Automated Trading Strategies: September 2022 Portfolio Update
Over the last 12 months our top strategies made over $2.7M based on NT8 backtest results. This is a leaner portfolio in both size and performance.
In case you didn’t know, we’re on the hunt for the holy grail of automated trade strategy. We define the holy grail of trade strategy as having the following attributes:
Profit factor (gross profit/gross loss) greater than 3
Annual max drawdown less than 3%
Annual return greater than 500%
Maximum daily low of -$1,000
Avg Daily profit greater than $1,000
Less than 5,000 trades annually
Greater than 253 trades annually
Every two months I like to look at how our top automated trading strategies are performing and update you on key findings, highlights, takeaways and what’s in the pipeline. This is our September 2022 update. You can view past updates below:
Portfolio In Transition
Sometimes I need to bump up against what’s wrong to know what’s right, and if I had to sum up the last year, I’d say that’s pretty much all we’ve done. With every update, the data grows richer and our observations more insightful. We’ve learned a lot and we use what we learn in each successive strategy.
In the beginning of the year I talked about volatility and its impact on automated trading strategies. What we’ve found is that like everything else, the primary impact, and therefore vulnerability, on automated trading strategies relates to the backtest. In particular, it increases the likelihood that backtest results will be misleading.
After Strategy 10, we took a station ID to learn more about backtest risk. I discussed the topic of backtest risk at length. I told you what changes we were going to make to reduce backtest risk in all of our strategies. I then introduced a revised portfolio that was less prone to backtest risk and therefore more accurate.
What was the major difference between the old and new portfolio? In the new portfolio we:
Switched all strategies to a minute-based data series--Range and Renko are more complex bar formations so the simulation has a hard time recreating the bar formation for the backtest. Switching to minute-based data doesn’t improve the simulation, but it improves backtest accuracy by giving the simulation an easier calculation.
Diversified contracts--Using more than one futures contract increases profitability and decreases risk without any associated cost. It also allows us to expand the hunt to the highest performing contracts.
Conducted all backtests using high order fill--We conduct a high-order fill backtest (tick based) on strategies with only one data series. This is particularly important for minute-based strategies.
These are huge changes, but they were necessary. We lost some of you in the process. There were definitely moments when I had to push myself through a rough spot or two. The good news is that I think we’re strategically better for the transition. The strategies are the same, but the frequency and contract types have changed. The hunt is the same, but we’ve discovered a better path.
In practical terms this means that we’re optimizing all strategies by data series and contract type to find the “best” performance. And, this data series can be re-optimized with every contract expiration. We’ve started down this path for many reasons including our own empirical research as well as evidence from research studies.
In the fall of 2012, Kim Man Lui and Lun Hu published an article titled: Does a Generic Price Pattern Exist? An Alternative Approach to Technical Analysis in The Journal of Investing. The two authors used correlation studies to look for price patterns. They concluded, “that price patterns, if any, as reported in technical analysis literature, should not be equally applicable to any time series of stock prices.” Indeed, we know these price patterns exist so there must be a certain data series for each instrument that is better than others at creating price patterns. We also know this to be true based on strategy optimizations.
So, we’ve created a master strategy list to align the best instrument and data series with the best strategy. The list and master strategy will be published as Strategy 54 shortly.
We have found that multi-contract, -strategy portfolios are better than portfolios based solely on one strategy or one contract. This is especially the case for highly volatile contracts like NQ. We love NQ because of the price action, but it also greatly increases risk and usually lowers profit factor. That said, highly volatile contracts also tend to do very well on strategies that rely on finding trading opportunities during market extremes. Perhaps the ideal strategy is one that uses highly volatile contracts with a strategy that plays on market extremes, and low volatility contracts with a strategy that performs better in range bound markets.
There are many ways to construct a portfolio: by profit factor, percentage of profitable trades, net profit, drawdown percentage, etc. We’ll be looking at how various portfolios test in real-time over the next two months to gain some clarity on this question. I’ll also be creating trade profiles to help subscribers to better visualize the difference between the cash flows generated by each strategy.
September 2022 Performance Chart
Based on the September 2022 performance chart, it appears as though we’re on the right track with our changes. This is a leaner portfolio that we believe has much higher backtest accuracy. The next two months will be focused on testing that accuracy.
