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.