[unholiness]

Betting on PredictIt has been a side-hobby of mine for the last 5 years, in which time I turned my fun $10 deposit into (just) over $1000. For me, it's a satisfying way of interacting with the news, motivating clear-headedness and accuracy over simplicity and filter bubbles.

The talk about "rational markets" in this thread is well-meaning, but I think it could better targeted toward BetFair and other uncapped foreign markets. PredictIt has a cap of $850 from any individual in any market, which means almost no market is dominated by "sharps" exploiting differences between prices and reality. Sure, PredictIt's prices are often more accurate than the average person's guess, but they still exhibit a lot of small and predictable biases:

- "Yes" positions are more popular

- Pro-Republican positions are more popular

- Cheap, improbable positions are more popular

- Positions confirming simple ideas are more popular

- Positions traders wish were true are more popular

- Things being discussed on the news are treated as more contentious than they are

Ultimately, almost all of my 100x growth just boiled down to finding these biases and maximizing my expected log(return) with them in mind.

I do believe PredictIt was good for discourse. Though it may have ultimately affected a small number of views, the "put-up-or-shut-up" mentality is much closer to the scientific method than media and its tendency to navel-gaze. The domain of PredictIt markets was small, but I often imagine a world which creates this kind of ecosystem for a broader array of scientific fields.

For me, though, this is the end of the line. The 2020 races (with Georgia runoffs) risk being unresolved by the 2/15 deadline, and it's impossible to predict what PredictIt will do then, much less if markets will properly price that in. It's been a fun run, and I'm glad it lasted as long as it did.

[ummonk]

The fees that PredictIt charges also make the market highly mispriced. E.g. you can't arbitrage two markets on PredictIt because PredictIt will take a cut of any profits you make in a particular market, without deducting any losses in other markets.

[unholiness]

This is true about arbitraging between markets on PredictIt, but I do think the impact of fees on market prices (and on ability to make money on PredictIt) was usually overstated.

PredictIt has two fees: A 10% fee on winnings, and a 5% fee on withdrawals.

The 10% fee on winnings is only on winnings. So when buying a share for 90¢, your $1 won will earn 99¢ and pay 1¢ in fees.

This means in a market with multiple candidates, betting "No" on all of N candidates will always earn you $(N - 1) without fees. That's is a winning bet iff all the "No" shares add up to $(N - 1.1111). In practice, the bias toward "Yes" shares alone is strong enough that buying all "No" shares, even at current asking prices, usually cost between $1.07 and $1.10 less than $N.

So, if you're only buying "No" shares in markets with multiple outcomes, the prices are set close to a point that entirely negates that 10% fee.

And of course the 5% fee on withdrawals is only on withdrawals. If you send the money straight back into another trade over and over, the fee is diluted enough to be irrelevant to any individual trading decisions.

[LittlePeter]

I don't understand where N-1.1111 comes from. Isn't it enough for all the shares to add up to less than N-1 for there to be a profitable bet?

If all no shares add up to S < N-1, then gross winnings will be N-1-S > 0, which when multiplied by 0.9 to get net winnings will still be positive.

[unholiness]

From the fees. 90% of a $1.1111 profit is $1.

Notably, you don't need to post $N to make this bet. PredictIt calculates your worst case scenario and holds only that money. A set of "No" shares worth N-1.12, if it existed, would be literally free to hold. Fees are included in that calculation, but also fees are based on the difference between that cost to hold and the return. So yes, profits mean different things in different parts of that calculation in this degenerate case.

A full set of Nos worth N-1 would be a guaranteed loss of $0.10 per share from fees. PredictIt would hold that 10¢ and you would never see it again.

[renewiltord]

Same, though I went $30 to $180 or some such shit and I didn't keep playing because it locks up the money. Concur with everything you're saying. On PredictIt the black swan was over-weighted.

You can take the suckers' money, but you end up still being a sucker because your money is locked up.

What I find amusing is that they have some Plaid-type integration for deposits but withdrawal still needs this routing number / account number shit.

[VoidWhisperer]

Deposits are likely more subject to KYC and things like being able to verify that the bank account exists - I imagine, atleast at the surface this might be why you see this

[shusaku]

Interesting. I always assumed that republican positions were favored because liberals were hedging their anxieties.