One of the Most Recognized Names in Quantitative Finance Shares His Views on the Market
The inventor of two models (FAM & AGR) that is used by asset managers worldwide with over $1 trillion under management shares stories and insights sparked by questions from members of the StockTwits community.
Ophir Gottlieb is the CEO & Co-founder of Capital Market Laboratories. He contributes to Yahoo! Finance, CNNMoney, and Reuters. Mr. Gottlieb was an option market maker on the NYSE ARCA exchange floor and CBOE, remotely, and then moved on to become managing director of algorithmic trading and client services at Livevol, recently acquired by CBOE. Ophir then moved on to become the managing director of quantitative research at GMI, a company which recently sold to MSCI. In that role Mr. Gottlieb invented the Forensic Alpha Model (FAM) and a co-invented the Accounting and Governance Risk Model (AGR). Mr. Gottlieb’s methodological approach taken in creating FAM was endorsed by the head of artificial intelligence for the state of Germany. Ophir is recognized as one of the first quantitative scientists to ever explore and apply deep learning with neural networks in finance and apply to billions of actual investment dollars. FAM and AGR are used by asset managers worldwide with over $1 trillion of assets under management. The FAM model has made Mr. Gottlieb one of the most recognized names in all of quantitative finance. Mr Gottlieb’s mathematics, measure theory and machine learning background stems from his graduate work in mathematics and measure theory at Stanford University. He continues to be cited by various financial media including Reuters, Bloomberg, Wall St. Journal, and through re-publications in Barron’s, Forbes, SF Chronicle, Chicago Tribune and Miami Herald and is often seen on financial television. For the full transcript please GO HERE.
Having already invented two models, what’s next for you? What type of new models are you working on now? — Jayhead13
I have one passion left in finance — break the information asymmetry that has benefited the few at the cost of the many for far too long. I was a part of that information asymmetry, albeit unintentionally. In my house, in our company, in my world, that unfair playing field will never exist, and I will pour my guts into making sure that future generations do not face the absurd task of trying to assess the value of an investment when the other side of the assessment already knows the answer. It’s like they watched the 6 o’clock news and we watch the 10 o’clock news and wonder why they already know how the story turns out. That doesn’t mean I’ll always be right, but you better believe, I am out to level the playing field. In that world, retail wins a lot more than they do now. And that means normal everyday people create wealth for themselves, their families and for their heirs. That is a sweet day, friends.
OK, so didn’t the neural models long ago figure out how to sniff out our stops? So how do I counter that? Leads me to a deeper Q about whether you think bot-charged volatility today is a qualitatively different breed. — bearcharts
My FAM model never looked at an order book. Stops, limits, that wasn’t the trade. The trade was finding a stock that would rise and a stock that would drop, hopefully in similar industries, to be stock neutral. I think the hyper focus on stops and limits getting run is way (way) over done. I mean, you canhide your limits, after all ;-)I think bots help retail — but don’t tell anyone. Back in 2008, my first year making markets on NYSE ARCA I picked up on one little bot trick — it was that Citadel made markets in 11 (21, 31, 41) lots, while almost everyone else made markets in 10 (20, 30, 40) lots. My best guess is Citadel did it so when trade allocation went through on a pro rata basis, they always got just a little more than everyone else. But, it was my first window into how to the machines could be seen as making mistakes. Citadel used a matching algorithm because it always wanted to be on the NBBO. Let’s just say someone could have walked that price up or up down where it was a better trade than otherwise, for oddly, an 11 lot of options… hypothetically.
As a option marker maker, did you play mind games with retail traders penny for penny or you dealt with big fish? — TraderBullAndBear
So, this is a good question in that it reveals a misunderstanding in market structure. I see a lot of messages on StockTwits blaming MMs (market makers) for losing trades. I think this brief read will both change your mind, and make you a better trader, which is to say, make you take responsibility for your own yarding plan and execution. Let’s be clear about this. Market makers make a fair and orderly market. Not only do they have little knowledge of the stocks they trade, they likely have less knowledge than even a retailer trader does. Market makers make markets in several companies, for example, I made a market in 140 companies. Retail traders, hey, we might focus on five stocks and we know every little bit about those companies. Market makers have no real idea if a company is on a hot streak of earnings beats, or if some brokerage was getting bullish. Not at all. And even further, let’s take a real-world example. There was a small pre-FDA biotech (let’s keep it nameless for this discussion but this is a true story), where some trader came in right before the closing bell and bought thousands of out of the money call options. Then, what do you know, after the close, the small biotech announces it was taken over and the option trader turned a few thousand dollars into millions. Which is to say, he stole millions from the market maker. This is how market makers lose not just their wealth, but their homes, their cars, even their families. I saw someone, first hand, commit suicide by jumping out of a window because of something similar. It was awful, and I will never, for the rest of my life, forget it. The market happens to market markers — they are not influencers, they are influenced. The job is simple, in a frenzied crowd where rumors are swirling and buyers are betting on some huge news, a market maker must stand up in that crowd, and calmly state that they are willing to sell while everyone around them wants to buy. In fact, it’s a tautology, when everyone wants to buy and is buying, that means someone must be selling. That someone is the market maker. Every time the bell goes off on a market maker’s computer indicating a trade happened, the first thing we all think is, “Uh, oh. Somebody knows something.” Every insider trading act that was caught (and the millions that were never caught) rip one person off — the market maker. This fantasy that market makers are the bad actors comes from two places, in my opinion. One is simply a misunderstanding of market structure. The second is much more sinister — it comes from a place where traders look to blame a mechanism beyond their control for a trading loss. Losing money on a trade and saying “it was the market makers” makes as much sense as playing basketball with a friend and blaming Michael Jordan for your own missed shot. I’ll leave everyone feeling better about this story — the trade in biotech above was caught by the SEC and no one was hurt — other than the cheater that likely went to prison. It was my second blog ever, having now authored more than 4,000 option dossiers. Quite a way to start and always my go to answer about anyone that thinks market makers do anything other than sit there, make a two-side market, and pray someone out there isn’t trying to cheat them. Of course, market makers aren’t angels, this risk is compensated for by something called the bid-ask spread. So, over millions of trades in a year, on average each trade accumulates a few cents of edge, market makers make good money. But one bad actor, one cheater, can literally end lives, and I assure you, market makers have zero interest in retail traders, zero interest in other traders either, as long they aren’t cheating, the bid ask spread, averaged out over millions of trades, is a good living.
For your QA approach what is your prefer language of choice and why? — likewhoa
Every task has a language that suits it. Matlab is great for mocking up, especially machine learning and unsupervised learning yet further. But, for production code, you need something better, or faster, like C++, JAVA, Python, etc. So, my favorite language to code in is JAVA, but most of my time is in PHP and java script for front ends or Matlab for machine learning.
What statistics do you look at specifically before taking an algo live. Do they match the backtest? Do you backtest? — psycandrew
This question needs a long answer, but I want to keep it brief. Every scientist has a method. It’s in that method where we define steps we intend to take, and the checks we intend to make, before taking an algo live. If an algo is meant to be predictive, it takes a lot of time. If an algo is meant as trading mechanism, so, if A happens, do B, that is actually pretty fast and much of the testing is actually done in real time with very small trades (like one lot option trades) That’s the type of algo trading I ran for Livevol on NYSE ARCA. The predictive modeling was on a much larger scale, at GMI.
I’d like to learn a programming language, which would you recommend for beginners (R, MATLAB, Python, other)? Algo/HFT is often blamed for increasing market volatility & reducing liquidity when it’s needed most — thoughts?— Lonesome
I think all programmers should start with C++, but I am old school. After that, it depends on what you want to code. For front ends, javascript and PHP (and of course HTML) are fine. Perhaps .NET is better. For machine learning, likely C++ or JAVA. Find your passion, and the language will find you. Algos do both, really. In fact, HFT algos make and take liquidity at the same time, that’s why they have to be ‘HF’ (high frequency). There’s no doubt that a fat finger can set the market on a wild move that was unintended, so in that realm, yes, it can create unintended and unprecedented volatility. Generally, each algo does something different. Some are liquidity takers, some are liquid makers — but most take liquidity. Generally, the market has been calm, and algos have never been more used, so we have some circumstantial evidence that algos relax volatility, though that is heuristic and not at all robust.
How did sale to MSCI go down (like can you tell the story of it)? What’s your top 3 things to look for wrt to Corp gov flags? (esp EM) Do you think that ESG factor still early stage? Is ESG style premia inextricably linked to quality/governance?— BadaBingCapital
A story I must take to the grave. When a CEO is selling shares and the company is buying shares, I get scared. I also don’t like it when there are “accounting changes” unless they are imposed by the SEC, in which case companies have no say in the matter. Finally, a lot of restructuring sounds like a slush fund — and that means earnings are not as clean as they may appear. I have done years of work in this area, and as far as I see it, E (environmental) and S (social) have no alpha. G (governance), does have a little — but just a little.
What are your thoughts on the future of cryptocurrencies? Thank you — LeoVeloce
I am very bullish on blockchain and I do think there will be a crypto currency that gains wide adoption — although it may not be an investment vehicle but rather a facility for exchange. I like the trend, I see similarities to the ICOs as the IPO market in the 2000 Internet bubble Everyone was right, the Internet got crazy big! But they were also right, a lot of the companies were going away and were crazy overvalued. Crypto will live, blockchain will thrive. Where they intersect, I just don’t know yet, but broadly, bullish on the technology and in a decade, I do believe 25% of people aged 50 or younger will use cryptocurrencies at least one a month.
