They Still Haven't Told You

But that's illegal. It's market making.

It is not ridiculous if the data is longitudinally consistent.

You clearly didn't read the abstract. There's obviously a conspiracy and if any of this is news to you, it is because the people you trust to alert you to such problems still haven’t told you.

Only Bruce knows the truth!!!

/s

This would make tons of sense with regard to actual HFT firms. If you're whole schtick is doing stuff over timescales that are (far) shorter than a minute, being locked into a position for hours is risk you really don't want to take.

Yeah this is so obvious I would hope it could be controlled for by the Analyst. There are simply a lot of funds and traders who, by policy, do not hold positions overnight.

Shades of Slartibartfast: "... Anyway, the recession came, so we decided to sleep through it."

No, since expected intraday gains are negative, this would be a losing strategy. Closing price seems to be lower than opening price on average.

The suggested explanation is: buy orders are placed before open, raising opening price. The gains on the holdings are then larger than the cost of buying in the opening, and selling during the day.

but this article is explaining that unless you are selling short in the morning and covering at night, there are mostly negative returns for intraday trading, i.e. "pil(ing) in in the morning, and exit(ing) in the afternoon".

There are lots of firms and funds and floors that never hold overnight, but this research demonstrates that that is basically a statistically losing strategy. If you follow markets it's almost impossible to not notice that almost all the real action happens after hours, and the day's trading tends to "erase" whatever happened overnight. Bruce's research is hard to contradict, unless the data is wrong or his formulae or off.

"They still haven't told you and I'm not going to either" should have been the correct name for the paper.

I don't think there is anything new here. The overnight trading anomaly has been observed for years

The author explains what he thinks is happening on page two of this paper (which he cites in the OP, but doesn't actually explain): https://arxiv.org/pdf/1912.01708.pdf

TL;DR: Stock prices in the US go up overnight and come down during the trading session. The only explanation must be that some shadowy trading firm with a lot of money is buying a bunch of $stock in the morning and selling it back later in the day (at a loss), because it inflates the value of their much larger stockpile of $stock that they hold onto overnight.

I thought anyone can publish to arxiv?

> between 0930h and 1600h there's liquidity to buy/sell your position at any time, for the prevailing price. Markets are open only 7h of the day but 24h worth of events takes place each day.

Yes indeed. The author claims that there is less risk in overnight positions than in intra-day positions. I think there's more. Firstly more time passes overnight and secondly the lack of liquidity means you can't unwind overnight positions if you need to.

But spreads aren't larger in the morning are they? If anything they are smaller, due to the action of the opening auction. Without actual data to back up that assertion, the rest of it doesn't really need reading.

Assuming his assertion is correct, the reverse is also a strategy. Selling in the morning, causing prices to drop, then buying back when they are cheap in the afternoon, so ending flat but having made money.

So anyone that buys and then sells or sells and then buys makes money. This is easy! What could possibly go wrong?

My knowledge of this is maybe limited, i've not worked as a quant, but i've stared at a fair bit of market data having written feed handlers for a fund.

Tangential question: why isn't the stock market open 24/7/365 (minus periodic maintenance of the computer systems)?

I am surprised that by now, decades later, no one has the goods on Renaissance . What is to stop someone who works there or former employee from uploading to the dark web the "Renaissance strategy", for a price tag of $10-100 million btc, monero or something. Who would know. Although no one would beleive him.

No.

He is saying that some firm own lots of $stock that they buy-and-hold. They then buy smaller amount of $stock early in morning to cause a swing up in price. Over course of day, the ability to influence price declines, so they can sell the amount they just bought without influencing price as much. Sell for profit or loss, doesn't matter.

The root goal (how they make money) is that they should have influenced the price enough that the large buy-and-hold stock they own has increased in value. Specifically, they want the "overnight" price change to be more positive than the decline across the day (again, by pumping up the morning price). They don't have to buy/sell, its more the value of their holdings are higher.

People have certainly tried, but I think that these comparisons don't make any sense. What makes the "waste" of crypto stand out is a qualitative concern:

HFT uses energy as a means to an end (i.e. computation), while proof-of-work mining has an incentive structure that directly rewards energy expenditure. In other words, one is energy- (and hardware-)bound, the other isn't.

As a thought experiment: If fusion energy and self-replicating nanobots were to become viable tomorrow, how would that impact HFTs and proof-of-work mining, respectively?

HFT firms use a tiny fraction of the energy of proof of work crypto schemes. Even if you take into account energy requirements they used to build their high speed networks.

The main reason being is that they are all co-located with the exchange, and the byzantine fault tolerance is completely side stepped by the exchange having a single point of serialization between the HFT participants and the matching engine. And then just ensuring everyone has the same length cables to that single point of serialization.

So the policy of everyone connects through this single point, solves the fairness problem.

And then the fault tolerance is often just solved by having an active standby architecture, so if the primary matching engine goes down, you fail over to the backup.

You would be surprised at the low number of servers actually in the primary path for financial exchanges.

Well for the actual trading they will be operating out of the same datacentres as the exchanges which are not particularly massive and will have relatively small limits on the heat (and therefore power). There will be other computers in other bigger cheaper datacentres for analysis and suchlike. And office buildings of course. But not really comparable to e.g. the energy usage of Argentina. I would guess the total is less than one of the massive internet companies like Google or fb but that might be a bit low.

Not sure, but I understand crypto uses > 0.5% of global electricity production. So you can pull up any chart that accounts for the 99.5% of top electricity users and verify HFT is not one of them if that's what you're seeking to show.

No, I don't think that can be done. No insult to the layman, whatsoever.

it reads like the sort of thing they would publish