Hacker News Reader

Comment by davidu | Hacker News Reader

I don't doubt the headline, but "We observe all individuals who began to day trade between 2013 and 2015 in the Brazilian equity futures market" seems like a weird way to come to that conclusion.

View on HN

20,000 people trading a broad based equity futures market over two years isnt a bad study

But yes a study of individual stocks and certain sectors would be better

Or following those individual accounts to see how they grew based on everything it traded would be interesting

Why weird?

Because it would represent an extremely narrow slice of traders. It's possible traders of these equities find it quite hard to make a living but a larger number of other day traders do just fine.

Also it says "impossible" but the article actually shows it's difficult and rare.

The paper's claim is "virtually impossible".

The post's title is click bait.

lots of things are virtually impossible to do for a living and people still manage it though.

Glancing at the abstract it looks like 97% lose money, what larger number in a more suitable sample size would make a difference?

I've heard low single digits of traders and investors deliver Alpha. Impossible? no. Highly efficent? Yes

Because that is not a representative sampling of the day trader population. Assuming the techniques of the study were sound, at best we can conclude whether or not day trading Brazilian futures is a sustainably profitable undertaking.

Wouldn't you need a basis for believing that day trading Brazilian futures has unique properties not present in other markets to assert this? E.g. the volatility or spread is significantly larger than European futures or the information flow is more tightly de/regulated etc.

Is there a basis for believing this?

That belief seems like it would be obviously justified. The same would be true for pretty much any geographically restricted set of day traders on a specific exchange tied to the geography, likely even for huge exchanges. There are huge effects from correlation with that country’s equity markets, currency policy, political events, etc. to such a degree that it would be unreasonable to assume they aren’t present until they are proven present. That’s reversing the burden of proof in a terrible way.

The number of confounders you need to control for before you could confidently not make this assumption is staggering.

There is a distinction between structural confounders and the variables which you are literally backing yourself to predict when you're a day trader. You've mentioned two in the same sentence which I would consider to be polar opposites: political events and currency policy.

For example if market data operated on a significant lag in Brazil for amateur investors, but didn't in the UK, this would represent a structural disadvantage which might make it impossible to beat the market except through dumb luck in Brazil.

It wouldn't mean it's definitely possible to beat the market in the UK, but it would be something one would want to control for in a study which makes a global prediction based on data from a sample in one nation.

However political events, as I understand you to mean, are one of the things most traders are specifically backing themselves to predict in a futures market.

So my question is: what's the basis for believing that there are structural impediments to retail day traders which are present in Brazil but not present in the rest of the world? Everyone in this thread is absolutely certain they exist, but nobody seems to be able to outline what disadvantages are present for a retail Brazilian futures trader which are not present in the UK outside of red herrings like "Brazil is less stable".

You are making a large misunderstanding. The confounders are not about what the day traders themselves are using in their own models. The confounders are things that affect researchers studying whole cohorts of traders.

Day traders may well believe they can predict political events in Brazil effectively. And yet, there may be some quality about Brazilian political events that impacts the cohort of Brazilian day traders systematically differently than say Korean political events affect Korean day traders.

For example, maybe a new cabinet member was just appointed in Brazil who likes to leak new political announcements early to an old business partner with connections at several large banks and trading shops, and so regular day traders are at a significant disadvantage in just this one market.

Or maybe during the time period of the study, there were political protests that disproportionately involved young people, meaning the average age (and perhaps skill level) of the observed cohort of traders was artificially too high just in the window of time of the study. Or any number of possible things like this.

You can’t assume these effects don’t exist when looking at different cohorts that have real reasons why they might not be probabilistically equivalent to a more general population. You have to actually account for the possible sources of confounding (in this case the country of the exchange), either by proposing some prior counterfactual model, or by collecting data across different cohorts and explicitly controlling for the confounder with whatever type of model you are fitting.

