US mean reversion system

I haven't been able to paper trade this new system of mine as initially attended because LOO and MOC order types simply don't work on IB's paper trader.

The guy at IB told me that I "can't expect" the paper trader to be able to execute such specific order types - but he would fully expect that I would have no problems with a live account.

So instead of paper trading this system I will start live trading with a small account until I'm happy that performance is broadly in line with expectations before scaling up.

The above table shows performance of the system so far this year, from 1 Jan through to 21 May. Monte carlo simulations were run 1000 times.

Re-defining equity

Since the last update, five trades have been closed out. Three losses and two winners. Overall slight dip in closed equity since that October chart.

Portfolio is currently fully invested in the market.

I've also decided to use another method to determine my position sizing. I previously was using the TradeSim method which is available cash + open positions at the price paid. While the intention of this method is to be conservative (not count chickens before they hatch) in practice I have found it to be rather aggressive and to take on more risk than necessary as you are increasing your size in what could be the beginning of a drawdown.

Big winning trades are usually closed out when the market is correcting or experiencing a downturn of some sort, which is precisely when you DO NOT want to be increasing your position size.

The other obvious alternative to this is to use pure open equity marked-to-market, which I think is too aggressive and will result in too many fluctuations. Plus I don't really count open equity as my own.

So what I have opted for is what I consider to be a middle ground and to calculate my total equity as available cash + open positions at their trailing or initial stop loss levels (whichever is the greater). I think this method is conservative initially but as your winners start moving and stops are raised, it allows you increase your position sizes more dynamically.

The downside is I have not been able to test this method. But common sense tells me that it won't get me into too much trouble.

I have also been reading some of Ralph Vince's work, and I have been impressed. There are some areas which I have marked down for re-visiting down the track once I have time.

In other news, I've been busy trying to automate my US mean reversion system through Amibroker and IB.

Quarterly Portolio Update

There were no trades closed this quarter. The system is fully invested in the market and has been for several months now.

Just thought I'd put in an equity curve - probably not the best you've seen!

Start up date was the 9th October 2007 with $66,000 in capital.

Since then 51 trades have been closed.

We are still in the middle of a drawdown - max.DD stands at 40.56%.

System flaw

The gap fading system that I was hoping to trade on US stocks has run into a bit of trouble. Actually alot of trouble. The problem is that it can't be traded, and here's why.

I believe the opening auction system works differently in the US compared to the ASX. In the local market, the most liquidity for the large cap stocks especially, always occurs at the open and at the close (usually higher at the close).

BHP or CBA, for example, each of which routinely turnover hundreds of millions of dollars per day , would do about $2-10million right on the open.

But over in the US, even some of the high turnover NYSE tickers like WYE and JNJ, you only get 100 shares going through on the open! And these guys turnover the same or more than BHP or CBA. What i noticed in the course of sales is, about 5-10 seconds after the open, the party finally gets started and you see the 10,000 share and 20,000 share parcels start going through. Which is useless for a system that is supposed to buy "the open" as you wouldn't be able to catch this liquidity using "limit-on-open" type orders.

The NASDAQ opening auction system must work differently to the NYSE because they seem to have much higher opening liquidity across the board (in the hundreds of thousands range, so still not the millions I was hoping for).

So its back to the drawing board for the short term system. It looks like I may need to turn to intraday timeframes if I want to maintain my 100% overnight cash position and still generate a super smooth equity curve.

The leveraged model

Considering the maxDD stats were acceptable, even during the 2008 Black Swan, the system will be leveraged up. Here are the charts of CAR and maximum drawdown over a 10-year test (01-01-1999 until 31-12-2008), using 30% margin as per I.B.

The beauty of trading short term

You just can't beat that super smooth equity curve.

This system uses a different entry to the one in the system posted below (US stocks) but was tested over the same universe of stocks.

Seems to have more of an edge. High Pfactor, lower maxDD, higher win%, higher payoff, less trades. And did I mention the equity curve?

The performance report shows the system in its inherent form, to provide a fair comparison with the system below. The equity curve and monthly profit chart show the system traded on 30% margin, which is what it would be traded on, if it gets to that stage.

US Stocks

I've recently been looking at US stocks and the potential to trade short term systems there is amazing due to the wide range of stocks and their liquidity.

The system that I've come up with is similar to the system described here a few months ago that traded ASX stocks through CFDs. Both systems are based on mean reversion and never hold overnight. But while the ASX system shorted overbought stocks and bought oversold stocks, with several days lookback and magnitude moves considered in the entry criteria, this US system just fades opening gaps and trades long only.

I have found gap fading (particularly going long gap downs) to be a robust edge that has worked well over many different universes of stocks. The idea came from James Altucher's book and also from this trader describing fading extremes as one of his favourite setups.

There are no CFDs in the US but my broker does allow 4:1 leverage intraday if required. The universe of stocks consists of 50 tickers from either the Nasdaq 100 or S&P 100, so I'm looking for liquidity and scalability in the system.

The system makes about 30-100 trades a month and performance is consistent with profitability on average 10 months of the year. 2008 was a tough year, but out of 10,000 portfolios, monte carlo testing shows that 98.59% were profitable last year, with the median portfolio gaining about 10%.

So the previous STMR system will not be traded; and this new one will take its place. I intend to use IB's PaperTrader to simulate it for about a year before committing real dollars to it.

Performance stats below are from 01-01-2003 until 31-12-2008. Position sizing is 10% of equity. Commissions are included (however, slippage, is not accounted for).