This page has been archived and commenting is disabled.
More On The Futility Of Groupthink Quant Strategies, And Why Momos Are Guaranteed To Lose Money Over Time
Following up on the earlier post in which Robert Litterman basically told his quant colleagues that they either have to stop competing with Goldman or die, we present the following presentation from GSAM titled "Maybe it really is different this time" from the June 2009 Nomura Quantitative Investment Strategies Conference. In essence Goldman finds that when one factors in liquidity and trading costs, the returns by momentum and value (as well as most other strategies) strategies have in fact underperformed miserably over the years. Furthermore, with everyone now enjoying the tools of the trade courtesy of $29.95/month TradeBiot programs, the crowding in the quant arena will generate more volatile returns as well as lower Sharpe ratios, while the stronger results will be available to the smaller managers who focus on less liquid stocks or markets. But we thought liquidity was what all buyers were demanding and benefiting from? It is ironic that the quants most likely to succeed are those who stray from the beaten bath. Of course the only ones who profit are those who provide the liquidity, courtesy of every more generous exchanges which are exponentially losing traffic to non-exchange ATS.
Goldman presents the industry as two specific pools of participants: stickers and adapters. The distinction is as follows:
The Stickers believe this is part of the normal volatility of such strategies
• Long-term perspective: results for HML (High Book-to-Price Minus Low Book-to-Price) and WML (Winners Minus Losers) not outside historical experience
• Investors who stick to their process will end up amply rewarded
The Adapters believe that quant crowding has fundamentally changed the nature of these factors
• Likely to be more volatile and offer lower returns going forward
• Need to adapt your process if you want to add value consistently in the future
Goldman logically poses the question of who is right as both camps can not make money at the same time.
The crowding in the quant arena is problem number one according to Goldman:
- Leading up to 2007, quantitative investing had become an increasing percentage of the overall market.
- When quantitative investors were forced to unwind their positions simultaneously in August 2007, it created the dramatic moves that were synchronized across factors and markets.
- There is always the potential for more risk and higher correlations for quant factors whenever quant crowding is evident.
And as all familiar with the space, August 2007 presented a very good case study of what will happen soon to quant strategies as once again the same crowding is prevalent in several core factors.
Focusing solely on backtested momentum strategies, would demonstrate nothing too surprising.
Yet an analysis of the momentum portfolio netting out trading costs, and one wonders how it is possible that momentum quants are still in business:
And while the inherent collapse in cumulative returns when presented objectively is quite shocking, adding on top the impact of crowding and you get a real head scratcher:
Yet the momos are not the only ones who should be worried:
With all this adverse data Goldman asks now what, and presents the following observations:
When properly adjusted for liquidity and trading costs, popular quant factors have indeed become less effective recently.
The Stickers solution:
• Live with it
• Sharpe ratios still attractive
The Adapters solution:
• Develop proprietary factors
• Dynamic allocation to popular factors
• Incorporate human judgment, but chastened
And in this environment where all in the top tier share the same infrastructure (traditionally the source of alpha), how can quants generate new alpha
- Alpha can only be derived when prices don’t accurately reflect all public information (i.e., price ≠ fair value).
- This can only happen when investors over- or under-react to information.
- Thus, the Value effect is really an over-reaction effect; Momentum is an under-reaction effect.
- Investors tend over-react to information that confirms their prior beliefs and under-react to information that contradicts those beliefs.
- We can develop proprietary factors by looking for other situations where investors systematically over- or under-react to information.
The last is where Goldman is particularly good: as in they will sell everything they recommend you buy - the most proprietary approach to asymmetric information dissemination (and let their several best clients in on it).
Because as the chart below demonstrates, when everyone is on the same side of the boat, the embedded risk increases exponentially:
And some very interesting observations on how to fix the crowding problem:
- Quants have embraced objective models to forecast returns because they do not trust biased human judgment
- An alternative solution is to train analysts and traders to recognize and control their biases
- Not strictly a quant approach, but uses insights gleamed from quantitative analysis
- Can also use quant/statistical models to improve analysts forecasts of the underlying fundamentals (EPS, ROE, growth, payout, etc.)
Goldman's conclusions:
- Historical results for many popular quant factors look much weaker when we adjust for liquidity and trading costs
- Crowding has fundamentally changed the outlook for many of these factors:
- Likely to prove more volatile and provide lower Sharpe ratios in the future
- Stronger results still possible for smaller managers who focus in less liquid stocks or markets
- Some opportunities for quant managers
- Develop proprietary factors that capitalize on over- and underreaction
- Dynamic allocation to popular factors based on perceived crowding
- Work with traditional analysts and traders to minimize over- and under-reaction
The other take home here is that momo (and most other quant strategies) end up being value losers over time. This is especially true when crowding forces everyone on the same side of the boat. Which is precisely the situation currently. A rapid unwind of positions would result in significant losses for all involved in the game theory structure. Whether Goldman has the holy grail is unknown, although for all its posturing the firm relies on less sophisticated quant managers (virtually 99%) to generate its P&L. As nobody has the depth of market vision that Goldman does, most strategies over a finite time horizon will prove losers, and Goldman will likely once again triumph courtesy of its flow and prop operating divergence. In the meantime, naive investors will likely continue to lose money either gently or in forced bursts like the August 2007 quant crash. After all, nobody can beat the casino, especially not when the casino just happens to also be the biggest player.
