How To Build Robust Algorithmic Trading Strategies

Algorithmic Trading Educational & Tutorial Videos
Algorithmic Trading Design Methodology Primer

This algorithmic trading educational video is a great overview on how we assembler our algorithmic trading systems, also called portfolios – presented by the lead developer at AlgorithmicTrading.net. Topics covered include: Can anyone predict market direction with 100% certainty? Is there a better way – using market direction pseudo-arbitrage? We cover in great detail how we define market states and how each of our seven trading strategies performs during up, down and sideways moving market conditions. In addition, he reviews the design rules that we utilize prior to releasing an algorithm to the public.

If you are a curious observer or die-hard algorithm developer yourself, this video might provide insight into how we tackle the problem that all designers face, namely: How do you assemble a portfolio of trading strategies that have expectations to outperform in ALL market conditions.

It is our opinion that the algorithms we offer are high-quality trading systems, however no trading system is perfect. Trading futures & options involves substantial risk of loss and is not appropriate for all investors. With that said, if you have risk capital available to trade and are willing to accept the risk trade off – with the potential for great gains, reach out to us for a live demo where we will review our algorithms in great detail!

[addtoany]

How To Design Algorithmic Trading Strategies

Transcript of Video: In this video I’m going to be reviewing our design methodology and how we use market directional pseudo-arbitrage to put together portfolios that are designed to do well in both up down and side ways moving markets.

First I would like to go over our disclaimer. We are not registered with the CFTC as a commodity trading advisor. We are what’s called a third party developer. We create algorithms and we license them for use on a personal computer through trade station. And they can also be auto executed through NFA registered brokers. Keep in mind that trading futures and options involves substantial risk of loss. These algorithms are really not for everyone. They should be traded with risk capital only in our opinion.  Lastly, the data that we show, unless otherwise noted, is based on hypothetical back-tested models. It does have certain limitations. Per the disclaimer here. Feel free to read this. You can pause the video and read it more carefully. The last comment I’ll make is that the data is for educational purposes only. Again, because we’re not registered we do not control client accounts. Also I guess I should mention that because we’re not registered. The data that we show has not been audited by any government agencies. So just keep all that in mind as we go through the data in this video. We’re going to be talking about performance. Again, that’s based on hypothetical back-tested models. It does have limitations. At times we do mention the live returns on the website or in the video. When we do that is from live data and we note it as such. When we do that, again, just keep in mind that past performance is not indicative of future results. Again trading futures does involve risk of loss and is not for everyone.

Steps To Create A Robust Algorithmic Trading Strategy

Okay, so now I’ll start talking about our design methodology. But first I’d like to just comment on predicting market direction and trading algorithms in general. So, obviously no one can predict the market direction with 100% certainty. There’s no such thing as a holy grail trading algorithm. There always give and take. If you have an algorithm that aggressively buys the S&P. More than likely it’s going to do well when the markets going higher obviously. The hard part is the market trade sideways or gets volatile or starts trading lower. That algorithm that does good when the market goes higher will take losses as the market rolls over and trades sideways. Getting stopped out of trades. So the more aggressive an algorithm is in trying to capture those gains. Usually the more likely it is to take losses. Because it will be getting into to trades. That might be a little bit more risky then get stopped out of those. Another example would be if the markets going sideways you could obviously just short the top of the range and then buy the bottom of the range and keep doing that. The problem is is that when the market breaks out or out of that range. Then you’ll get stopped out of those trades. Or if it sells off then you’ll get stopped out of the long trades that you take towards the bottom of the range. So, that’s just in general the problem that any algorithm developer faces. So, you want to try to capture the gains when it goes higher. Capture gains when it goes sideways and then capture gains when the market goes lower. So if you knew 100% certainty that the markets going higher in thirty days from now. If you had a crystal ball then obviously you could buy. Load up on the S&P and then just sell after 30 days. But the problem is no one know with 100% certainty. So with that said, I’d like to move on to the next slide.

Combining Multiple Algorithms

So at Algorithm Trading.net we do feel that we have a better way. Instead of creating a single algorithm that does good in all market conditions. What we do, and this is really the core of our principle at Algorithmic Trading.net is that we use a market direction pseudo-arbitrage algorithm. So what that means is that we attempt to be market direction agnostic by implementing a pseudo-arbitrage technique. It’s not 100% arbitrage because it’s not a 100% paired trade. Where we’re equally long, equally short the indexes. It’s more of a design that would overlay multiple trading strategies into one complete system. So what we do is we create multiple algorithms that trade together as part of a portfolio. We attempt to have one to two algorithms outperforming in up markets. One to two that would outperform in down markets. Then one to two that would outperform in a sideways moving market. The key in the real R&D effort when you do this kind of technique is trying to make sure that these hero algorithms don’t loose too much in contrary market conditions. Either they go on the side line or they take small stops. Or you could potentially have small gains. But that’s really where the brunt of the effort lies in our opinion. Is to create these algorithms that do good in certain market conditions and then don’t loose too much during contrary. Cause what happens is when you overlay all these algorithms together. Then you have a complete system that’s doing it’s best to outperform no matter what the market conditions are.

Step 1: Define Market States

Alright, so the first step in implementing our trading methodology or design methodology is to define market states. So we need some way of comparing each one of our strategies to how it does based on if the markets going up, down or sideways. The way we did that is we took our back-tested period. Which the one we did most recently was from October 2003 to October 2016. Which has about a 163 total months. What we wanted to do was to categorize each month of the S&P into three equally distributed states. Either strong up sideways/drifting or down. Again the market can trade higher, lower or within a range for any given time period analyzed. So someone can do this on a one minute chart and analyze each one minute candle as its own kind of performance state. You can do it on days, you can do it on months. I guess you could even do it on years. We chose to do it on a monthly basis because we felt that that would give enough room to see how the strategy trades throughout that period. So again, we define the market states into three categories.

The market is either going up, it’s either going sideways or drifting higher or down. So to kind of show you what I mean. Let me pull up a chart of the S&P 500. So this is the E-Mini S&P’s. So essentially the S&P 500 on a monthly chart. So each one of these candles is one month. So for example, right here is October. Where the market traded down in October of this year. Here’s another down month in 08. So you can see 08 here when we had that bear market. If I zoom out and go sideways you’ll see every month since 03 on this chart. So what we wanted to do was create three categories. So that we can compare our strategies to each category to see how they do. One way to do it was we could have just said well if it’s a green candle that’s an up month. If it’s a red candle that’s a down month. But we wanted to characterize the kind of in between sideways moving conditions. Where there’s either a lot of all utility. The market goes up and down within that month. But then ends up trading very close to where it started. The way we did that is we partitioned each month into a category.

Strong Bull Market State

So the strong up, we define that as the S&P closes higher by 30 points or more. Which would be roughly about a percent and a half. Maybe a percent in some of the earlier months in our period. What that equals on the S&P E-Mini’s is a gain of more that $1,500. So if somebody bought the S&P that month and then sold at the end of the month. They would have made $1,500 or more. That’s how we categorize a strong up month. In the history of the 2003 to 2016 there were 55 months in that category. The other category that we did was down.

Strong Bear Market State

So if the S&P close down by four points or more. Which would be a loss of $200 or more. We categorize that as a down month. There were 55 of those.

Sideways Moving Market State

There is also an in between state. We call it sideways/drift higher cause it’s really sideways or maybe up a little bit. And so that’s the in between category that we call sideways. Where the S&P would have been down between four points or up 30 points. Anything in between there. So, either a loss of $200 or a gain of up to $1,500. There were 53 of those months.

An Equal Distribution of Market States

You can see that we created these values in order to have three equally distributed states. Because we didn’t want any one state to be over representative of that period. So, that’s how we decide or define our three market states.

Step 2: Evaluate Trading Strategy vs. Each Market State

The next thing that we do is we compare how each one of our algorithms does in each of the market conditions that we’ve defined. Now, what I’d like to do is quickly show you a chart of our S&P Crusher Package. This is the portfolio that trades all seven of the algorithms that we currently trade. What you’re looking at here is our momentum plus the covered calls strategy. This one here is the iron condor. It trades the S&P E-Mini’s. This one also trades the S&P E-Mini’s. This one is the ten year note treasury algorithm. This one trades the ten year. Usually when the market goes lower. When the S&P goes lower. The ten year normally will go higher. It doesn’t always, normally it will go higher.

