Welcome to NexusFi: the best trading community on the planet, with over 150,000 members Sign Up Now for Free
Genuine reviews from real traders, not fake reviews from stealth vendors
Quality education from leading professional traders
We are a friendly, helpful, and positive community
We do not tolerate rude behavior, trolling, or vendors advertising in posts
We are here to help, just let us know what you need
You'll need to register in order to view the content of the threads and start contributing to our community. It's free for basic access, or support us by becoming an Elite Member -- see if you qualify for a discount below.
-- Big Mike, Site Administrator
(If you already have an account, login at the top of the page)
I am curious what people out there think. If this was your trading system, would you still be trading it?
Please answer a 1 question survey here: https://www.surveymonkey.com/s/XS6X7DN
(If this is against forum rules, I apologize and ask Big mike to delete the post).
Here are some performance curves to help you make a decision. For the sake of the survey, assume you started trading with real money on the green portion of the curve.
Good question. Up to now, I haven't addressed this at all. The answer for me involves the "next best alternative."
Every 6 months or so, I look at every system I am trading, and also the systems waiting in the wings (ready to be traded live, but currently not.) If I find a better system than the one I am trading, and I don't have enough capital to trade both (or perhaps because of correlation issues I do not want to trade both), I will replace it. Thus, even if NGEC is performing decently (making money as opposed to its current situation of modestly losing), I still might replace it with the "next best alternative."
I only do this exercise 2x a year in part because it is involved (correlation studies, etc), but mainly because it is only fair to a new system going live to have some time to prove itself. Most people don't have the patience to do this, and they jump from system to system, never giving any one of them a fair chance. It would almost like pulling a pitcher in baseball as soon as he gives up just one hit.
Of course, even with this analysis, the quitting point is still in effect. If I hit that, I am out, regardless of the next best alternative (which may be cash).
One of the tricks unscrupulous vendors play is to assume all limit orders are filled as soon as the price is touched. You recognize this by looking at a price charts of their trades. If the method shows trades being bought at the exact low of a bar, and/or sold at the exact high of a bar, you can bet this game is being played.
Of course, the reality is that it is hard to buy the low and sell the high. My experience is that, depending on the market and when your order is placed, you can probably do this 20% of the time. The other 80% of the time, price has to trade through your price to get you a fill at the limit price.
This can be an issue with backtesting. If your backtest engine assumes limit orders are filled when touched, the results will be too optimistic. If the backtest engine assumes price must be penetrated to get a fill, then the backtest results will be a bit too pessimistic. I always go with the pessimistic approach. My actual results can then only be better than the backtest.
I bring this up because today my NGEC day strategy bought the exact low of the exact on a limit order. Of course, part of the reason I was filled was because I am trading a one lot - if I was trading a ten lot, I probably would have received only a partial fill.
The interesting thing is that when I refreshed the chart, the trade went away, according to the strategy engine. Since the price did not go one tick below my limit price, the strategy engine assumes there was no trade. But my real account says there was a trade, because I was indeed filled. A nice surprise for once!
So, this is roughly a $400 trade in my real account's favor, when compared to the backtest engine. This makes up for the order problem that cost me $500 a few days ago.
I dont know if its right for me to chime in. But I see this topic on quitting this system all the time on this tread. IMHO the biggest take away here is doing 2 things. Pre-defining your plan, in this case specifically a quitting point before activation of a system. Then having the perseverance to stick with it. Most people here on the thread keep saying the same thing, you should quit, when will you quit, why are you still holding out? For me that was answered at the start of this entire exercise post#10 in fact.
I think what @kevinkdog is doing is absolutely correct. But it also shows the difference from highly seasoned algorithmic traders vs say the "rest of us". Most of the people wouldn't have had the stomach to cope through this 10 weeks of misery, even if they had the pre set quitting point. IMHO everyone who would have quit before the pre defined criteria, changed the criteria (even with some pretty justifiable reasons and maths) or failed to follow the pre set plan in anyway. Made a huge mistake. That's where the failure is, in the trader not following the plan. Not in any single trade, or system being profitable or not profitable.
Thanks Trey, I think you hit the nail on the head!
I can tell you there is nothing more deflating than having high hopes for a trading system, only to see it not immediately deliver. I sometimes wonder if this only happens to me - am I the only one who experiences drawdowns, almost without fail, right after I start a new system?
I think a lot of this impatience is because of the age we live in. Nobody wants to wait for anything anymore. I see this a lot with trading systems. I've heard of people abandoning trading systems after 4 or 5 trades. Seriously?
Please weigh in on this topic: Answer this one question survey, and I'll share results with everyone:
I don't believe (in my case) that patience is the issue. It's that the system is not conforming to the original idea. I do understand you set initial stops and quitting points, which is a "must", and you want to see it through. But I struggle with the way the system is performing today vs the way it performed in the tests, to me they are not the same. I would have made that part of my "quitting" points.
However, if the system was performing the same, then you make an excellent point. Most people cannot sit on their hands very well or resist the urge to make changes/tweaks to the system, which will make analysis of it near impossible. You can go back to Richard Dennis's quote about how he could post his trade signals in the WSJ, and most people still wouldn't make money on them because they would try to "improve" the calls or they have different emotional capital than Richard, and therefore would wind up with drastically different trades and results.
Nope I have this happen to. My live algo trading journal shows the same thing. Even worse my own errors compounded the problem. To be honest it does feel better to know that even you experience it. Misery loves company
That's the $100,000 question: is the system performing the same as it did in historical tests, or did something fundamentally change?
Trying to be objective, here is what I currently see:
* Performance after 26 days of trading is not even close to the average expected performance.
* System is currently performing at about the 15th percentile. If it was average, it would be at the 50% percentile. If it completely fell apart, it would be at the 0 percentile.
* Looking at the equity curve, there are 7 similar flat/down periods in the equity, of roughly the same shape and length.
* There have not been any big winning trades, which is what this system needs to succeed long term. On average, I should see 4-6 large winners per year, or one very 2-3 months.
Based on all this, I am not convinced that the system has stopped working. But, I am open to hearing everyone's thoughts on the matter, especially if you can conclude that from the numbers (I can send you whatever data you need). Maybe there is a measure that says "stop! It is broken!" that I am unaware of-- if so, I'd like to understand it. For example, I might play around with a 2 sample T test, and see what that says.
I entered the historical data, and the live data, into the T test calculator here: Student's t-test: Results
Results are here:
If the result was 0, I believe that indicates that the 2 samples (historical backtest and live testing) are definitely different (ie, the system has changed). If the result was 1, then the 2 samples are definitely the same.
Just for kicks, I took some random groups of 25 historical trades, and put them as sample #2. These are obviously from the historical sample, but the null hypothesis result varied from .09 to 0.82, with a mean of 0.51.
Based on these results, and assuming I did this right and that the analysis is appropriate in the first place, it looks like the two samples are leaning towards being different, but it is not conclusive.
One other interesting tidbit: If I add one $1000 winning trade to my actual results, the null hypothesis test jumps from 0.26 to 0.47. This tells me that the one big winning trade I have been waiting for will move actual results much closer to backtest (which I kind of knew anyhow)...