Hacker News new | past | comments | ask | show | jobs | submit login

Putting aside the potential quality issues of the study, the headline here in HN isn't accurate and made a lot of people comment about how wrong it is.

The study doesn't literally say "impossible" but "virtually impossible" and they do say that a very small percentage was able to make about $310 a day which is about ~110k a year.




0.0051% of the study group, to be precise.

The more interesting trend, which isn't covered in the abstract, is that profitability is inversely correlated with how long you stay at it. It was 30% of people who day traded for one day and gave up on it, and steadily dropped down to 3% for people who kept at it for more than 300 days.

As a stats person, that struck me as a very evocative distribution. The paper put it nicely: "This peculiar pattern is similar to what we would fi􏰄nd, for instance, in the casino roulette."

The next thing I would want to know is, if you model the stock market as a machine that simply gives random payouts drawn from some distribution, what is the probability that a set of 1,551 groups of 300 random draws from that distribution would contain at least one set that averages more than $310? If the paper ran such a model, I didn't see it.


> if you model the stock market as a machine that simply gives random payouts drawn from some distribution, what is the probability that a set of 1,551 groups of 300 random draws from that distribution would contain at least one set that averages more than $310? If the paper ran such a model, I didn't see it.

The abstract says the trader who made $310/day had a standard deviation of $2560 (presumably per day). Over 300 samples, that's a standard deviation of 2560/sqrt(300) or $150/day. So if he really had a breakeven strategy and got lucky, he was 2 standard deviations from the mean, or in the top 2.5%, and you'd expect 40 other traders to have done just as well, and it would be vanishingly unlikely that nobody else did.

Not quite what you were asking, but I don't think there's enough information at least in the abstract to answer your question. The payout distribution will be largely dependent on the size of the traders' positions and riskiness of their strategies.


Yes, it certainly looks like what would be dictated by statistics with random chance that eventually reverts to a slightly negative mean.

In the short term it's not that unusual to flip a coin and get 5 heads in a row, but the more you throw the closer to a 50-50 split you'll see.


I'd also be very interested to see what would happened, hypothetically, if everyone just stopped trading after some fixed number of hours, held that position until the point when they decided to give up trading, and then liquidated it.

My hypothesis is that their average returns would end up being much higher than the returns they realized by day trading that whole time, due to less money being lost to fees.


ICO bagholders disagree


> which is about ~110k a year

$310 * 5 days a week * 52 weeks a year = $80,600


110k pre tax in the USA is 80k post tax per year-- it is often helpful for us to think in terms of both numbers when comparing money making methods. That's my guess at least.


Probably more that they didn't account for weekends, $310 x 365 is $113k


They are also trading on a small account most likely. If you have $500k to trade with making $300 a day is much easier.


Yeah you’re right, I just multiplied by 7 days. Then again you don’t work 5 days every single week (depends of the location) but 80k is definitely closer to the actual figure.


If for each of those people, there are dozens of long tail traders who lost their shirt and tapped out after a few thousand, the point stands. It could be just random chance.




Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact

Search: