We all know what LTCM propeller heads did in terms of risk modeling, and how they ended up, due to huge leverage (and unexpected correlation with the rest of the world, lack of liquidity on the other side, etc. etc.). Is Quantopian looking for algos which fit LTCM's criteria? Low beta, high alpha, low vol, low drawdowns, high sharpe, etc.....the only way to make money with these type algos is to add leverage, and that was the LTCM recipe for failure. Heck, I submitted a simple algo with 2X leverage and it is up about 6% in 20 days, and all it trades is bonds, stock and gold. SPY is about even during the same period. Is Quantopian loking to fund the next LTCM-style algos , or ones like I have submitted half-jokingly, i.e. 3 tickers for broad asset classes, low maintenance, naturally hedged, 2X leverage, low maintenance and high return?
Hello BT,
It's been well-established that Q is a bit of a seat-of-the-pants, learn-as-you-go operation (which is sorta the definition of a start-up, unless it is doomed to fail). This may sound like a criticism, but the context is that I don't think anyone has ever attempted quite what they are trying. They can fail by being knuckleheads and repeating the past (the LTCM path, I guess), but maybe they'll have an edge because the new thing they are attempting will succeed. Access to tools like the research platform and the backtester, coupled with data, all for free (well, not quite), available to some guy in his pajamas in the middle of nowhere is new, I think. It might work, it might not. At least it is different. But I don't have a finance background--maybe it is obvious that they are doomed to fail, at a gross level. Anyway, what do users have to lose? If Q succeeds, they make some money. If they fail, well, researching and coding algos is less mindless than watching TV and drinking beer.
Grant, the concept is cool, Web 3.0, start up-style, hip, you name it. But whoever is setting the criteria for selecting algos is setting it up for failure. That is the main message. The tooling here is great, but it is not the best, and it is not even close to institutional level (tick data, options, etc. are missing). I have said in the past and I stay true to this: I would pay a premium if they offer options backtesting, even with end of day data. I think a reasonable membership fee PLUS the hedge fund if it takes off should make Quantopian a more viable business...relying only on the crowd to overfit some algos in-sample is not a viable strategy, on the other hand.
whoever is setting the criteria for selecting algos is setting it up for failure
I'd be interested in hearing what they should be doing differently. Are you saying it doesn't make sense to do an equity fund? And if so, for what reason? Simply that a stock/bond ETF portfolio would do just as well in the end, over the long haul, and so they'll never be able to attract outside investors? Or is it that the guidelines and selection criteria are wrong and should be changed, but otherwise an equity fund could work?
You may have a good point that no Holy Grail will be found. No reward without a proportionate risk. But there is the advantage of not needing to pay quants salaries and benefits. No additional expensive downtown Boston floor space. No hiring-firing costs. Etc. A staff of 30 full-time quants might run $10M per year or more, I figure. And they'd be under the influence of a few Quantopian managing directors who would inevitably herd them to be less creative and diverse than a crowd-sourced approach.
LTCM thought they found the Holy Grail. They figured if you combine a bunch of uncorrelated assets, the net portfolio correlation and its correlation to the market can be zero, i.e. the aggregate protfolio will have no risk. They figured since there is almost no risk and the return is small, why not use leverage to increase the return multifold? And so it all started. Is this eerily familiar to you already? Replace "assets" with "algos", and you have Quantopian value porposition.
The other thing is that derivatives must be used if you can call something "hedged" with a straight face. That is the main drawback. But as I have said in the past, every step in the "active trading" direction must be jstified in terms of risk adn return (yes, it makes no sense to enrich IB or whoever the prime borker will be by trading millions of shares, if you can achieve the same results with a simple 50/50 buy adn hold). I will write some more, but these are the main points...and finally, the mainest point - they MUST relax or change the criteria, if they want a real world algo, and not some overfit code which will crash one month out of sample.
Well, by their own analyses, they are getting a lot of over-fit algos:
Personally, I think they may reach a point where they'll need to understand algos in detail (i.e. line-by-line code inspection), versus the black-box approach. It just doesn't make sense to me, in the end, if their goal is to have $10B AUM of institutional money. My hunch is that there will be some legal problems, or some big honkin' investor will stipulate that the code must be reviewed line-by-line. They can pull it off at this point, because it is their own small pot of money at risk.
"investor will stipulate that the code must be reviewed line-by-line"
If they cave into this, then it is over for them, since they will lose all their free(lance) algo writers such as yourself.
