A quantum walk down Wall Street
economist.com
economist.com
They are fiercely and aggressively secretive, so no one really know for sure.
However, their founder worked in that field before moving to business: https://en.wikipedia.org/wiki/Chern%E2%80%93Simons_theory
Famously, he was sacked by the NSA for giving an interview opposing US involvement in the Vietnam war.
lolol. i will bet all of my 401k that they absolutely do not.
i've interviewed at a couple of hedge funds and have lots of friends at various prop shops/hft firms/etc. most of them are using linear models. the ones that have market maker businesses deploy those linear models to fpgas.
people act like these places are spooky magic cauldrons of physics + math + computer science. they're not (but it's definitely in their interest to promote such rumors). what they do have is very very robust data collection/aggregation infrastructure and backtesting systems. the rest is just correctness of execution of strats.
RenTech is especially good at acting like they're from another planet, which I'm sure helps them attract those stellar mathematicians.
The claim, the secondary (tertiary?) source, and the primary source make that all hard to not smile at.
66% annualized returns are more likely to be pyramid schemes - but that couldn't happen, could it?
This isn’t true once you include equity options, which have several orders of magnitude more data.
However your general point still stands, because most options trading strategies don’t need such extreme granularity of data. Much of it can be ignored, or close to it.
However I think you're underestimating RenTech here. They're genuinely just in another league compared to what most people consider the "elite" quant shops. You're not getting in unless you're an actually impressive academic with a track-record, so I wouldn't be surprised if they're trying some weird stuff that other shops can't even understand (although I would imagine in small size vs. more vanilla stuff).
IMO places like JS/CitSec do a really good job of bombarding campuses to inculcate this idea they're the absolute apex of mathematical wizardry, but the places that are really doing some dark magic shit aren't trying to get undergrads to apply to them. Ofc maybe I'm falling for the same kind of propaganda for RenTech
Just from snippets I've sort of heard/read about, I think they were one of the earlier ones to move into HFT (although maybe not the super fast infrastructure heavy type these days). In some interview Simmons said they realized returns became more predictable they shorter the time frame they looked at and they pushed it to the extreme. I think there is also reason to suspect that they may have adopted some NLP strategies early on as well since Mercer was involved in that or something, and they initially hired a bulk of their team from IBMs NLP research team. Also they did not dodge 2008 completely, in some interview they said they lost close to / more than half their portfolio value in the market crash, but because they didn't have stupid leverage or outside investors or something like that, and also because they trusted their models, they didn't sell and held on. So maybe just slightly better execution but mostly the same.
Anyways, I was reading about these guys back in 09 when quant trading wasn't so blown up. Now every kid whose decent at math seems to want to be the next renaissance, which just makes me feel like the best years for that are over.
As far as I know the NLP stuff is more to do with similar techniques being applied to market data, rather than actual speech recognition or whatever. Hidden Markov models and the like.
That is one example.
If you hire very intelligent people to do just that, you are doing it right. But the info is out there for everybody to see. They just arrive earlier than others. How? That is the secret.
People said the same thing about LTCM. They had multiple Nobel Prize winners on staff, literally the people who wrote the Economics and Finance textbooks.
Then a decade later, history repeated with all of the prop trading outfits doing securitization.
It's always the same ingredients: 1) a theoretically sound strategy for taking advantage of some arbitrage opportunity 2) the assumption that positions can actually be liquidated on demand at the prevailing price and 3) enormous amounts of leverage. Then something happens that wasn't accounted for (e.g., sovereign default, counterparty default, etc.) and suddenly the strategy is no longer sound ("in a crisis all correlations go to one"), the liquidity assumption is no longer true ("where are all the buyers?"), and the leverage puts you out of business ("the market can stay irrational longer than you can stay solvent"). People never learn.
Sure, you've got the quintessential marketing induced ML overlay that many firms do, but in all cases I've seen so far it's completely defanged and really there only so that it can serve as a marketing move.
At lower frequencies when the data gets thin and noise/overfitting is a major problem, yeah it makes sense to use simpler models. Bias/variance tradeoff in action.
