The opinions expressed in these forums do not represent those of C2, and any discussion of profit/loss
is not indicative of future performance or success.
There is a substantial risk of loss in trading. You should therefore carefully consider
whether such trading is suitable for you in light of your financial condition. You should read,
understand, and consider the Risk Disclosure Statement that is provided by your broker
before you consider trading. Most people who trade lose money.
I’ve developed a LFT long-only equity portfolio system that normally rebalances every 6 months unless there are exceptional market conditions for exiting the market.
Algorithmically driven, the trading method is loosely based on published academic research of price momentum from a number of well known researchers such as Fama and French et al. but also uses proprietary research ideas.
It’s been comprehensively backtested from 1996 using a historical point-in-time research database with delisted stocks to ensure there is no survivorship bias, which is a common problem with standard price databases.
For a specific typical diversification configuration, the backtest results from 1996 to the present show a compound annual return (CAR) of 32%p.a or risk adjusted return of 40%, with a max system drawdown of 27% at the peak of the GFC. The profit factor is 3.6 with sharpe ratio of 0.59. The recovery factor is 6.87. In monetary terms, $100K invested at start of 1996 produced a net profit of $23.3M.
The diversification configuration is adjustable which affects volatility of returns and drawdowns. i.e increased diversification results in lower return and drawdown volatility and therefore slightly lower CAR.
It trades only highly liquid equities in order to support large portfolios and/or large number of subscribers.
It also produces dividend income but this is not included in the backtest results.
It can also be a tax-effective strategy due to the low trade frequency.
It purposely designed to be a low frequency system to exploit the long-term “weighing-machine” characteristic of the equity market instead of the short-term “voting-machine” characteristic that is essentially just noise.
I’m interested to hear from C2 developers and users if this type of system would be viable for C2?
The average holding period (average Bars held) is 71 days.
The data frequency is daily.
The risk free rate was set to 0.0%.
The exposure is 80.72%.
Trades are on market close but this is not critical due to the long holding period.
I can publish the full stats and charts as generated by Amibroker if needed.
@WilliamTonks
Hi Will,
perhaps I should publish the full stats and charts to provide more info.
I will try to do this over the next few days.
Cheers,
Ern.
It’s quite different to TAA or diversified model.
These don’t provide the same level of market out-performance.
It’s also a much lower turnover.
It uses the concept of pure price momentum but is applied in a unique way compared to these models. The approach is unconventional compared to most trading systems which effectively trade on noise.
It also uses a novel market timing method that protects capital during extreme market events while not compromising returns following these events.
For an example please see the post-2002 and post-2008 returns below.
It’s also possible to trade an inverse market ETF during extreme market events to further improve returns but I haven’t fully pursued this as the returns are still good without this.
I don’t understand. It rebalances every six months but the average holding period is 71 days (bars?), which is less than four months?
In any event, such a long period before rebalancing (and consequent limited number of trades) raises the issue of over-fitting because you’re essentially ignoring all the intervening data and thereby effectively limiting the quantity of data that you’re using for your backtest. In other words if you have daily data, but only look at that data every six months, you are effectively using bi-annual data which is not a lot of data to gain confidence from. Many developers commit this error when they for example rebalance only on Fridays. If you do that, you’re basically cutting your data set by 80%, raising the question whether you can draw valid conclusions from such a small sample.
Is this a stock system? Did you develop it using the individual components of a particular index? If so, try testing it on a totally different index. Say, the Russell 2000 instead of the SP500 (exclude the SPX symbols from the test).
Norgate has a survivorship bias free data product compatible with Amibroker for US stocks that is in Alpha testing. Maybe you can ask them if you can be an alpha tester for it. I am one. Makes bias free backtesting very simple.
Yes - it does look low I agree, but there are periods when the system is totally out of the market. Another aspect is that for ease of back testing at the rebalance points the system sells all positions and then buys new positions instead of scaling. Some of these new positions may well be the same stocks held in the previous period but the back tester wouldn’t account for these positions as continually held. The only affect of this is on the stats for holding periods.
This looks unconventional and is one of the reasons why it outperforms most systems and does
so with low risk and very little effort on the part of the trader.
The algorithm is not based on the usual methods that just about everyone else follows. As I’ve mentioned it exploits the long-term momentum characteristic of the market that not many people realise is present because most people trade on short term behaviour that is just noise.
It’s also not looking at the data only every six months - this is only the normal rebalance frequency. It monitors the market on a daily basis for handling exceptional market conditions as needed, or for handling any corporate events or for position specific exceptional events.
For example, over the back test period from 1996 to 2015, for the typical diversification configuration, there were a total of 898 trades.
Without giving too much of the key ideas away, it’s based on a combination of pure price momentum with ranking. The pure price momentum method and the ranking method are proprietary but are based on proven academic research from various well known researchers in the momentum field. I’ve combined a number of different proven research ideas with proprietary research to develop this system.
One of the best books I’ve read on the concept of long-term market behaviour which initially directed my research is:
“James P O’Shaughnessy - Predicting The Markets of Tomorrow 2006”
Yes - it’s a stock system. It’s been tested on many different equity indexes with similar results as all markets exhibit the same long-term momentum behaviour that this system exploits. I personally prefer to use the S&P500 as it offers the most liquid equity market in the world, but the system performs as well on any major equity market index.
All my testing has been done using this database. I have been an alpha tester for Norgate for some time.
In the past it has been particularly difficult for independent system developers to test with research quality databases such as Compustat, but the Norgate product has changed the ball game
Also I have been trading this system myself since the start of the year.
It trades only the most liquid stocks in particular market indices such as the S&P500.
I personally prefer to use the S&P500 as it offers
the most liquid equity market in the world, but the system performs as
well on any major equity market index.
Hi Met,
based on the interest so far I’m currently in the process of setting up this system on C2.
I’m also considering offering different variants of the system with different risk profiles
and rebalance frequencies, so that traders can select a variant that suits their requirements.
Regards,
Ern.
I’ve published the back testing results from 1996 to the present (June 2015) for a typical system variant:
Trading universe: S&P500 stocks
Rebalance Period: 6m
Number of positions: 16 equally balanced