So this, by going long the ten year note (TY). It’s very similar to going short the S&P. Although it’s not 100%. It doesn’t always do that so just keep that in mind. This algorithm here is our gap short algorithm. It’s the newer one that we recently added. This one is our break out day trade algorithm. This one is our short day trade algorithm. So these three are day trades. This one is our ten year note treasury algorithm. This is our iron condor options algorithm. Then this is the momentum plus the covered calls algorithm. So, what we did is we take each one of those seven algorithms. We call those strategies. On the website you’ll see each one of those partitioned on the website as a trading strategy. These are the kind of discrete strategies where each one has a different expectation of performance. Based on either a up move in market, a sideways or a down. Again, up is it closes up by 30 points or more for that month. Sideways the S&P close between down four points and up 30 points. Then down is a loss of four points or more. So what this shows is if you take the strategy. We’ll take the momentum first. That’s the bull fire we use to call it. But we changed the name to momentum. This one is the algorithm that we have that does best when the market goes higher. You can see that we took the monthly back-tested performance of the momentum algorithm. We lined it up against any S&P month that closed more than 30 points. The average gain for this algorithm is $1,600. A unit size, if someone was just trading the momentum algorithm, would be something around $20,000. So this is a gain of about $1,600 in the month per $20,000 traded. Again, this is trading one contract and even though it says 1,600 here. This is the average over all the up months. I believe there were 55 of them. We average $1,600 gain in the back-testing for each one of those months. What’s really interesting is if you then look at how this algorithm does in a sideways moving market. You’ll see the performance drops off quite a bit. It’s only up $378 for that month. Again, this might represent 10 trades in that month. But it just depends, each one has a different average. Then again, when you go to a down month. This momentum algorithm does not do well. It actually looses $812 per month on average. So again, this is the example of if we knew that the S&P was going to be up for that month. If we have the crystal ball. All we would do is turn on the momentum algorithm and shut off all the other algorithms. Just put everything in this. The problem is though is that no one knows with 100% certainty. I’d like to also just look at this algorithm and kind of show you the example. So this is the momentum algorithm. Again, this is the one that does good when the market goes higher. If you look here. The algorithm is basically applied to the S&P 500. That’s this top chart. This bottom one is the VIX and it’s in this chart because the covered call algorithm uses it. But what you’ll see is that as the market goes higher this momentum algorithm starts getting in. So you can see here we bought on the 9th of November. At 2158.75 and then we got out two days later at 2165.5. So this is a gain. Anytime you see the blue dotted line like this. That’s a gain and so as the market was going higher we had a winner right here. Another winner here. Another winner here. Then another winner here that we closed out yesterday. Then Friday at the close we got back in on this algorithm. As the market goes higher we had one, two, three, four winning trades in a row. That’s to be expected. Because this is our long bias algorithm. It should do well when the market goes higher. Now, you’ll see those as the market was going sideways through here. We did have a few winners. I’ll zoom in on that so you can kind of see. Looks like I jumped past it. So right in here as the market was going sideways. We didn’t have any trades. The algo really did the right thing by shutting off through October in here. Then it got back in right here on the 6th of October. It got out when it hit its target the next day. Jumped right back in on the 7th. Hit it’s target on this big gap up. It got back in on the 10th and then got stopped out. Jumped right back in on the 11th and got stopped out again. So you can see as the market was going sideways. It did kind of okay. But then as the market rolled over it got stopped out. That gives you an idea of how at least this algorithm, is designed to do well when the market goes higher how it does. So when the market goes higher. We generally speaking, are going to expect this algorithm to do well.

Now, it’s not 100%. I believe its of those 55 months where the S&P close at 30 points or more. I believe this algorithm is 85% profitable. So, 15% of the time it gets those conditions wrong and result a loss. But here’s another sequence of trades from back in July of this year. Where as the market went higher. You know we had one, two, three, four, five, six, seven. Seven winning trades in a row then we got stopped out. We had another one, two, three, four, five, six, seven. About another six winners before taking another loss. Again, we don’t expect this algorithm to work 100% of the time in all market conditions. This is what we call our hero algorithm for the condition where the market is going higher. So again, that’s the momentum algorithm. It averages 1,600 per month back-testing. When the S&P is up it averages 378 per month. When the S&P goes sideways back-tested. But then it’s at a loss on a down month. Now, I’m going to quickly go over the other ones. Now that I’ve shown that example. But the iron condor also does really well when the market goes higher. And an iron condor essentially. It’s selling a call and selling a put at the same time. What it’s doing is creating a kind of a bracketed range that we want the S&P to close at for that week. But the iron condor actually does really good when the markets going higher. Even though it’s selling calls. The reason why is because the calls that it sells are deep out of the money. The puts are as well. So it does good when the markets going higher. When the market goes sideways it does really good as well. You’ll see that in sideways conditions this algorithm does the best. That makes sense. An iron condor is a trade where you’re expecting the market to go sideways. That’s why people place the iron condor trade. So as long as it closes between our two strike prices. It’s a full profit trade. Now when the market is going down the S&P closes for that month four points or more. This algorithm averages a loss of 195. Let’s look now at the day trade. That’s another algorithm. This break out day trade that does good when the market goes higher. It’s also barely profitable in sideways. We really don’t count that as doing well sideways but at least it doesn’t loose money. Believe it or not it actually does not loose money really in down markets either. That’s because it’s buying strength in the morning for a day trade. In down markets this break out algorithm just doesn’t really trade that often. When it does it can capture some pretty big short covering rally’s but then it gets out that same day. Let’s look at the covered calls now. So those. The covered call also does the best during sideways market conditions. But it also does good when the markets going lower. The reason why it does better in sideways is only because it places more trades. When the markets going sideways it’s selling calls that are tied to the momentum algorithm. As the momentum algorithm gets into a trade it’ll sell a call for that week. That’s why the numbers 523 is higher than when the markets going down. It’s only because it places more trades. But it also does good when the market goes down. That should be real obvious. Because we’re selling calls and then the market closes lower for that month. More than likely those are going to be profitable trades. This break down short day trade is similar to the break out day trade but it goes short. So it shorts into market weakness for a day trade. It does it’s best when a market is going lower. So this also make sense. It’s going short, so of course it will do better when the market goes lower. It’s kind of interesting, it also does pretty good in sideways market conditions. This is not a surprise. That when the markets going higher, it is in a average loss for the month.