On niche markets with low liquidity, one doesn't have such tight latency envelopes but those markets also offer more opportunities in general so again there's no real justification to use fancy ML models or GPUs in general.
For me the "spooky magic" is more the ability to pull this off in size.
Edit: I do have friends on the buy side that use pretty fancy models, but those places are not in the HFT/market making game. But using those models does not take away from needing to be able to translate the edge consistently and efficiently.
Rumor is they use magic.
I suppose he could’ve been lying, but the believable accounts suggest that the majority of their magic is simply having smart people use simple tools very, very well.
E.g knowing what's causing missing values in your data and what implications various fixes on that might have on bias in your linear regressor is probably way way more valuable than fitting some shiny non-linear toy
Transcript with time stamps by @clausok: https://news.ycombinator.com/item?id=19065226
One thing I tell new quants that I hire is that your job is kind of like a magic trick. To an outside observer the trick looks almost supernatural, but once you understand the trick you realize it's so unbelievably simple and straight forward you wonder how it is you were ever fooled by it.
That said, even if someone tells you how a magic trick is performed and then hands you everything you need to perform it, you are almost certainly going to mess it up. It takes a lot of practice, skill and discipline to pull off even a simple magic trick even after you fully understand it. All of the difficulty of a magic trick is in the execution, not in the idea.
It's the same with quantitative finance, the techniques do not involve anything remotely complex like quantum mechanics and in fact I am almost certain to reject strategies that are overly complex... but even if I revealed how our simplest of trading algorithms work, it's still incredibly difficult to actually execute them. Taking an idea and translating it into a high performance algorithm that is bug free, dealing with networking, collecting and dealing with petabytes of data, having a tight iteration loop, risk management, and most of all, having the creativity to identify something simple among a sea of complexity, those are what make my firm and other firms successful.
It's unbelievably difficult work and most quants do end up failing, but it's not difficult in the way that many people think it is. If anything, most quants I've worked with fail by overcomplicating things and not being able to work from the ground up, off of first principles.
If you want to know what it's really like to be a quant, review stuff from the ARPM; everyone I hire goes through their 6 day bootcamp but they have other materials as well:
And as for books, Algorithmic and High Frequency Trading covers the foundations:
https://www.amazon.com/Algorithmic-High-Frequency-Trading-%C...
Those are pretty good sources to get an overview of what actual quants at successful firms know.
Quantitative trading is technical, I don't want to give the impression that it's not technical... but it's not "fancy" technical. It's more along the lines of rigorous and iron clad instead of flashy and sophisticated. Every strategy is built up step by meticulous step in precise detail and every step needs to be rigorously justified and experimentally verified.
Generally the thought process starts from the assumption that there is no money to be made on the stock market, either due to perfect efficiency or things like fees eating up any potential profits... when we talk about models, the models we construct describe how the stock market would behave if it were perfectly efficient, ie. free of any arbitrage opportunity.
Then given our model of a perfectly efficient stock market, we simulate what we should expect to observe in such a perfectly efficient market... we then investigate empirically whether these observations happen in reality. Is the market genuinely efficient all day every day across every security.
For some phenomenon it really is, but sometimes the market deviates from the model, so our model is either incorrect or an arbitrage opportunity has presented itself. If an arbitrage opportunity presents itself, we investigate how feasible it is to capture it, things like engineering effort, risk factors, profitability etc...
If all of that works out, then we get to work constructing a state machine for an algorithm to capture that opportunity. We implement the state machine, write tests for it, run it through our backtester, then run it through our live simulator, and after everything checks out we deploy it live.
Every algorithm is treated like a person, it's given its own human-like name, has its own account, its own set of permissions, capital allocated to it, risk profile, and algorithms are evaluated on a daily basis to reallocate capital to them and modify their risk profile.
Not that I know, how they actually do it.
We do prevent this by using a fairly standard component called an internalizer. Almost all of our orders go through such an internalizer and if it sees that two orders are matching one another, the internalizer will perform that match directly instead of having it go to the market.
This happens quite frequently in fact, and it's good to be able to catch it beforehand, avoids paying fees, having any kind of market impact, as well as obviously complying with regulations.