The following performance chart has been updated with backtests starting on 9/01/2021 and ending 9/01/2022:
The first thing you’ll notice is a shorter list. This is for several reasons:
We aren’t trading the FDAX contract any longer. It failed the data audit three times.
We’re in the process of retesting the entire portfolio.
Which strategies performed the best?
Strategy 2 (also available in the ATS Mini Newsletter) performed well.
Strategy 2 (30 minute CL) - made 322 trades on a profit factor of 1.64 in the July update, and 319 trades on a profit factor of 1.88 in the September update. Net profit is also considerably higher at $124K, up from $90K in the July update.
Strategy 6 (also available in the ATS Mini Newsletter) was a big surprise.
In the May update, Strategy 6 made $142K on a 1 year backtest with a profit factor of 1.29.
In the July update Strategy 6 made $185K on a 1.35 profit factor.
In the September update Strategy 6 made $234K on a profit factor of 1.43.
Whatever is happening in the NQ market, Strategy 6 is benefiting from it. We’ll see if the trend continues in November.
Strategy 31 had a nice increase in profit factor for the ZB 30 minute contract from 1.30 in the July update to 1.40 in the September update. Strategy 32 also had a solid performance for the ZB 30 minute contract from 1.53 in the July update, to 1.52 in this update. While it’s nice to see profit factor increase, ideally this is the kind of stability we’re looking for.
I also want to point out that:
Strategies 2, 48, 48a, 48c, 48d, 49, 51, 52, and 53 had a profit factor over 1.81. Our goal is to substantially increase this number over the next two months.
Strategies 15 and 53 have the lowest daily net loss — neither strategy lost more than $1,000 on any given day of the backtest year.
Strategy 12 never went below zero. That is, cumulative net profit was always positive from day 1. This is more about when you start the strategy rather than the merits of the strategy itself, but it’s worth noting.
Strategies 47, 48, 48a, 48c and 53 have a max drawdown less than 2%.
84% of Strategy 47’s (GC) trades were profitable with this update as well as July’s update. Unfortunately, the net profit is very low.
The September update did not disappoint. Even though we’re in transition, it continues to confirm our theory for selecting the best data series and instruments.
Where Are We In The Fight For More Accurate Backtests?
Most traders aren’t testing their own backtests and if they are, they certainly aren’t sharing the results. It’s a level of transparency that could make you look foolish, but we plan on using this research to conduct a live test of our own so the fear of losing money is far greater than the fear of playing the fool.
We started out thinking that we’d found the holy grail in Strategy 10, but got impaled instead. What can I say about Strategy 10 that hasn’t already been said? If it were a person, I’d say it was my best and worst friend, which is appropriate since this update serves as a kind of eulogy for Strategy 10 — may it rest in peace.
One thing is certain, we wouldn’t be where we are in the hunt without Strategy 10. The betrayal of Strategy 10 led to an all out investigation. The investigation culminated in strategies that produce more accurate backtests, but we still need to go back and test each strategy again. We’re testing them to see which minute-based data series has the highest performance and we’re testing them to see if they perform like the backtest in live trades.
Next, we need to decide on our opening move in the live hunt (starting in January 2023). We also need to make some decisions on how to handle things like rollover. I like the auto-roll feature, but you want the logic to follow volume, not a specific date, which means you either do a manual roll or you don’t trade. For the first test, I’m hoping to start at the beginning of the contract and end two weeks prior to contract expiry.
Another big question: what’s our funding source? We’re currently weighing the option of using a funding partner, but that requires a profit split and usually comes with rules that make it prohibitive for automated trading, however, I think a shift in funding is coming as more funding shops enter the market. The consequence is more favorable profit splits and fewer rules. In other words, a year ago I would have said that isn’t an option, but now it might be feasible.
Backtest Critics
There are some people that say even a live simulated test isn’t going to help backtest accuracy and that the only way to test a strategy is with a live test using actual dollars.
I say they are wrong and have probably lost a lot of money.
We’ve managed to carve out around 85% of manageable backtest risk over the last 6 months. The only thing left between a simulated live account and an actual live account trading real dollars is the potential for slippage, which happens in both directions and only tends to be a real concern for scalping trades and/or trades that use more than one contract. In fact, I would say that slippage probably only accounts for 10 to 15% of backtest accuracy, and there are degrees even within this range.