One comment on this is that when the market is going higher. It’s pretty rare that the short day trade will fire. Because it requires market weakness for it to get into the trade. But when it does, and the market is going higher it’s more than likely going to be at a loss. So then the gap short day trade is another day trade algorithm. Does best when the markets going lower. This one shorts morning gaps and then has a target. It’s actually a newer algorithm. But it helps out in the down market conditions. In the sideways market conditions. Then when the markets going higher it really just doesn’t. It just kind of average. Doesn’t trade a lot. When it does it’s probably half winners, half losers.

Alright, so the treasury note algorithm. We used to call it the push pull. But it trades the ten year note. This is our best algorithm for down market conditions. So again in the back-testing. From October 03 to October 2016. This algorithm averages about $1,000 per month when the S&P closes lower. That’s also to be expected. Again this treasury note usually rally’s when the S&P is going lower. It’s a swing trade that we take and that’s why it does good when the market goes lower. It also does good in sideways market conditions. Actually, really good in sideways. Yeah, it’s the third best algorithm we have for sideways. I think it’s mainly because as the markets going sideways after going higher. We start getting into the treasury note trade. Then as the market starts rolling over, going sideways. A lot of times the ten year note will begin to rally and so that’s why it will have gains. When the markets going higher, the treasury note doesn’t do that good. It still averages $48 per month. But that’s pretty much flat. So, it might having a winning trade here and there. Or it might just be on the sideline. But so, to summarize this slide. After we’ve defined the market states. Of what an up, down or sideways market is. Again, we wanted them equally partitioned in the three categories. So that they weren’t weighted to heavily to one side. This shows how each algorithm does for those market conditions. The next slide I’m going to show you how we put that together in a portfolio. You can probably begin to imagine how it works. But, basically what we do is we look at. Okay, this algorithm does good in up conditions. We want to have an algorithm like this so that we’re covered when the market goes higher. The treasury note does good when the market goes down. Doesn’t loose a lot when the market going higher.

As we overlay all these algorithms together, you can start seeing how we have this pseudo-arbitrage/market direction agnostic portfolio of algorithms.

Step 3: Combine Algorithmic Trading Strategies into Final Trading System

Okay, so in this final step. What we do is we assemble the algorithms into a portfolio. What you’re looking at here is. Again, this is based on the back-tested models. But this is the S&P Crusher v2. It trades the iron condor. The momentum, the break out day trade, the covered call, the break down short, the morning gap and the treasury note. So, the Crusher is a collection of these seven algorithms. And what I’ve done is just copied the values from here. Into a new slide just to show the totals.

So, when you sum up everything. It gives us an idea of how we expect this algorithm to perform based on what the market is doing. So the total is when you add all these seven algorithms together. It’s about 3,541 and that’s per $30,000 traded. So, our kind of general expectation is that if the markets going higher. Closes up by 30 points or more. We expect the Crusher to do really well. And that’s primarily because the momentum algorithm and also the iron condor. But the other ones also contribute. The only one that doesn’t contribute is the short day trade. Which actually does have a average of losses during up conditions. Then we look at sideways market conditions. So, when the market is going sideways. For a month, we expect the iron condor to do really well. On average to make about $1,000 in that month. The covered calls to also do well. Then the treasury note to do well. So when you sum these together you’re at 2,661. On a 30K account, so that’s just shy of about 10% on a sideways moving market. When the market goes lower. This algorithm still does really well. It averages 751, which is about a 3%. I believe that’s about. I’m sorry, about 2 1/2% for the month. That’s mainly because of the treasury note algorithm. You’ll see that the momentum algorithm cuts into to some of the gains that you would have seen. One question that we look at is: Do we really want to have the momentum algorithm in any package if it takes these kind of losses during down market conditions? It’s our opinion that the answer is yes. Because it does so well in up conditions. In sideways. So, that’s kind of how we put together these portfolios.

