Quant Funds and Automated Trading Strategies

[Future 10]

There was a topic somewhere on the boards about insitutions automating their trading more and more. Hence discretionary traders are amongst the rare breeds in many insitutional firms. The recent subprime problem has lead to a domino effect leading many hedge funds to post massive losses.

 

These funds are whats called Quant Funds that use automated computer models to trade in various equities and markets. The article from bloomberg about Goldman Sachs Global Equity fund explains their automated strategies of long/short. Article: http://www.bloomberg.com/apps/news?pid=newsarchive&sid=atrLyIlkKSjg

 

Futher information regarding this strategy can be found here: http://en.wikipedia.org/wiki/Long_/_short_equity

 

The second article is about Citigroup which may also be in trouble of around $3billion in losses: http://www.bloomberg.com/apps/news?pid=20601087&sid=ahW1pLE7WZ7I&refer=home

 

What interests me is the $3billion "rescue package" of Goldman Sachs fund. Furthermore, a fund that just lost 26% or so states "We don't think the strategy is going to change". With volatility going back to where it was, it is possible that Goldmans fund will recover. However, what intrigued me was how these so called mathematic based fund managers/traders/programmers have this much faith in their automated models. Did it never occur to them to view the market with their own eye? It seems like day traders are more informed of current market conditions becaus we breathe it tick by tick.

 

These funds go on to post 10-15% returns annually, yet when they lose over 25% in a matter of weeks their reason is "the models (ours included) are behaving in the opposite way we would predict and have seen and tested for over very long time periods (45+ years)". This sort of excuse is unacceptable in my opinion. A good trader is informed of market sentiment changes. A good trader can distinguish opportunities vs warnings. A good trader will know something doesnt seem right. These bots will never know what hit them until they are down in the hole and unable to unwind their positions. It seems like history repeats itself over and over again. Did we all not learn from LTCM crisis? I personally find it unacceptable that fund managers are panicking worldwide. Should they even be appointed as fund managers?

 

I understand hedge funds with that much capital must apply different strategies. But I think it would be beneficial for investors and the securities market if these funds simply hired and listened to a good discretionary trader instead of loading the boat over computerized strategies. Any thoughts?

[darthtrader 10]

I think we are in a "Quant bubble" right now. I noticed that last year looking at the autumn gold ctas that it seemed like a dirty word to put anything other than "0% discretionary" when it came to strategy. The no name state university here now has a "financial engineering" degree. When you have billions and billions all doing the same stat arb mean reversion stuff it seems like someone is going to have to be left holding the bag. Its not like these strategies are so mysterious. I've read Rentech model how they themselves will effect the markets from using their models but I don't think you can model what will happen to your model when everyone else is using the same model.

[torero 10]

There has been many questions about the markets being automated little by little and leaving discretionary traders as a bunch of minority lose cannons (Soultrader brought up this topic a while back). I think it is going this way, but the fundamental problem remains is who owns the money that they're losing? Machines? Computers? or Humans? If they answer correctly, then it's down to human psychology driving the markets, not machines (I've never seen a machine screaming "Oh my GOD, I'm losing millions, I got to scale down, scale down!!!!!). Of course the end result is the investors lose the money but Wall St. keep ticking with their fees and comms. End game: no matter how you play, they win.

 

Maybe we'll go back to basics. This is due to the fact that quant has enjoyed a nice bull market. Heck, anyone can make money at this point, engineer or not. Once the cycle changes (i.e. volatility, bearish, consolidation, range-bounce, what have you), that's when the real test on which funds/traders can cut it. This was exactly happened up til the tech bubble.

[Nick1984 10]

That video posted up a while ago on the forums about the crash of Long Term Capital was great. Putting 100% faith in a rigid system is a recipe for disaster. There are countless philosophies across a wide range of disciplines which derride rigidity as the downfall of any person or system. Look at Bruce Lee's philosophies on martial arts!

 

Like torero said, in a bull market you don't need much skill to make money, discretionary or automatic. For a while now these automatic systems have enjoyed the run up and the programmers have been taking the praise. However with the recent volatility, and the market falling like a knife (just try picking the bottom on this one!) these systems aren't coping. They are too rigid.

