That’s not selection bias, that’s a functioning marketplace. Something is worth what someone will pay for it. If demand is increasing, prices increase.
The point that the op is making is that this is only looking at the "successful" trades. It is not looking at the set of "attempted" trades. So it is only looking a selection of the possible trades, and so there is a selection bias involved in the data set. This is very much a form of survivor bias.
It is likely that it also "only" includes those that managed to buy a house at under market rates, and sold at, or above, market rates within 90 days. Anyone who is buying to live in is likely paying as much or more as the "sale" price.
This is further stated in the market, since the annualised rate of return is 650%, but house prices are _not_ going up that quickly.
Which, in turn, means that distressed buyers are not necessarily going to see a 15% increase in the price of their house in 90 days, nor are they going to see a ~30% in 180 days, since their property may not be "undervalued".
There is clearly selection bias in the quoted figure of houses sold within 90 days, because people try to maximise what they'll get.
People only tend to drop the price after it's been sat around for a while. By definition, such sales are not very likely to happen within the 90 day window and therefore do not contribute to this average.
There's selection bias at any time scale. Those who manage to sell their property at all are getting better prices than those who fail to do so.
The point is that you can't calculate a probability of making +15% within 90 days just by looking at how many of the houses sold within 90 days were sold for +15%. Or any +X%. Or any time period.
If you really want to calculate such a probability, you need to buy a bunch of houses at random, then try to sell them, and then see how much you made and how long it took. The data used in the post is not suitable.