The S&P Crusher has very high expectation for up and sideways market conditions. Also high expectations for down conditions as well. Again, keep in mind this is based on the back-testing. This doesn’t mean that we guarantee that every month these algorithms make money. They don’t, we have loosing months. This is just the average based on the back-testing. It’s also how we track the performance in the live trading to see how they match up with these expectations that we have.

This next slide is looking at the same data. Just in a little bit different way. It’s using this kind of stacked column chart or bar chart. What we’ve done is. We have each algorithm a different color. Then we have it categorized in the strong up, sideways or down. What you can see is that. You know, just another way of looking at the same data. So when the markets going higher, the iron condor contributes a lot. The momentum contributes a lot. These other ones contribute a little bit. The break out day trade does well. When the market goes sideways, you can see the iron condor also helps. When the market goes lower, the treasury note does the best. These other ones help out. But then we do loose some because of the momentum algorithm right here.

S&P Futures & Options Trading Strategy Example

So this is our S&P Crusher. It is our flagship portfolio. It’s designed to do well in all market conditions. We do feel that this is our best package. It requires about $30,000 to trade it. That’s based on what the brokers. What the brokers require. In theory someone could actually trade it with less. And so that $30,000 just a starting account. It can take a draw down of. I believe over 50% before it would have to turn off. In these algorithms there’s plenty of buffer built into it. But that’s kind of our case study for S&P Crusher. Where we use this kind of market direction agnostic design methodology. So put together a portfolio that we have high expectations for moving forward.

S&P Futures Trading Strategy Example

Okay, I’m going to quickly go through the other portfolios that we have. This one is a ES/TY Futures V2 portfolio. This algorithm is very similar to the Crusher only we took out the iron condor. So, it trades still the momentum, the break out, the covered call, the short day trade, the morning gap and the treasury note. Because it requires only $20,000 in the account. It does have a slightly better performance on a percentage basis during down market conditions. It’s about 5% per month as opposed to 2 1/2 on the Crusher. So this is a great. A great package for someone that maybe wants to be a little more focused on a potential bear market that might be coming. But you can see that it still does really good in up market conditions. Really good in sideways as well. This is kind of a similar case study picture. That we had on the other slide. Only with the ES/TY Futures. Just another way just to look at the data. The momentum, once again, helps out on the bull. Takes away some in the bear market conditions. But then these other algorithms kind of make up for those losses. So that we still have a really high expectation for it.

Day Trading Strategy Example

The day trader is a portfolio that we put together mainly because there are some people that do not want to hold positions overnight or over the weekend. They don’t want to do options trades. So this is only the day trade algorithms. The unit size on this one is $10,000 so each unit trade is 10K. You can see that the amount that this one is expected to make is lower than the others. So you’re at about 3% for an up market condition. But it does do really well in down. This package really would be for someone that just really doesn’t want to hold any positions overnight. Because when you do hold positions overnight, there’s always a little bit of extra risk. Due to geo political events that might happen. Overseas markets that could impact the US markets. So, because it’s a day trade. We’re in and out the same day. Most of theses algorithms have pretty tight stops. They’ll get in. Usually in the morning and then either get out at the close the same day or when they hit their target or stopped out. Again, this is a similar way to look at this data. Where you have the strong up, the sideways and down. So I’ll quickly go through this one.

Bear Market Trading Strategy Example

This last one we call the bears trader portfolio. Really what we did is we just took the four algorithms that did the best in bear market conditions or down markets and we put them into a single portfolio. Even though this, I know most people are bearish right now in the markets. If someone asked my opinion I would probably agree. I know there’s probably a temptation for people to gravitate towards this portfolio. You know, we can’t give advice. All we can do is show data. But I do want to caution people that this portfolio in up market conditions is really not. We really don’t expect it to do very well. Because it doesn’t really have a hero algorithm for the up condition. It does still have good algorithms for sideways. Has really good ones for down. But this portfolio is out there just in case someone doesn’t mind taking that risk trade off on the up market condition. Just really they just want to be hedged for potential downside. So, this is the data for the bearish trader portfolio. Then if we look at this chart. Again, as you would expect. This looks really good on the down. Looks pretty good on the sideways. Then strong up though is not that great.