 

All automatic systems are based upon the thoughs of humans. Their advantage is that when it comes time to act the system will act without hesitation. However what a distretionary trader has an advantage over is in adaptation. A discretionary trader would have made a killing over the last 24-25 days where as these large funds are posting losses.

[Soultrader 12]

That video posted up a while ago on the forums about the crash of Long Term Capital was great. Putting 100% faith in a rigid system is a recipe for disaster. There are countless philosophies across a wide range of disciplines which derride rigidity as the downfall of any person or system. Look at Bruce Lee's philosophies on martial arts!

 

Like torero said, in a bull market you don't need much skill to make money, discretionary or automatic. For a while now these automatic systems have enjoyed the run up and the programmers have been taking the praise. However with the recent volatility, and the market falling like a knife (just try picking the bottom on this one!) these systems aren't coping. They are too rigid.

 

All automatic systems are based upon the thoughs of humans. Their advantage is that when it comes time to act the system will act without hesitation. However what a distretionary trader has an advantage over is in adaptation. A discretionary trader would have made a killing over the last 24-25 days where as these large funds are posting losses.

 

Yes very true about adapting to market conditions and adjusting strategies. This is something discretionary traders clearly have an edge over automated bots. Fund managers fear volatility while day traders crave for it. For those who havent got a chance to watch the LTCM documentary take a look here: http://www.traderslaboratory.com/videos/Trillion%20Dollar%20Bet%20Nova.wmv

 

Its a great documentary so worth your time. If anyone else has any financial films/documentary it would be appreciated as it is relatively hard for me to obtain them here in Japan.

 

I also want to add that learning market profile gives traders a tremendous edge even if its just analyzing end of day market action. Day by day thorough analysis gets you in the zone and alert to changing market conditions. Although candles and volume do help, being able to visualize a profile can offer clues not visible through just a volume and price chart.

[Nick1984 10]

Market Profile has been one of the hardest things that I've been trying to learn. Initially I thought it would be as easy as treating them as pivot points...how wrong I was! The information generated by a developing market profile is priceless if you're good enough to spot what type of profile is developping. An automated bot can't do that in real time effectively. The only way a bot can analyse MP is to have someone plug in some paramaters which identify a market as either being balanced, trending, range extension etc.. and therefore will always be making its decisions in retrospect to an already developed profile!

 

I have no idea how I'm going to begin to speed up recognition of MP development, I guess just keep on staring at it!

[torero 10]

I think these funds are another reason people should invest their own money with their own wits and knowledge. Most of us got caught up in the tech bubble (guilty here!) so another 10 years or so, another batch of wall street guys pitch the same spiel again to catch the next wave of losing funds. We sure have short memories don't we?

 

As for MP, reading the book is opening up new ideas for me. I'm already using the POC with success so can't say it's useless anymore. I'm going to delve into it more tasty setups.

[weiwei 10]

Market acts in a non-linear form, for years, mathmatians have being trying to solve non-linear problem, If it can be solved, those big investment firms would have paid what ever price to get it. but so far it remains a challenge for them.

 

Since so far it has no solution to solve a non-linear form, then next best thing they come up with is quant fund which is to build a model, with some assumptions that gambler's ruin would have 1/1000000 chance of happening(I made up that number, but the logic is that it is small enough that it would not happen).

 

"gambler's ruin" - It is gurantee 100% to happen if you play it long enough, no matter how small the chance it is.

 

any good statistian knows about this rules, and guess what, with more and more quant funds out here, the "gamble's ruin" that will go against them gets bigger and biggers.

 

Another other thing about building a model, it can never formulate all conditions into its mathmatic formulaed model. To make long story short, Non-linear form can not be solved by mathmatic, thus market action can not be solve by any mathmatic. Given it long enough time, it will run into that "Gambler's ruin" condition.

 

weiwei

[BlowFish 10]

Risk of Ruin or Gamblers Ruin as you describe it is the most important statistic you can be aware of. Even discretionary traders with long enough statistics can work it out. Of course the art is to stop long before you are close to ruin and re-evaluate whether your method and/or models are still working. Recognising you (discretionary) or your system (quant) is operating outside 'normal' parameters is time for serious action

 

Recalculating RoR based on current performance may be shocking enough to pull he plug immediately.

 

Cheers.