What's wrong with algorithmic trading?

When I first began learning about trading, I always wandered why traders still had jobs. I saw that fundamentals were of little use when trading shorter time scales (is this correct?) and that the technical/quantitative/statistical methods used my most traders could easily be programmed algorithmically. Not only because this would mean that you wouldn’t have to sit behind a screen all day, but also because it could potentially be quicker, more accurate, may notice signals that you wouldn’t and for the added benefit of trading without the influence of emotion!

Anyway, I’m still not entirely sure why it’s not all algorithmic these days? I’ve read somewhere that algorithms cannot adapt their approach as the nature of the market changes, but I’m not even sure how I would adapt my approach to satisfy changing market conditions. How would you?

Thanks

There is plenty wrong with algo trading as several flash crashs and Knight Capital aptly demonstrated. Computers are just big dumb machines and people like me who code them are prone to making mistakes. Everyone makes mistakes, but it takes a computer to really screw things up.

Anyway, here is the latest article I read on High Frequency trading which should give you an insight into the state of automated trading and for the future of trading

How the Robots Lost: High-Frequency Trading’s Rise and Fall - Businessweek

[B][I]Resistance is futile. You will be assimilated![/I][/B]

LoL! Sorry, I couldn’t resist!

That is the great puzzle for all of us no matter our approach. Complexity and conditions constantly evolve making it very difficult to compensate for the new shift before it has its way with us. - JP

What makes you think that an algorithm would be worse effected by a flash crash than an actual trader? You could always use a few lines of code that would command it to stop trading if losses reached a specified level or if the market was behaving ‘strangely’. You could also ensure that it used an economic calendar to trade around major news releases, as not to be blind-sided by unexpected events.

I just don’t really see how a properly designed algorithm would be that much riskier than a human?

Algorithmic trading has a lot of advantages such speed, not being emotional and creating and applying systematic trading rules. But the reality is a lot of the automated forex software available to retail traders just isn’t that great. You can write your own EA but then how good is that going to be compared to something created at Citigroup by a team of quants? If it was easy to do investment banks and hedge funds would have cracked the FX markets years ago with their robots. They’ve put billions of dollars into program trading systems and as CodeMeister’s article says the results haven’t been that great thus far. There’s plenty of analysis around the web about this such as on forexop where factors and weaknesses in robots are considered. In short the reason traders still have jobs is that there isn’t anything around yet that can replace human judgment. A machine has no way of understanding the market impact of the default of a sovereign nation, a Bernanke speech or a riot in Brazil, so for now people still rule and traders will continue to have jobs.

Perhaps if you were a coder or had experience with a software development project, you would understand. First of all the design process is at risk when people can’t communicate properly. Requirements are usually imcomplete or poorly written. Coders are under unrealistic pressures to produce something that works. This usually leads to them taking shortcuts or not testing code adequately. Then somebody wants a change made late in the development phase and this can lead to new problems.

Its no wonder so many software projects fail or produce flawed results. The rollout of the Obamacare system is just the latest software fiasco. I have little doubt that the underlying cause is very similar to what happened with Knight Capital.

Yup, to add onto CodeMeister’s link posted to business week, an interesting read is Dark Pools: the rise of A.I. reading… by Scott Patterson, is all about how it started and progressed to where we are now. The big boys’ idea of algo trading is different from what we are talking about. They go after latency of the network to trade differentials in price across the network and sniffing out the other big boys’ massive deals hitting the wires and then spiking prices around those events since they the market makers. but the principle is the same for us and them, they need an edge, the first bunch hitting high frequency trading had an edge and they coined it. Now all the big boys all play the same game and their edge disappears. So thanks to CodeMeister posting that article, fascinating to see that high-frequency algo world is already calming down.