So, I think that covers how we put together portfolios. I think one other thing I want to comment on is we are talking about design methodology here. So, really what I’m not talking about is how we create the strategies. What we’re showing is what we already have. The strategies that we’ve had. That have been under development for the last three years. Some of them have traded live already for well over a year. But there are a lot of strategies that we develop that we don’t put out to the public. Because they don’t pass all of our tests. So, the seven strategies that we have. That I showed on this first slide. These are the seven strategies that have passed our design requirements. Once it passes our design requirements. That’s when we put it in a excel spreadsheet like this to see. Okay, how does it help out in up, down or sideways? This is really showing our kind of 10,000 foot view on how we go about algorithmic trading. So we don’t try to create one algorithm that does good at all times. We create multiple strategies with each one having an expectation based on what the market might be doing.

More Tips on Creating Robust Algorithmic Trading Strategies

Back-Test 10+ Years

So now I’m going to be talking about some optional kind of good practices that we use. So, what this is. Is just kind of a graphic showing some of the design criteria that we use. When you put together a trading strategy. The biggest risk is that it’s over optimized. The way that we try to fight that or do our best to ensure that the live returns will continue on the path of the back-tested hypothetical. Is we’ll do things such as back-test as far back as possible. Minimum 10 years. In a portfolio we want to have multiple uncorrelated algorithms.

Ensure A Reasonable Profit Factor

We want to have reasonable profit factors so anything between 1.2 and 2.6 in our opinion is good. Some people are looking for profit factors you know, bigger than two, three, four, five. 10 even. In our opinion, I’m just not sure that those really exists in live trading. I have coded algorithms at profit factors of 20 and higher. But you pretty quickly realize that those are over optimized. That their just not practical. So we like to, instead of trying to create that holy grail. We try to get more simple algorithms that have reasonable profit factors. Then as we overlay them on top of each other. We create that more kind of robust trading system.

Large Average Gain per Trade

We like to see large to average gain per trade. We don’t like scalping day trade algorithms. I have coded algorithms before the the trade on. One minute candles and have equity curves that go pretty much straight up. But the average gain per trade might be $12 which is the bid ask on the S&P. So, in our opinion those are not good. We look for kind of more, a bigger average gain per trades.

Look Inside & Intra-Bar Order Generation

You always want to make sure you use Look-Inside, Intrabar Order Generation. That’s just something to do with trade station where if you don’t have this set. It can make assumptions about whether or not a stop was hit first or a limit order was hit first.

Include Slippage & Commission

You definitely want to include slippage and commission.

The Smaller Number of Technical Indicators the Better

Generally speaking, the fewer the technical indicators used in the strategy the better. We like to see three or less. One reason why we mention this is because if you have a trading strategy that uses one technical indicator. That’s ideal because that means that you have a good algorithm that only looks at basically one indicator. Really the only reason why people would have more than one is because when they had one it didn’t work. So they might have had a second so that they can get rid of some of the gains. Or the losses in the back-testing. They keep overlaying more and more indicators. Which really just creates kind of a recipe for over optimization.

Monte Carlo Simulation

If possible and if it makes sense. We do the Monte Carlo simulation. We don’t always do that on every algorithm though.

After Back-testing, Modify Inputs

We like to modify the inputs after we optimize by plus or minus 10% to ensure that there’s minimal impact.

Large Number of Trades

We like to see 200 trades or more in the back-tested history.

Evaluate in Live Trading

And then we like to trade it live prior to offering. The duration of how long we trade it live just depends. It could be as few as two weeks as in the case of the gap short algorithm that we recently launched. Or six months as was the case with the momentum algorithm. But we do like to trade it live before we offer it.

Draw down Should Be Scalable

We want the draw down to be scalable to meet needs. So, we want to make sure that there’s a way to lower the draw down if possible. So instead of trading a $30,000 per unit on the Crusher. Someone could trade one unit per $60,000 that cuts the gains in half and also the draw down.

Optimize Inputs In Large Increments

We do not over optimize. So, when we do our optimization. For those that aren’t familiar with optimization. We will have another video out that will talk about that in more detail. But we don’t want to over optimize. So we optimize on fairly large increments. We do not optimize down to the tick level.

Independent Third Party Evaluation – if possible

When possible, independent third party evaluation. Early on we did have this done on some of our older algorithms. We haven’t done this on the more recent ones.

Only Trade High Volume Futures Contracts

The last thing is a scalable system that can handle volume. So, there’s a ton of different futures instruments that can be traded. We like to focus on the ones with the most volume. So the S&P, the ten year. Those really have the most volume and so we stick to those.

