Ask me anything - quant/quantamental Hedge fund manager/Consultant - Everything on liquid hedge fund strategies

combining a good technical knowledge (ml, econometrics, optimization) and deep knowledge of the products I trade.

Math is for granted, you need creativity and a lot of practical experience once your notional becomes "relevant" in the market

q1 was an impressive quarter for me, in the hf space it's ok to lag the market in bull markets, but in a bear mkt like this at least you should try to avoid big drawdowns.

1) it depends, true alpha comes from mixing many niche strategies, two sigma and citadel do it, aqr not really... 2) smart beta 100%, in fact it's basically a factor zoo, that in few years you'll buy at the cost of a plain vanguard etf, act accordingly for your careers 3) many funds are smart beta ish, aka 0-0.5 sharpe, don't be surprised if they underperform. It's hard to generalise, there are many niches: CTA is commotized (you can run AHL/Winton/Aspect programs from your cheap laptop), stat arb did well until march, but in general is strong, index arb got smashed in march but with all this passive flow is going to be good going fws. Many have strenghts and weaknesses, but mixing them like the multi-strats do is the magic (the only free lunch in finance is diversification, when you kind of know where to put your eggs)

there still plenty of money to be made, the problem is always to have the right size to justify the costs. For instance if you use a lot of alternative data your bill is probably few millions, so you need at least to make 15/20 millions do justify the costs per year,

Fundamental pms that pick quantamental tricks properly have an edge that is hard to replicate, but in order to have all the pipes almost plug and play you need a proper seat (millennium, citadel, exodus, etc.) coz if you have to build youself it would take a year (with a strong team).

Have to double post because my revised comment is not saving down.

It's been a crazy past few weeks. I've never seen this kind of volatility since back when I was trading equities and Getco's algos blew up in 2012 or 2013. I can't remember exactly what year. but that was only for a day or two. Anyway, I understand why a lot of hedge funds blew up with their relative value strategies with correlated instruments breakout down and diverting to unprecedented levels. Typically in this kind of environment, would quant strategies focused more on intra-day momentum, reaping ticks be more advantageous? Is this common or sought for compared to RV strategies? The point im getting at is these mega hedge funds took huge losses and these PMs are no rookies. I feel like realizing PnL from small tick moves and going in and out all day is safer. I mean the transaction cost might be a burden but I feel like overall, as long as you're right most of the time, it would be a better bet? The point I'm trying to make is these PMs are no rookies and they took a huge loss. Typically RV strategies don't tend to do well during market turmoil. I've created an algo where it takes advantage of intra-day mean reversion, scalping ticks in and out and I just feel like what I'm doing would be a safer bet being as long as there is a volatility, I will always make money. Of course, what I'm doing is not scalable to manage hundreds of millions but I just feel like it's safer. Yes, the transaction cost would be a burden but as long as I'm net positive, why not?

applying a proper quantamental approach is hard, you need to be a discretionary guy with solid edge in your sector/asset class willing to change/adapt. In Equity there's more things to do coz equity is mainly "stories".

A good example is central bank statements/speeches, in the past discretionary ppl were arguing they are the only ones with experience to manage the fed mood, but nowdays it's all there with a millisecond fight to get the text before others, and the info is useful for both worlds.

You need a lot of money to run quantamental books, so the only ones who are able to do it properly are the usual suspects in the edge fund space. In the long only I see a lot of hopes, but they are going to be smashed anyway by my fee compression so they play just tactical (ie marketing...)

you don't need to complicate the equation to have a balanced portfolio, the problem that kills the industry is retail money and offering liquidity that doesn't make sense: if you own a business with let's say 10m opex and you run 300m aum with daily liquidity, that means it could be wiped out in few days.. it doesn't make sense to have retail and offer daily liquidity for hf strategies, that's way UCITS funds is a big failure and the big winners are cayman structures with quarterly or annual liquidity...

Re your second point, from ir/marketing team perspective, the less the better, you don't want to deal with crazy retail investors and spending millions in marketing costs (like long onlys do) to convince someone to put a chip.

easy one: - if you are serious in HFT/market making is kdb c++/java and a bit of python - if you are serious in quant funds: probably the same mix, with a bit of mongodb/postgres

1) both pretty much all (with people monitoring it) 2) python is the standard pretty much everywhere for r&d (someone is using julia, but I have never tried), production is c++/java/go

1) they do it to filter the ridiculous amount of applications, I have a PhD, but it doesnt mean I am smarter, you don't need a master, just a solid foundation of math/stats the rest you can pickup online through courses and practical projects (eg quantopian, etc.), if you can get in early don't waste your time in grad school, experience is worth thousands of degrees (even top notch ones) because only hard work and persistence matters in this game, it's not PE, even harvard grads get slaughtered easily

2) it depends, they need to mix skills so physics, math, stats, or finance doesn't matter as long as you get quickly a grip about numbers and trading. I am not sure pure math is the ideal path towards a successful career in quant trading.