Re: monetization, one other thing they can do is entice their 70K users trade using IB or another broker, and have IB pay for order flow and/or take the other side of the Quantopian user trades. This is what major brokerages like Ameritrade do with firms like Citadel, and I know for a fact that IB plays the order flow game. It is a huge money maker, but in the end, it boils down to a bet (on the IB side) that the Quantopian users represent dumb money. Not very motivating, I know., but I suspect this is where they might be headed, if the hedge fund thing fails to gain traction.
Personally, I wouldn't mind sharing code with Q, but I recognize this may not be the sentiment of many. I'm not planning to trade my own money with the kind of long-short algos they are looking for: 50-200 stocks, scalable to $5M or more, etc. Besides, it doesn't seem like the kind of thing where there will be some kind of alchemy, a mysterious inefficiency known only to the author, which will fulfill his dream of getting rich on Wall Street, but only if he keeps it secret. And if what you are saying is true, everyone should just put their money into a basket of ETFs and forget about it. So, revealing code would not have a downside, but only increase the probability of a risk-free gain.
If folks are worried about Q just using their code (i.e. stealing it) and not paying them, well, Q could do that anyway.
Another angle is that some users may be using information that they shouldn't (i.e. information that belongs to current or former employers). Maybe they are worried about getting caught?
The thing is, at $10B AUM, they'll have non-freelance (and perhaps also free-lance) algos writers, I bet. Ends up being a recruiting effort. Who knows.
Maybe order flow is already being sold. I haven't heard them deny it. I kinda wonder if they have this thinking in mind with Robinhood, since I can't understand how it fits with the crowd-sourced fund model.
Lot's of good points, but let's not forget that inefficiencies disappear as soon as they get crowded, so the goal of everybody on Wall Streeet is to remain as nimble and under the radar as possible when trading. LTCM and plenty of other still active funds gathered billions in AUM while being total black boxes. Arguably they had some Nobel winning propeller heads at the top who pretty much represented the entire selling point, all else was a black box. So, yes the goal is to be black box, to varying extents, otherwise there is no reason for someone else to hand you over their money. I speculate that 50% of the users here have written "algos" which they do not care if they get used/stolen/etc. (I am one example of this group), 40% have written algos which they think are proprietary to them and are hoping to get an allocation, and perhaps 10% who are willing to share their code, but are also hoping for an allocation. Just speculation on my part. 10% is too much, in my opinion, as an industry average, since those "in the know" would never dare write an algo on someone else's platform for fear of it being copied/used/stolen.
Anyhow, when/if they add options trading, even with end of day pricing, the algos may resemble more the real world algos out there, but not before. Adding futures was a move in the wrong direction, in my opinion, since you are not going in the direction of a "hedge fund" but you are going in the direction of a black box "CTA". My 2 cents.
LTCM and plenty of other still active funds gathered billions in AUM while being total black boxes
Yes, but in these cases, the principals ("propeller heads") presumably know what is inside the boxes. Q is different, I think, in that there could be some real surprises in those boxes. The risk would seem to be different. Asking investors to trust you is different than asking investors to trust the guy you found on the internet from the middle of nowhere, who wrote the un-audited code, trading $25M of your hard-earned (or inherited) money, who has nothing to lose. It's hard to fathom how they'll stay on the current path to $10B AUM.
"hard to fathom how they'll stay on the current path to $10B AUM"
Indeed, but always rooting for the good guys, hoping that if that does not pan out, they will monetize the platform. I personally would not care if they sold my order flow, since I do not care about minutely pricing. But I do want to be able to overlay my strategy with options, i.e. add the secret sauce. I would pay a premium to test my secret sauce strategy on hundreds of underlyings.
@BT
Agreed...Seems like I always find myself running into a wall of volatility, and then wanting to back test if said volatility had been trading at a discount on the CBOE.
The material on this website is provided for informational purposes only and does not constitute an offer to sell, a solicitation to buy, or a recommendation or endorsement for any security or strategy, nor does it constitute an offer to provide investment advisory services by Quantopian.
In addition, the material offers no opinion with respect to the suitability of any security or specific investment. No information contained herein should be regarded as a suggestion to engage in or refrain from any investment-related course of action as none of Quantopian nor any of its affiliates is undertaking to provide investment advice, act as an adviser to any plan or entity subject to the Employee Retirement Income Security Act of 1974, as amended, individual retirement account or individual retirement annuity, or give advice in a fiduciary capacity with respect to the materials presented herein. If you are an individual retirement or other investor, contact your financial advisor or other fiduciary unrelated to Quantopian about whether any given investment idea, strategy, product or service described herein may be appropriate for your circumstances. All investments involve risk, including loss of principal. Quantopian makes no guarantees as to the accuracy or completeness of the views expressed in the website. The views are subject to change, and may have become unreliable for various reasons, including changes in market conditions or economic circumstances.