Team, I know there are many threads on Algo trading but I needed to start a new one to ask 1 very simple question.
I am not looking for setups, I just need to know if anyone has had a successful trading algo, specifically with NT or Bloodhound Sharkindicators, that continuously makes money with auto trading. I am concerned if this is possible. I am not using backtesting because I know how inaccurate that can be. I do use my NT sim account with live data, not market replay.
I have had success with automated trading in NT. So yes, it is possible to be successful. I have a trading journal that documents some aspects of it, in case you are interested. If you want to know anything specific I would be glad to share,
Thank you Iantg. And thank you bughatti for this thread.
I am beginning to develop an automated system. I have technology background but I am not too savvy with C# hence I using a coder. I would like to get some insight from your experience on how some specifics such as - how to prevent from trading in choppy market. You have to get the system to be your eyes and feelings to gauge how the market is behaving and this I am finding a bit tricky. If I tighten up the logic then I prevent chops but I also miss good opportunities.
Anything you can share will be of immense value I think.
You can see choppy times using Ichimoku cloud.
I use daily cloud: if price is inside - then NO trade taken by automate.
Helps to prevent a lot of negative results.
As price gets out of the cloud some good directional trades are easy.
Though I don't use this approach myself because my style of trading uses chaos theory instead of more traditional methods, there are several great ways to filter the market. Here are a few good ideas:
1. Time filters: Everyone knows that the market tends to trend from the US open until 2 to 3 hours after, and then from there it may or may not continue to trend. I would say 9:30 am Eastern time to 11:00 am eastern time you have 75% to 90% trends. But beyond this you will likely have a 50% / 50% chance of sideways action. One of the oldest sayings among trend traders is that you make your money in the morning and you lose your money in the afternoon. In NinjaTrader you can set up a basic if statement to say if Time is between (X and Y) then trade. Otherwise you will stay out of the market.
2. ADX: This is the only indicator I know of that claims to tell when the market is trending or not. it does a decent job from what I have seen. There are a variety of settings to play with but if you used a larger time frame say 10 minutes or so, you could tell if the market was in an up trend or down trend and how strong it was. If the trend was below a specific level you could set your code to just not take trades, this would essentially filter out sideways action.
3. Finally my personal favorite for simplicity and overall accuracy is using a Range indicator with something like a simple moving average. This will give a picture of the overall movement of price in the last N number of bars / X minutes, etc. If you are looking to trade X number of times every Y number of ticks, or Z number of minutes then just set it up like this:
A: Initialize your value like this: double rangevalue = SMA(Range(), 30); This will show you the range of the last 30 bars, in this example we are assuming that you would trade if the range in the last 30 bars showed the market was moving and not stagnating.
B: Then call it in your code as part of your entry condition like this:
If(rangevalue > 1.75 && whatever other conditions you like here) This is basically saying if the range in the last 30 bars was greater than 1.75 ticks. In the ES for example this would be 4 ticks = 1 point. So 1.75 points = 7 ticks.
//enter long or short depending
What this will do essentially is allow you a way to only trade when the market is moving. I.E range of price is = 5 ticks, 10 ticks, you decide. If the range of price is only = to 5 ticks for example, and you are trying to catch a move of 10 ticks as your profit target but you have a stop loss of 5 ticks, you are 100% times more likely to get stopped out at a loss vs. catching the 10 tick profit target. Alternatively If the range is moving 10 - 20 ticks and your profit target is only 5 ticks than you have a much higher chance of hitting this. So knowing your odds of hitting your profit target relative to the market conditions can really help you. Depending on how you set your profit target relative to your stop loss, knowing the range is also paramount. If you are a trader that uses a profit target that is > than your stop loss, you should use one of the methods I described or you will get stopped out at a loss almost every time.
Hope this helps. Happy trading
Last edited by iantg; March 4th, 2016 at 10:54 AM.
The following 5 users say Thank You to iantg for this post:
I tend to lurk in the shadows, and when I do post, its usually in the hoity-toity 'elite' section, but I came across this question whilst browsing and thought I'd add my two cents. .
Automated trading is my life, I eat and breathe it, go to sleep and wake up thinking about it, and devote my entire day to it. . and this has been the case for quite a few years now.
Its one of the most difficult paths a trader can take, with a far lower success rate (in terms of those that make consistent money, relative to the total number *attempting* to make consistent money) than discretionary trading.
One of the main reasons for its difficulty involves the premise behind your question. . . namely, that its really impossible to know whether you've created a truly profitable, robust, consistent automated strategy, even if its producing profitable, robust, consistent and stable returns. The reason for this is that if you were to create 100 automated strategies completely at random, using entirely random nonsense entry/exit logic, approximately 45 to 50 of these are likely to be net-profitable, even over long stretches of time.
Because of this, one of the first rules any skilled systematic trader learns is that live-walk earnings are *not* the foremost predictor of an automated strategy's strength, and certainly isn't the end-all be-all. The judgment of an automated strategy must be based on quite a few factors, the majority of which are in its historical record (as well as the historical record of other instruments in *relation* to it. . hint hint). . and this is the 'secret sauce' comes into the picture, the things that most successful automated traders are less willing to share.
However, I'll say this. . yes, its certainly possible to create profitable automated strategies, and even to do so regularly and consistently. I have approximately 100 strategies that I'm currently running, 24/7 everyday, and the cumulative result is in the black most days/weeks/months. . . but I'll also add something even more important/meaningful, its also possible to create an objective means of judging/'scoring' your automated strategy creations, with enough knowledge and hard work (largely the same thing), to where you can measure automated strategies relative to one another, using a barometer with true predictive-measurement-ability. This, to me, is the end-all-be-all of automated trading. . figure out how to assign them a powerful and at least reasonably accurate 'likelihood of success' score, and you've officially made it. As soon as you find a 'just' means of scoring all of the various different aspects of any given automated strategy, and then boil this down to one *single* score, you then have a clear cut path forward, and the rest is easy by comparison.
Good luck. . I'm always open to questions, either posted here or via private msg.
The following 2 users say Thank You to Dionysus for this post:
You'd mentioned avoiding back testing because you felt it was bugged/flawed. . you'll want to reconsider this. . unfortunately its simply not possible to create an automated strategy you can 'trust', an automated strategy that is likely to have forward walk results somewhat similar to its past results, unless you use a reasonably sized data pool to evaluate it. . and in the realm of automated trading, a reasonable data pool is actually quite large (I'd advise going back to 2011, at least, if not 2009. . most data feeds available, which are relatively low cost, will provide data back to 2006 to 2009, for virtually every major instrument.