3) it depends on the role, if you end up in a research zoo you need basic skill in python (but typically you low paid), if you run the show in your pod you need to be more capable to do many things. It's a simple business: solid data ingestion, db mgmt, r&d, production, monitoring (and sometimes conferences/clients meetings)

4) data science in other industries, yes the industry is volatile and commodization is accelerating things. To give you an example my 4 ppl team was able to run a book across listes/otc products ingesting anything from tick price to alt data like twitter thanks to the right tech stack. It's a business that doesn't need so many ppl, so opportunities will be less (or simply you will get paid like a bi guy in any corporate).

never heard a q researcher bringing home 7-figures, I don't expect this to change. There's a totally different pressure though (as you might know).

I agree with you, you need an edge and managerial skills to branch out, some networking, trackrecord are stories (nobody should disclose his/her own one, you know the stories of the poor chaps on the press when they tried to move...)

In my opinion it's not about experience (met many quant researchers with 10/15y experience who are frankly hopeless for managing their own book), you need to answer to a simple question: could you build and manage everything from scratch (data pipes, db, r&d, production/execution, monitoring, client/team management)? If it's yes probably in 3/5 years you will manage your own pod.

if you have scalable (prop) strategies is easy, the requirements are 1.5 sharpe gross (in macro, in equity 2/2.5) for a 250m book. Why you want to move?

a pm package is 200-250k base

there are many online gigs like quantopian, worldquant, linking a mt4 portfolio to myfxbook. Good books are the ones of Marcos de Padro on machine learning, the rest is math/stats you study in uni.

sharpe 2+, 5% on gmv, 2% ish, 10m+

7.5-20% for mm, yes it's in the contract, not really depending how desperate they are to have on board, all mm are basically mind your own pod

Now I do consultancy for the big ones, I am one man show (I can handle everything alone), average pod is 4/5 ppl, 1 pm, 3 researcher/trader, 1 quant dev. A good year can be a 5m pot: 3m to the pm, 2m to the remaining 4 (probably the quant dev much less than the quant researcher.

kdb for fast strategies/HFT, mongo/postgres for the remaining ones (1h+ holding period)

Great, thanks a lot for the detailed reply. I have some follow up questions.

1. Are there any particular subjects in machine learning/stats that is useful for quant equity? (i.e. regression, svm etc..)

2. Which python libraries do you use the most?

400k out of college is fake news, it's possible to get 1 time of your salary, which is similar to IB comp

quant trading is different from quant pricing, the former is heavily biased towards stats, the latter is more fin math.

I think a tech career is comparable nowdays (as entrepreneur even better), but you don't get the thrill of money mgmt. I do a lot of cool stuff in fintech, but the thrill of being right (and making money as second derivatives effect) is not comparable.

Quants are typically not business ppl, the ones who are business ppl too are pms. It's a niche field, my advice is always to check what's going on in the real economy, do side things to keep your brain ready for everything.

Downside is you have 10 years experience (probably around 35) you are not a pm, recruiters want researchers with max 3 year experience, you are probably screwed.. It's like IB if you don't make MD you are not going to stay forever VP or Director...Plus the industry is shrinking so, there are less opportunities going forward

hedge funds is a tough career, quant is the same, the fact you might find a job doesn't mean you are going to stay a long time in the game, there's a tremendous luck component (ie your shop perf, you pod perf, your boss, fund raising regime). I recommend it only for the ones who really love mkts, if you chase only the money a developer career gives a better risk adjusted return.

A grad from Stanford in my previous from shop helped on the inflation model (which is a great place to start if you and code and have ML experience as there's a lot of big data), he got paid almost 400k USD bonus because the team has made a killing doing front end inflation fixings. It's not unheard of. But in general a data science grad is on 100-120k base and bonus is around similar mark.

not true, it's a good combo in my opinion, but you need to be a good geek with data.

re covid I don't have a complete picture, but that's what I hear from few friends

strat in sell side is typically a glorified developer (70/80% of the cases), quant structurer might have a chance to jump into a hf, but it's hard coz they sell smart beta strategies in the bank so to be honest I don't know what they can bring to the table (it was easier in the early days, but nowdays it's unlikely).

The typical path is to start with a grad program in a hf or a prop house, sell side is not a natural path for systematic trading (I started in a US bank, but I was lucky to find a seat in a multi-pm platform straight away basically)

a strat career is a good career, but it's not hf comp.

from strat you could do data science in any industry, some pay well like fintech or insurance, others a bit less (typically the capex intensive industries)

pretty technical, you should really figure out what drives asset prices and what's against your trades/strategies (that's pretty much in any style), quick decision making, lean/lateral thinking

I think the perfect combo is applied math for bachelor, pursuing all levels of the cfa to get a general understanding of finance, then maybe a master in computational finance, phd is not necessary (I did it, it was fun, but probably spent too many years in uni).