Do I algo trade, nope, I place my trades manually, but being a coder, I have created my own price analyzer, so i stream my platforms live prices into my little app and that will do all the initial analysis for me across a number of pairs. Is it fancy, not at all, ugly as hell, but it tells me exactly what I need to know for my trading rules. Once the app throws an alarm, I will eyeball and then take it from there. so if a normal day was 100% effort to trade, i’m down to probably 10% effort now across more pairs. Once you have your edge, you make more money by scaling it out and doing what works more often if possible.

This is such a huge point but most traders never really get a handle on it. If you have something that works in a market the minute the knowledge begins to leak is the moment it begins to fail. Implementation leads to nullification. It has too, it’s how competition works. This is why, unfortunately, all of these open source EA’s will never make the majority of their users any money. - JP

It would be interesting to see at what stage a retail account would start to have its edge reduced due to stipulation on the trade sizes passing through such an account by the broker. Just how anonymous are retail accounts when they enter the realm of “medium sized speculators” with a multi-million dollar account value. Sure this account would still be considered small in terms of large investment banks, but when would these liquidity providers (the investment banks) start to manipulate such a retail account, knowing that it has an edge, and fully exploiting this edge for their own benefit.

It’s understandable that commercial investment banks are not interested in small time retail accounts which are turning a nice profit, and that’s not the point of this post. I’m just curious to know what the liquidity providers can see on their order books when large orders flow in from such retail accounts, accounts which are at the uppermost end of successful and capital value exceeding millions of dollars.

Not only are we not allowed to share a profitable trading system because the edge will diminish over time as the number of users increases, but do we also have to stay under the radar of investment banks once hitting a given level of success and account size? Worth a little thought, at least that’s why I posted this, lol.

Would love an ‘insider’ to answer this one, if you’re the biggest fish in a small pond (small broker), then surely they going to sit up and take notice of your transactions? This might blow up into a dealing/non-dealing desk argument again…

The short answer is no. Your account size and success won’t attract the attention of institutions, most retail trades end up aggregated into bigger chunks automatically via the brokers software, that are then moved into a larger pool of liquidity in the open market.

This is one of the reasons retail has wider spreads, they have to allow for the slippage in price that occurs over the few tenths of a second it takes to aggregate funds to an executable size. When all of this first started spreads were a lot wider. It wasn’t unusual for us to have to eat five to seven pip spreads on the EUR/USD in mini-acounts. As FX became more popular and the participation based increased it took less time for the brokers to meet the minimum execution size and spreads tightened. They tightened even more as more competition entered the market place.

So we never really have to worry about institutions eyeballing individual account orders, but we might need to worry about them eyeballing a large convergence of pooled orders with a boat-load of stops sitting just a few pips out from their entry.

Despite the appearance of there being a 1000 roads to Rome, most of the intraday methods, stops and targets of retail traders will fall into a pretty neat distribution. A few stops will be tighter, a few will be wider, but most will be clumped up together in a pretty tight price range. Often there will also be an overwhelming bias in direction. Take a look at the sidebar in the Alpari News Room. Alpari News Room Despite them having thousands of traders, which might lead you to believe the orders would be more balanced, their orders are currently 73% short on the EUR/USD.

What many institutions and large traders know that many of us don’t consider is, if you are trading at 50:1, 100:1 etc. you can’t take much volatility, as a result there will be some easy targets out there for improving a very large position.

I’m pretty sure when someone is looking at order flow it’s easy to see where the retail blocks are simply by the accumulation of stops within the normal range of volatility. I’d wager a pretty large bet, the majority of the stops on those short EUR/USD trades all converge into a pretty tight area, much closer to the current price action than the average institutional trade that is sitting out there at 1:1.

Probably also worth noting all those short EUR/USD positions are getting beaten up right now as the pair has ground higher over night… Retail stops are dropping like flies. - JP

Always nice to go against the crowed, I believe their are strategies which use this exact data for analysis, how profitable they are though is questionable.

I agree with you that stops are typically loaded into a given range, and are not always evenly distributed as one would expect give the size of the over all market/no. of transactions.