Final Word on Designing Robust Algorithmic Trading Strategies

So the purpose of this video is to give everyone a real high level view of what our trading design methodology is. You know, how do we look at the market? How do we try to out perform it and provide value? As an engineer, we’re always trained to solve problems. When I looked at the market. The reason why I enjoy it is because to me it is one big problem to solve. There’s a lot of different ways to solve this problem of trying to out perform the market. Or to profitable with algorithmic trading. Lots of different ways to do it. I’m not arrogant enough to think that my way is the best way or the only way. But I did want to just kind of share the way that I deal with the market. The way I do that is I just recognize the problem. That you just can’t predict market direction. I mean, we can make estimates. You know, when the bull fire triggers we know that the wind rate is about 70% on that in the back-testing. The live trading so far it’s for the most part matched that. But we don’t know for certain. There’s always that 30% chance that we’ll be wrong on that trade. So, instead of trying to make that algorithm have a 100% wind rate which is impossible. We create other algorithms to cover these other market conditions.

We build a suite of portfolios that have high expectations. Perform well in all market conditions. I just want to emphasize that it doesn’t always work that way. I mean you could have the market go higher and the bull fire for whatever reason could take losses for that month. So it’s not 100% certain but in our opinion it does stack the odds in our favor for future of success. So then once we have a strategy. Then we analyze it. You know, we run all the tests on it. We back-test it. We do walk forward as well on some of them. I haven’t really talked about that. We will in another video though. But we do all kinds of testing to try to break the algorithm. To try to figure out is there something we’re missing on this. Where in going live it’s not going to do as well. One we feel good about it then we kind of analyze it. See how it does in up, down, sideways to see how it can fit into our portfolios. Then there’s the various sign-off’s that we do prior to offering to the public. Also want to emphasize that in some cases we don’t do all the sign-off’s. In particular the Monte Carlo simulation which is essentially randomizing trades to make sure that there’s no hidden patterns that are being missed. We don’t do that. One reason is that we don’t do it on all of them is because we trade the indexes. There is a even flow to the market where it will kind of go higher, then sideways, then either roll over, go higher again. That’s some of the patterns of wins and losses. In our opinion are built into the market. So we don’t always do that. We haven’t always had the independent third party review. But what we do think we have is a really good suite of algorithms. We’ve been around for about three years now. We’ve kind of every six months to a year. We might come up with upgrades. New algorithms that we add into these packages.

We would love to talk to you though. So if you have any interest in what we have to offer. Than just give us a call. You can also email us. We have people that can do a live demo if you like. They will share the screen. Similar to what I have here. Kind of walk through each one of these algorithms to show you all of the performance reports. To go in more detail. Also on the video page of the website we’ll have more videos. Where we’ll be talking about these strategies in detail. I really enjoy talking about algorithmic trading. As the Lead Developer I really enjoy it. Far more than doing kind of the sales stuff. So, I’m hoping to get a lot of videos out there. Hopefully if anyone out there is a developer and they do their own algorithms. Hopefully, some of the data that I provided might be helpful to you as well. To think of your algorithms as how they do in different market conditions. Then use that information to try to plug in any holes that you might have in your portfolio. But again, would love to talk to you more. Answer any questions, provide a quote.

Again, trading futures and options involves substantial risks of loss. Not appropriate for everyone. But for those of you that are active traders or investors that are looking to take advantage of algorithmic trading. We definitely feel that we’re kind of a leader in this space. We’re doing our best to provide quality product to people. Hope to hear from you soon. Alright, have a great day. Thanks for watching. Bye, bye.

 

Start Auto-Trading Today With One of Our Algorithmic Trading Systems!



AlgorithmicTrading.net provides trading algorithms based on a computerized system, which is also available for use on a personal computer. All customers receive the same signals within any given algorithm package. All advice is impersonal and not tailored to any specific individual's unique situation. AlgorithmicTrading.net, and its principles, are not required to register with the NFA as a CTA and are publicly claiming this exemption. Information posted online or distributed through email has NOT been reviewed by any government agencies — this includes but is not limited to back-tested reports, statements and any other marketing materials. Carefully consider this prior to purchasing our algorithms. For more information on the exemption we are claiming, please visit the NFA website: http://www.nfa.futures.org/nfa-registration/cta/index.html. If you are in need of professional advice unique to your situation, please consult with a licensed broker/CTA.

DISCLAIMER: Commodity Futures Trading Commission Futures trading has large potential rewards, but also large potential risk. You must be aware of the risks and be willing to accept them in order to invest in the futures markets. Don't trade with money you can't afford to lose. This is neither a solicitation nor an offer to Buy/Sell futures. No representation is being made that any account will or is likely to achieve profits or losses similar to those discussed on this website or on any reports. The past performance of any trading system or methodology is not necessarily indicative of future results.