There are many big rollers in Europe: Marshall Wace, GSA, Squarepoint, Qube just to mention few of them.

1) 5/10% gmv 2) it depends on how much he/she has lost, and his/her references 3) 10 years ago many had many shots, I doubt in this cycle ppl get more than 2 shots. If you have a 10 years exp and you miserably fail in the last gig, your hope to get a researcher seat, but you are too expensive, so any recruiter will look at you like dead meat because recruiters are simple chaps they need to exactly match the requirements of the "buyer", they don't waste their time in some "value" play with you

From an organization perspective the big ones are all the same.

You have a ds team involved in handling data sources (structured and unstructured from a variety of sources, twitters, satellite, nowcasting etc) and some calculations so pms could leverage information on potential trades, and portfolio intelligence. Ds team does the initial validation and the data monitoring (ie api doesn't work anymore they fix it)

Visualization is in many forms (tableau, shiny, etc.)

for each data source cost is splitted among users/pm so 100k/year licence and 3pm using it, 33k/each per year. You have some variations from this, but at the end of the day someone has to pay the bill, and it's not going to be the mgmt team...

Then you have some "evangelists" how are like a bridge between the discretionary ppl and the quants, they promote what's inhouse and hear what pms need.

It's very hard to make it work, many failed and reshuffled the ds team several times in the last 5 years. Moreover incentives are not particularly aligned because ds teams are paid kind of fixed and pm is all relative, but the ds team knows where's the alpha more than the pms because they have the total visibility across the books, so at some point they either join a pod or get promoted to run their pods.

if you have visibility on the short term alpha for execution you have already all the answers you need, it's easier to move from hftish to mid-frequency coz you know how to handle data, if you have passion for mkts you can come up with ideas

a quant fund needs: a manager - pm, manages ppl down the list, sometimes code some researchers - testing ideas developers - building and monitoring infras traders/quant traders - making sure execution is smooth

research zoo is typical in cta shops: strategies is not very different in the last 10 years, you need people to do run some exotic research, typically the principals are 50+ years old that have already made their money, changes to the portfolio are not often so from idea to execution you see very low flow or you might need ppl to impress that you have a lot of phds. Typically these structures don't pay well. If you are in a team of 10/15 researchers for one big pot stay away, the chances of you having an impact on the pnl are quite low

1) 9 years ago, engineering with strong focus on applied econometrics, no top 25

2) phd is not necessary, they helped me but you can get the same with an applied math bsc+cfa and maybe a 1 year master in computational finance (or some coursera courses)

The best ideas I got are coming from discussing them with big rollers, I was lucky enough that one of my mentors is a former LTCM guy, brilliant minds accelerate the way you get intuition in this game. No degree will provide that, research papers are written by people who are paid to write papers don't forget...

lean multi-pm platforms, they are the only ones that could handle the costs of running a business like this charging the fair amount of fees (2-20+), unfortunately thinking that this business could be handled with 0.5-15% is a dead end.

We are going to have more hf concentration and tech market makers/prop houses. I don't think is going to change massively we will still have equity l/s, event, macro, quant, etc.

the key aspect in the industry is to survive, your employers want you to run a very niche strategy to reduce visibility and bargaining power when your niche becomes low alpha, act accordingly.

it's PE vs HF, the rest doesn't pay as much and is a bit boring, it's an endless debate in my networks, I see more lux sport cars in PE, so following the money I would say PE.

I think risk-adjusted PE is better, PE chaps are smart ppl with good interpersonal skills, but do they know where their target will be in 5 years time? Nobody knows... Unfortunately is an industry built on moral hazard (aka a debt load hoping the fed/treasury will help in the downturn), perf are roughly a russell 2000 with some leverage (plenty of academic papers prove this, it's not only my opinion), plus allocators love it because it's not mark-to-market...I think we are reaching the limit in terms of money managed by private players, so my bet is that a lot of things will be commoditized as well there. You definitely don't wake up at 5am to check that your asian pnl is smooth.

A HF career is in structural decline, they are too many useless funds out there, which are feeding the big/smart ones. it's more challenging than before, and fee compression is a hammer for all. But you will never get that thrill anywhere else...I think it really depends on your skills/personality, all the young chaps I talk to are always saying I want manage X, work in X,Y,Z because in uni I am the top one etc... but the reality is that is really hard work, my family says I am obsessed by markets, you need to be.

A tech career could play in two ways, as employee or entrepreneur, the former in in terms of excitement/risk taking is not comparable to HF/PE, but it's more secure. I think the big opportunity is in tech entrepreneurship, capital is cheap, you can run a team from remote, there are many hard problems that need to be solved. and financing is becoming even more smarter (instead of raising capital for marketing, firms are happy to finance your campaign for a chunk of your revenues for x quarters).

I am into VC these days for seeding and running my side gigs, it's fun and interesting, but the market thrill is not comparable...