»
- 6920 reads
- Printer-friendly version
- Send to friend
- advertisements -
Let's not confuse these Quant strategies which for the most part are low frequency and not latency critical. This is billions of dollars moving slowly and really in the same camp as mutual funds, with the attitude being why do the research - either invest in a factor portfolio or go with momentum.
The HFT (liquidity provider, yeah right) strategies that everyone here complains about happens in seconds and steals money from the strategies above. Momentum is measured in seconds or minutes not days or months, and factors are irrelevant. HFT strategies are also information disymmetry strategies as most of us dont have access to the information the co-located SLPs can look at or at the speed they do.
But please let's not confuse the two. Personally the daily stuff is working amazingly, our neural net optimized VWAP strategies which re-optimize through out the trading day have been working better than ever - which tells me that we are close to failure. We actually start reducing size after a while once things work out too well and you are not loosing money on almost any given day.
Couldn't they have just condensed it to, "Crowded trade?" I guess you can't charge exorbitant fees for pithy analysis.
How does this affect prop desks? One of them asked me to interview with them for developing their HFT software. I asked what magic formula do you have to make money consistently, but they didn't want to get into details. Seemed like a shady business to me so I didn't pursue. Any insights here on how these so-called prop desks are making money?
I can tell you. But then I will have to kill you.
off topic
so the "war on terror" has been renamed the "war on extremists"
consider the implications of that
Looks like dark pool systemic risk:
http://www.youtube.com/watch?v=Th_aBzrV37M&feature=related
These strategies have been loosing steam for some time now and they are not to be confused with the HFT strategies that are so profitable right now. The truth is that HFT strategies can be done intra-day at 14x leverage and do not require so much equity. I know a firm where the top developer (trader) made $20m this past year and I don't think the firm has over $40m in capital. You don't need investors for this kind of business because the capital requirements are low and the profits are high. On the other hand the other strategies are great for raising a ton of capital, being lazy and collecting fees. Much easier to invest in a screen of low PE, stable growth and high cash than actually doing research. It is also easier to go with the momentum than be a true trend follower and have your sharpe ratio decimated by drawdowns (which are very common with trend followers).
It is impossible for most people to replicate the most complex HFT strategies but anyone now can copy momentum or three factor strategies. I assume this will change. Our genetically optimized VWAP strategies which re-optimizes several times a day and trades on tick bars and is executed by a co-located broker has not lost money in I don't remember when. This is not normal and we have been reducing size since we expect the party to stop. Who knows when? I know that a transaction tax would basically make these strategies obsolete. This will allow the guys getting rebates to be the only game in town and make more money. It's a rigged game. The public markets are for calculated gambling (that's why I am a quant) and private companies and real estate are for investing. Anyone who socks a public stock it their 401k thinking that they are investing is just mistaken.
Only in the short term, my friend. The fundamentals catches up to everyone in the end.
As a CFA, I'm all about fundamentals, and I do think that there are some good long term buys here and there but overall there is very little value in buying public companies (maybe not small caps). I just dont think the average person or financial advisor has the skill set to invest in public companies properly, most funds dont outperform the market so indexing makes sense, but the indices are overvalued and generally not what I would consider a good investment. Good investments generally perform irregardless of market price. So say I buy a bond for 40 cents on the dollar and gives me a great yield, this has very limited chance of default, if I am happy with the yield then it is a good investment as I am getting a good return and will eventually get my capital back. The price the next person will only matter to me psychologically. You can't make that argument for the S&P500 let's say. Maybe you could at 800 but not here, and those buy opportunities come at most a few times a year.
Really enjoy your seeking alpha write ups BTW.
off topic again (sorry)
strategy -> we'll begin withdrawing in 18 months
counter strategy --> lie low like brer rabbit (for 24 months)
The recommendation is to go to portable alpha like PIMCO, no?
Take a strategist idea, play with the tails and build a risk-adjusted trade, and then leverage it up. Alpha morphs into beta.
Some poker games the house takes a chip from every pot and they always come out ahead, everyone else is gambling against the odds.
In a casino, you know the odds. The rules are predictable. Markets are nothing like the casino.
just a basic question. are markets chaotic or random? is there an underlying pattern? maybe if you abstract out? or is this whole thing completely and truly random?
momo works over time IF:
you have discipline to honor your stops
you have discipline to get out when the trade breaks its momo
you don't put all your eggs in one basket
you don't put on too much size.
momo works best for individuals / smaller funds. NOT good for big players.