet and correct the market, because they'll get a much better risk free return just by putting their money in a term deposit and collecting the interesting. And so the market remains un-corrected. Basically prediction markets have an accuracy upper bound determined by the current risk free return. What if the deposited USD automatically earned interest at the current risk free return (think something like USDC on Aave in crypto land)?
- enjeyw 3y agoYeah as far as I’m aware that’s the solution. Though then the market can’t use the deposit to make money for themselves, and needs to rely on other mechanisms like trading fees etc. So really it comes down to how much the market cares about the accuracy of long tail events.
- zhivota 3y agoIt's more than the risk free return. It's also fees. Whatever the fee percentage is, all in, to get a dollar out of the platform and back into cash on a winning bet, plus the risk free rate over the time period, that's your upper bound. Given the comments above about fees it doesn't surprise me at all to see 5-10% mispricing.
- tqi 3y ago> Even completely rational prediction markets generally have a structural problem with events with likelihoods close to 1.0 or 0.0. This leads to cases where conspiracy type events have much higher likelihoods than they should, leading people to assume the market has lost all rationality when actually it hasn't. Is there academic research on this by any chance? Recently I've been wondering if there is a similar issue with VC funded of startups, where people overestimate the odds of a building 10B or 100B business (especially relative to a 1B or 100M business), causing them to pursue investments and strategies with lower expected value.
- alephnerd 3y ago> Is there academic research on this by any chance Read "Secrets of Sand Hill Road". It's written by Scott Kupor - one of the Managing Partner at A16Z - and explains in depth how the VC model works. It's also required reading if you want to interview at a VC fund, and highly recommended by other founders to learn how to better strategize when raising funds. Secrets of Sand Hill Road, Monetizing Innovation, and Working Backwards are the only books you need to successfully understand how to raise and build a VC funded startup. The pricing strategy guide by I think Sequoia is also useful.
- hipadev23 3y agoWorking Backwards, the Amazon book, or something else?
- alephnerd 3y agoYep. The Amazon book. The YC App is basically a modified PRFAQ. There's a reason why the PM-to-Founder pipeline exists.
- tqi 3y agoOh nice, thanks for the tips. I'm actually asking out of curiosity, born of working in and around SV companies over the last 10+ years, rather than anything practical.