What is your favorite options trading strategy and why? What options trading strategy do you dislike the most? — scheplick
I love momentum trades, especially as spreads. So, for example, buying a call one-week before earnings in MSFT has been a winner 3-years in a row (12 wins, 0 losses).
It very much defines the personality of this bull market, which is optimism. My least favorite trade is easy — one where commissions and slippage have a strong negative impact on the PnL. That means, really complex multi-leg option trades, especially in thinly traded names, are just the worst. If a trade fails bc it’s wrong, fine. But if a trade is right and commissions kill it, that is the worst.
Suggested reading on Forensic Alpha Model? — wellbri4
Read the little blurb that Rotman did on it. It’s on SSRN (author name: Ophir Gottlieb). They watered it down a lot, which was painful given where finance went, and other than a few high-powered scientists, no one really gave me the credit for breaking down the walls of finance and machine learning. I would say Izzy, that is to say, sub funds under Izzy’s Millennium were open to it — but they were open to anything that found alpha. I can personally tell you which firms, out in the open, where dead wrong about machine learning, said to to my face, and will forever be emblazed in my memory for the ones that missed… everything.
Can you explain a bit about how you applied deep learning w/ NN to finance and trading? — michaelbozzello
Man, there’s a lot to that question, as concise as it seems. Mathematics is a symphony and I am so fortunate that I took the time to get trained so that I could hear it. I’m no Bach, Mozart or Beethoven, but at least I can hear their symphonies. With that training, I took a constructive approach to model building, which is to say I built up from scratch, rather than down from a more complex model. Sometimes that felt like the wrong move — I don’t know, maybe it was the wrong move in retrospect. But finally, after months upon months of maddeningly little progress, I heard it. Like a flute that is suddenly given space to be heard over the rumbling Timpani drum in a symphony. It wasn’t the first time I heard the flute, but this time, that one time, it led me to an ocean of undiscovered methodologies that were just waiting to be found. I actually thought of them as puppies (I love dogs), and when I found them they jumped up and down absurdly like puppies do and started chasing their own tails and tried to tackle me, because they knew, now, finally, they were found. From there, I had to temper my excitement with rather strict and at times suffocating rules. A scientist is only as good as his/her methods and there was no way I was going to let my method fail those little puppies. This sounds absurd today, but I must tell you, when I showed the greats in the field that neural nets worked to do this one trick, they were both thrilled and utterly stunned. It wasn’t supposed to be that way. In fact, my professor at Stanford, a wildly brilliant and successful scientist and practitioner as of today, back then, when he taught my class, was not aware of the power of deep learning with neural nets, and in fact urged us toward other learning algorithms (like SVMs). But even still, his genius came through, just his one hour lecture on neural nets was enough to empower me to try myself. Man, what a day that was when it ended up fitting. What a crazy year it led to.
Gentlemen, what can you tell us about the position sizing model algo on such scale? — Cyclentrade
Position sizing with an algo is no different than you or me sitting in front of our trading platforms. Usually it’s a percentage of assets, or in MM terms, it’s the amount of haircut.
Do you think the gap between high volume traders and private traders become unprofitable for the latter group? And do you think arbitrage will continue to exist with constantly improving algorithmic traders? Why? — KevInvest
No, I think the opposite. HFT will lose. Longer-term holders have an inherent edge — stocks tend to rise, that is just a fact. Arbs will shrink in some areas, then gain in others. Relationships that demand arbitrage always exist and the more securities out there, the more of those relationships exist. But, the edge will shrink since everyone is looking at the same thing.
What’s your favorite sell signal? — bxvets
If I’m wrong on the first day of a swing trade, even if it is supposed last 3-days, I’m out. I don’t like trades that start as losers if they are short-term. In the longer-term, all I look for is any information that my thesis (bullish or bearish) has somehow structurally changed. If I see that, I’m out.
Can you show me how to use FAM and AGR. — Bullobo11
MSCI owns it. You want it, they will sell it to you. They killed FAM (I think) bc it was a massive risk to their structure of market factors. Luckily, we ran a hedge found around it so, payment was earned in other ways beyond the takeover.
What are the biggest changes you see or challenges for traders both retail and non-retail in the future? — coverthecall
The biggest challenge will be a reckoning between those that want to do research and those that do not. For the longest time, retail investors simply didn’t have the tools to compete with institutions, and in that sense, it made it easier. Now the tools are out there, and that means, some retail traders will build wealth and that wealth will likely come from those that do not study their craft (technicals, fundamentals, events, options, etc.)
A final shout out to our company; we have an option back-tester that is helping us break that information asymmetry and here is that back-tester (video).
We hope you enjoyed this Q&A! Press the 👏 below if you really liked it and feel free to highlight or comment on any part. We also have a newsletter for anyone interested in getting daily updates about the stock market.
I am the Community Manager for StockTwits. Follow me @michaelbozzello