Our interview process involves data structures, algorithms, probability and statistics.
For example if I asked you to write an algorithm to generate 4 random uniformly distributed positive integers (that means excluding 0) that sum up to 100, could you do it? That question or a similar one like it will eliminate about 80% of applicants, including those with PhDs, who are unable to produce an even remotely viable solution even after considerable assistance is given. Then you have the remainder who can produce a competent solution but it's not exactly uniform, there's some small bias in the answer but it's good enough and I can usually walk them through it to work out the kinks, and then maybe 5% or less are able to work out a perfect solution.
Some other questions you should be able to handle confidently would be... how many times should I expect to flip a coin before I see heads 3 times in a row?
More challenging questions will be about Markov chains, optimal stopping, random walks. All questions can be solved either by writing an algorithm, or providing a mathematical solution.
Then just go over your resume and ask you technical questions about it... usually I will find something you worked on, you will describe it casually and then I will ask a technical question about some specific area you bring up that you seem interested in.
Are these questions asked on a computer? As in do you allow the applicants to program it up and show you a working version, or do you have to perform it there on the spot on the blackboard?
doubt it.
> Famously, he was sacked by the NSA for giving an interview opposing US involvement in the Vietnam war.
Another cult of personality in business. Bill Gates, Jeff Bezos, Henry Ford, JP Morgan, etc. managed to do a lot without cults of personality.
The more BS I see, the more I question what they are hiding, and the more I question their judgment in distracting their organizations from practical business, and distorting reality.
Honestly, I don't really get your perspective. Simons legitimately was one of the best mathematicians in the world and was sacked from the NSA. This isn't some Musk-esque 'I sleep on the factory floor and sign off on everything despite lacking the bona-fides' nonsense, he was 40 before entering the investment industry and before that won top research prizes, published quality research and ran a math department.
> ran a math department
OT, based on what university faculty have told me, so maybe not true everywhere: chairing a department is a thankless political nightmare. It's foisted on people who can't say no; it isn't an honor.
FWIW, I don't believe this myself. But that's where the interest comes from.
I am struggling to convey the magnitude of how skeptical I am of this claim.
https://en.wikipedia.org/wiki/Michael_Crichton#GellMannAmnes...
For those wanting more recent additions, the author seems to have more updated thoughts here[1], as linked on the LinkedIn profile to which their former blog redirects.
[1] https://blog.headlandstech.com/2017/08/03/quantitative-tradi...
This interference creates a very different probability distribution for the asset’s final price to that generated by the classical model. The bell curve is replaced by a series of peaks and troughs.
-- No, it's not replaced by "a series of peaks and troughs". This is nonsense. It sounds flashy, as it reminds of the peaks and troughs seen in the double-slit experiment, but it does not accurately describe what could be done to improve modeling with probability distributions. Looking at it from an information-theoretic point of view, peaks and troughs in a probability density distribution would just mean lower entropy, i.e. it would be implicitly assumed to contain (quite a lot) more specific information than another, smoother PDF. So where does this information suddenly come from?? If this is not what the author meant, then it is at least an unfavorable choice of wording to write that "the bell curve is replaced by a series of peaks and troughs". To have mercy on the author, one could maybe assume they meant to speak about a characteristic function (https://en.wikipedia.org/wiki/Characteristic_function_(proba...) but that does not seem to be the case.
Furthermore, any probability distribution may be used to model financial instruments, depending on how well it appears to be suited for the purpose of modeling reality. However, if the author already speaks about it so specifically, it is almost misleading not to mention that normal distributions (the bell curve) are in practice not used in the way described, at least not by people who know what they are doing. Consider why Nassim Nicholas Taleb (author of "Fooled by Randomness" and "Black Swan") said that no one in the industry uses Black-Scholes, or ever has. What he was referring to - correctly - is that (in the options space) nobody uses the normal distribution assumption to be correct for the modeling of asset prices per se. It is rather used as a stepping stone with some convenient mathematical properties to describe things analytically.
Broadly speaking, the classical random walk is a better description of how asset prices move. But the quantum walk better explains how investors think about their movements when buying call options [...]