I conducted a little test on three simulated accounts and one live account. I ran the same strategy at the same time. I did this every day for a week. At the end of the week I found that all four accounts were different. It seems that one set of live data can create multiple performance results, regardless of the account type. Sometimes the live account gave up points and sometimes the sim account gave up points, but they were all within 15% of each other. Like systematic risk, this is a kind of risk that cannot be managed away, but the other 85% can be quantified and accounted for within your performance requirements. In other words, you want to make sure that your profit factor is at least 1.15. This isn’t an issue for us because we’re working towards a portfolio with a profit factor of 3 or higher.
It’s important to note that slippage also tends to be a larger issue for highly volatile markets with quick price movements. So, you’re going to have a higher degree of slippage for NQ than you would for ZN. In fact, we’re finding that our strategies tend to work better (read: have higher profitability) in markets with less volatility; as a day trader, this is an unexpected observation. It’s also another reason we’ve started diversifying the instrument list.
What’s In The Pipeline
Over the last two months ATS has published:
Strategy 51 - When traded as a portfolio using a 5 contract risk management system it made an annual net profit of $656K and had an average profit factor of 3.71 on 600 trades. Average max drawdown is only $6,679 (per contract), and on average 67% of the trades were profitable.
Strategy 52 - When traded as a portfolio, using only 1 contract this is a strategy with an annual net profit of $209K and profit factor of 2.88. It also has an average win/loss ratio of 3.49. That means the average winning trade is 3.49x more than the average losing trade.
Strategy 53 - As you can see from the performance chart below, this is a strategy marked by shallow drawdowns and steep profit runs. Based on our backtest, the portfolio made $126K for the year with a profit factor of 2.54. So this is a 1 contract, multi-instrument strategy that never went below -$30 on a cumulative basis and never lost more than -$750 on any given day. This is telling us that this strategy is very good at cutting losses and increasing gains.
We also published Is The Market Rigged For Retail Traders?, a little advice for those overwhelmed and frustrated by the market, and 3 Examples of Price/MACD Divergence.
Over the next two months, I’ll publish Strategies 54, 55, 56 and 57. We’ll also be refining the portfolio with higher performing strategies and testing the accuracy of backtests. Finally, I think it’s time to admit that we need a better tool for analyzing portfolios.
NT8 has many pros and cons. It’s great for strategy development, but not portfolio analysis. We’ve reached a point where the analysis just isn’t there. In particular, I’m looking for a platform that can backtest and optimize an entire portfolio of strategies. With Nt8, I can show you how the entire portfolio performs on a real-time basis based on the instrument, but not the strategy. And, there’s no way to optimize a portfolio of different strategies. So, I’m looking for a platform that can backtest and optimize an entire portfolio of strategies that trade multiple instruments, from multiple exchanges, in multiple time zones, using multiple bar types, and even order types. In a perfect world, a 3D heat or cash flow map would help to visualize the performance of each individual strategy. I’m currently testing a few platforms with these capabilities.
I also want to give a recommendation to all those that want to be a better trader. No matter where you are in the evolution of your trading skill-set, you will learn something new from John Hoagland of TopStep in a series of videos about trading strategies. It’s a four part series and is worth its weight in gold. You can access the videos here. Note: I have no affiliation with TopStep.
Finally, as part of a pre-test to our live test in January, we’re going to test two strategies using two different funded trader programs. The two strategies will be selected shortly. Results of this test will be shared with subscribers in the Mudder Report. Tests will begin on Monday, September 19. I’ll continue to update subscribers over the next two months and will provide a general update in November.
I want to encourage working in groups so if you’re a group of four or more, click here for information about signing up as a group, and here for the discount, or click on the button below.
Look for Strategy 54 to be published on September 15 to ATS, and Strategy 9 to be published to ATS Mini.
If you have any questions, comments or recommendations, please reach out by responding to this post or emailing at automatedtradingstrategies@substack.com.
Click here for strategy links.
Hi, what platforms are you testing? I believe TradeStation has a lot of capabilities and 3d visualizations.