To be clear: none of this requires you to know in advance what the confounders actually are, and indeed in most causal inference models you literally cannot know what they are. The best you can hope is that you measures things like, e.g. country of exchange or age of the traders, that correlates with the hidden confounders well enough that they allow you to control for the confounding effects in your model.

It is not a thing in professional statistics to expect researchers to already know the sources of confounding before postulating that there could be sources of confounding that need to be controlled prior to believing the results are generalizeable.

Making an assumption about the population based on a small non-random sample is usually the belief that needs to be justified by evidence.

20,000 people is not a small sample. Sampling bias is still an issue with large populations, but generally significantly less important.

Yeah what you are doing is nowadays known as the "Trump approach".

Asking what the basis is for disbelieving a scientific study with a statistically significant sample size cannot reasonably be described as analogous to anything Donald Trump does.

Kind of not all exchanges are equal, professional investors still don't trust the main Chinese exchange.

As per the top comment at the moment of writing, most of these day traders were likely to be clueless people roped in by scammy "day trading courses".

Yes. Brazil is a relatively high-friction market, and you'd need to know the actual trading costs. The studies I've seen in the US are that most 'active' traders lose money, but 10-30% do well. There is an entire industry of day-trading hedge funds, so many people believe this.

Obligatory joke about the three scientists visiting Scotland in the first time:

Biologist: "Whoa, you see that in the distance? That's amazing! All cows in Scotland are brown!"

Physicist: "Hold on -- all we know at this point is that there's a population of cows in Scotland that's brown."

Mathematician: "No, all we know is, there exists at least one cow in Scotland such that this side is brown."

---

With that said, I agree that it's an excessively narrow sample for the kind of conclusion they're trying to draw.

wouldn't the Physicist first need to assume a perfectly spherical frictionless cow?

It's very informative for Brazilian traders, however ;-)

Isn't it the case for most studies that we read about online? Anything that ranges from "X can cure Cancer" to "Sleeping X hours is better for you" has a sample mostly consisting of American individuals. That doesn't stop them from generalizing nor does it make us question said generalization.

That being said, this seems like a double-standard.

Swing trading with baskets as opposed to daytrading seems to be a better way to go according to historical performance, for example https://www.elastic.co/blog/generating-and-visualizing-alpha...

Anything that worked consistently yesterday will no longer work tomorrow as the HFT's arbitrage that market inefficiency away. That's one of the reasons why even extensive back testing isn't a reliable indicator that a trading strategy will be successful.

That's not so simple because swing trading is not looking for Arbitrage. 90% of traders lose but a good portion of us make money. It is about building a system which works in bull and bear markets.

My firm uses a similar system that hedges long and short baskets and can confirm that hedging is really the only way to swing trade profitably in the long term.

I do well with options trading for a fairly long time now. Anything over 25% a year is good for me but I aim for 100% per year. I am a hit over that this year at this point. Firms have a huge amount of money so they have to use other strategies. For someone with less than $1M to trade with going long or short depending on the market conditions is adequate. Also money management is a large factor.

I assume you trading US markets? The US markets have been very volatile the last 1.5 years which makes for more optimal swing trading. If the market was either going down or up in a linear line, most hedge funds underperform in that scenario.

Not for me. Volatile is bad as I made more money the years before. End of last year sucked and parts of this year had unusual losses. I prefer trending markets.

I only trade US equities and in particular large caps or soon to be large caps.

That document shows that as algorithms might decay, you can swap out the datasets to out-perform. They also conducted a year long study on the results.

HFTs don't arbitrage the single person day trader because those guys are are like a mouse farting in a hurricane. There is no point trying to identify and scalp someone who trades at most some $ millions a day, unless the person is doing it on really swampy stocks with no liquidity. Even then, the capacity in those markets for HFTs is also extremely limited.

HFTs will target big insitutional traders who have to shift a billion dollars at rebalancing time in a single day.

Even that is a dying business today because they themselves have been arbitraged away to a degree.

Reply on news.ycombinator.com