Unless otherwise noted, all returns posted on this site and in our videos is considered Hypothetical Performance. HYPOTHETICAL PERFORMANCE RESULTS HAVE MANY INHERENT LIMITATIONS, SOME OF WHICH ARE DESCRIBED BELOW. NO REPRESENTATION IS BEING MADE THAT ANY ACCOUNT WILL OR IS LIKELY TO ACHIEVE PROFITS OR LOSSES SIMILAR TO THOSE SHOWN. IN FACT, THERE ARE FREQUENTLY SHARP DIFFERENCES BETWEEN HYPOTHETICAL PERFORMANCE RESULTS AND THE ACTUAL RESULTS SUBSEQUENTLY ACHIEVED BY ANY PARTICULAR TRADING PROGRAM. ONE OF THE LIMITATIONS OF HYPOTHETICAL PERFORMANCE RESULTS IS THAT THEY ARE GENERALLY PREPARED WITH THE BENEFIT OF HINDSIGHT. IN ADDITION, HYPOTHETICAL TRADING DOES NOT INVOLVE FINANCIAL RISK, AND NO HYPOTHETICAL TRADING RECORD CAN COMPLETELY ACCOUNT FOR THE IMPACT OF FINANCIAL RISK IN ACTUAL TRADING. FOR EXAMPLE, THE ABILITY TO WITHSTAND LOSSES OR ADHERE TO A PARTICULAR TRADING PROGRAM IN SPITE OF TRADING LOSSES ARE MATERIAL POINTS WHICH CAN ALSO ADVERSELY AFFECT ACTUAL TRADING RESULTS. THERE ARE NUMEROUS OTHER FACTORS RELATED TO THE MARKETS IN GENERAL OR TO THE IMPLEMENTATION OF ANY SPECIFIC TRADING PROGRAM WHICH CANNOT BE FULLY ACCOUNTED FOR IN THE PREPARATION OF HYPOTHETICAL PERFORMANCE RESULTS AND ALL OF WHICH CAN ADVERSELY AFFECT ACTUAL TRADING RESULTS.

With the exception of the statements posted from live accounts on Tradestation and/or Gain Capital, all results, graphs and claims made on this website and in any video blogs and/or newsletter emails are from the result of back-testing our algorithms during the dates indicated. These results are not from live accounts trading our algorithms. They are from hypothetical accounts which have limitations (see CFTC RULE 4.14 below and Hypothetical performance disclaimer above). Actual results do vary given that simulated results could under — or over — compensate the impact of certain market factors. Furthermore, our algorithms use back-testing to generate trade lists and reports which does have the benefit of hind-sight. While back-tested results might have spectacular returns, once slippage, commission and licensing fees are taken into account, actual returns will vary. Posted maximum draw downs are measured on a closing month to closing month basis. Furthermore, they are based on back-tested data (refer to limitations of back-testing below). Actual draw downs could exceed these levels when traded on live accounts.

CFTC RULE 4.41 - Hypothetical or simulated performance results have certain limitations. Unlike an actual performance record, simulated results do not represent actual trading. Also, since the trades have not been executed, the results may have under — or over — compensated for the impact, if any, of certain market factors, such as lack of liquidity. Simulated trading programs in general are also subject to the fact that they are designed with the benefit of hindsight. No representation is being made that any account will or is likely to achieve profit or losses similar to those shown.

Statements posted from our actual customers trading the algorithms (algos) include slippage and commission. Statements posted are not fully audited or verified and should be considered as customer testimonials. Individual results do vary. They are real statements from real people trading our algorithms on auto-pilot and as far as we know, do NOT include any discretionary trades. Tradelists posted on this site also include slippage and commission.

This strictly is for demonstration/educational purposes. AlgorithmicTrading.net does not make buy, sell or hold recommendations. Unique experiences and past performances do not guarantee future results. You should speak with your CTA or financial representative, broker dealer, or financial analyst to ensure that the software/strategy that you utilize is suitable for your investment profile before trading in a live brokerage account. All advice and/or suggestions given here are intended for running automated software in simulation mode only. Trading futures is not for everyone and does carry a high level of risk. AlgorithmicTrading.net, nor any of its principles, is NOT registered as an investment advisor. All advice given is impersonal and not tailored to any specific individual.

* Published percentage per month is based on back-tested results (see limitations on back-testing above) using the corresponding package. This includes reasonable slippage and commission. This does NOT include fees we charge for licensing the algorithms which varies based on account size. Refer to our license agreement for full risk disclosure.