-- Nonsense. There is not even a hint of an explanation why either of the two would be so. It is merely an empty sentence that reads well. Quantum walk explains investor rationale and psychology? And only when buying call options?! This is quite funny actually.
A call option is generally much cheaper than its underlying asset, but gives a big pay-off if the asset’s price jumps.
-- Not always so. It depends on many things. Calls actually consistently disappoint some buyers by moving much less on the way up than what they expected / had hoped for. Having looked at a call's delta as per Black-Scholes, they end up wondering why the call did not move as much as the delta would have predicted. It has to do with spot-vol correlation (and other things), but I won't go down this rabbit hole now... (I would say you can PM me if you are truly interested and want to know more, but it does not seem to be possible on HN.)
The scenarios foremost in the buyer’s mind are not a gentle drift in the price but a large move up (from which they want to benefit) or a big drop (to which they want to limit their exposure).
-- If you are a buyer of a call option you would certainly not hope for a big drop in the underlying (!) and neither would you limit your exposure to such event by buying a call. This is, unless you hedged it either delta-flat or fully, which essentially transforms the call into a synthetic put. Nothing of that sort is mentioned here.
The prices of such options closely match those predicted by an algorithm based on the classical random walk (in part because that is the model most traders accept).
-- No, they do not match a price "predicted" by an algorithm (assuming the author is referring to market prices of the options here). It is the other way around. The assumptions ultimately used to make the algorithm fit the market are what is "predicted" by the market. The "algorithm" referred to here is likely the Black-Scholes formula and it does not predict any market prices. It gives you an idea where the expected value of the option would be if all of its unrealistic assumptions were true (which they aren't). So you have a function with many parameters, one of the most important ones in this context being implied future volatility (average future variance to be super-correct). But you still have to make a choice of what such inputs you want to use for them (the formula will spit out almost anything for the right choice of inputs). In practice, a subset of these parameters differs for each option from strike to strike, so there is no "close" match found to market prices at all.
But a quantum walk, by assigning such options a higher value than the classical model, explains buyers’ preference for them.
-- No. Rubbish. How is this comparison even made. Assigning a higher value, based on what benchmark or standard of comparison? The same input parameters to the pricing model? Hardly, as a quantum model would likely have quite different parameters than a "classical" one. This is just textual fluff.
Such ideas may still sound abstract. But they will soon be physically embodied on trading floors, whether the theory is adopted or not. Quantum computers, which replace the usual zeros and ones with superpositions of the two, are nearing commercial viability and promise faster calculations. Any bank wishing to retain its edge will need to embrace them. Their hardware, meanwhile, makes running quantum-walk models easier than classical ones. One way or another, finance will catch up.
-- Hype paragraph.
Regardless, finance isn't playing catch up.
RenTech funds which are not Medallion have experienced losses similar to any other .
The idea is great: looking at past events to understand how a particular event such as a sunny day in Manhattan or what did the Yankees do the night before (and millions of other things) influences the price of a stock.
If you have enough data you can go back and test if a signal is really a signal or just random coincidence.
But somehow it's not working as well as it sounds on paper, I think one of the reason might be that signals like that are very weak and they fade over time so you need leverage to make sure that you cash in while the signal there (because it's weak and subtle) and also you end up losing a bunch of money by chasing a signal which is on its way out.
Hard to tell what the book is about. Sounds a bit like a Deepak Chopra take on economics. Throw a few mentions of "quantum" in there to sell books.
I would say its mostly in communication.
People set up laser beam networks on rooftops to get faster pricing data.
The processors and hardware are all pushing for faster reading and trading, with continual research at the physical level.
Some formulas still go into financial models or back into physics.
Just like the article started off its examples with.
Which is to say, to get your brokerage's orders in ahead of a large institutional order.
IOW: Legally-sanitised insider trading.
Maybe informative for others.
My actual opinion on that is that you used to have to go to the square where people were exchanging. People farther distances away simply were not there and had to wait in the newspaper to know what happened with prices. I don’t consider advances in communication to be controversial in that regard. I do think people should be aware of who sees their communications.