collaborations

hi All,

I have been working together with a friend of mine to build strategies, although over the last year, it's been mostly a one way street, where I would implement and back test almost everything while my friend would forward test the strategies after I send him the executable to run. Recently, I have been becoming fed up with his attitude and am thinking it's time to move on and do this on my own. The reason is this: he is fundamentally a non-analytical person. He does everything by gut feel, although he never admits it. I am a hyper analytical person requiring a lot of statistics and back testing before I can start to believe in a profitable set up. So far, I have been able to find a few strategies that are profitable for some finite period of time, but then becomes unprofitable. I can accept this, and I am willing to use these strategies while the opportunities last and trust that I will stop using the strategy after it stops working (e.g., after a 10 or 15% draw down, I might stop live trading and go to paper for a while).

What drives me crazy is this: I am working on a strategy and find out a few problems with it so I test a bunch of hypotheses to try to correct this problem. I would run many different tweaks until I find a way to reduce the draw down or improve the win rate, or improve the mean profit and reduce the mean loss.

But then my friend would then come in and propose something completely out of left field, a completely different strategy that has nothing to do with the problem at hand. When I ask why he believes this will work, then he says he looked at the charts and "knows" that it works. So I go ahead waste a week to test it out, and invariably it doesn't work consistently. There may be one or two days that it works, but then there are 30 or 50 days that it absolutely doesn't work and it makes things absolutely horrible, losing 10s of thousands in a matter of months. So I tell him this, and his response is: I don't trust the back tests. The past is not a reflection of the future. I "know" that this works now because it worked yesterday and today.

So I try to tell him that although the past is not a guarantee of future behavior, it is our best guess and therefore we must use this information to guide our investments. But he doesn't budge. He has this favorite saying that the markets should not be approached as a science.

Perhaps he has his own way of doing things, but I simply cannot work with a person who refuses to accept statistics. We get into huge arguments because of this, and he always ends up making anti-scientific statements, although he never admits that it is illogical to make such statements.

So my question to the board: what should I do? Should I bear with this aggravation and keep working with him or just tell him it's over and I will develop whatever I think is right on my own. I know that we will keep getting into arguments, and sometimes I wonder if there is any benefit I am getting out of this collaboration. A key fact I have so far left out, though, is that he is the one who instigated this whole project by approaching me with this idea of developing an ATS, but it's just that I can program better so I ended up implementing everything...

 

If you know how trading works and how strategies are built, do it yourself. I'd never put up with someone like that, and I also believe in statistics in back-testing very much. He just sounds like a waste of time. Creative people aren't realistic. They come up with good ideas and ways to think outside of the box, but when it comes to practical implementation, in my experience - forget about them. Interpret his ideas the way he doesn't want you to interpret them and create your own idea from it. Otherwise, you're better off on your own - if you can think up enough ideas on your own (ie know something about trading).

 

Your friend sounds like a mess, and is definitely a waste of your limited time. I had one of those time-sucking vampires waste a good portion of my life in high school and I regret my decision to not distance myself to this day.

Your friend wants an ATS but he doesn't trust the science? Sounds like someone in need of psychotherapy. He may not know what he wants or he may just thrive on the conflict he generates with you. Whatever, it's taking an unnecessary toll on you. Just sever the relationship as amicably as you can and get your life back. You won't regret it in the long run.

 

Thanks for the reply. Well, yes I guess I wasn't clear when I said he approached me to develop an ATS -- it's not that he had specific ideas on how to do it, just that he wanted to build something, anything, whereas I wasn't even thinking about trading at that time. In the beginning, we brainstormed and came up with ideas and I implemented them but after about 3 months of failures, I started taking control and building purely my own ideas after a lot of statistical analysis of historical data -- that's when we finally arrived at a handful of profitable strategies. so yes, over the last year, I think I have gotten a pretty good feel for how to find a profitable strategy, at least in the context of the SPX movement during the day. More importantly, I built up a lot of tools and infrastructure to design and evaluate strategies, both with back testing and forward testing. This has greatly accelerated my understanding of the algorithms in the last few months.

Also, I initially fell into the trap of trying to find the holy grail. But after stumbling upon these handful of strategies that work for a period of time (and work fantastically over that period), I have realized that trading is like any other business. There is always a window of opportunity that you must seize while you can to make profits. Any business is not a sure bet at all times. There are ups and downs, and there are years with losses, as well as years with phenomenal gains. I think I have learned that trading should be approached the same way as a business. If you see an undeniable opportunity, you must seize it while it is available. If your business is not working anymore, you just have to step back and evaluate and then re-launch.

 

I used to try to put it to him gently and mildly that it doesn't make any sense to not believe in an objective, statistical approach to finding a strategy. But just recently, he completely blew my fuse and I told him how he is no different than a dumb Las Vegas junky with delusions of grandeur -- that he thinks that he can 'transcend' the statistics of the past is utterly ridiculous and a pure waste of my time. I feel bad because I have known the guy for years, but before we worked together, I did not have the slightest clue that he was a completely superstitious mystical fanatic who didn't believe the universe is governed by a set of immutable physical laws but rather is only explainable using our gut instincts alone.

 

Sounds like curve-fitting and, I am sorry to say, if that's the case you have limited experience with system development.

You see, in my view you are wrong and your friend may be correct. You never use backtests to guide investments. Backtests are used to rule out investments. Huge difference there between the two

I was also the victim of the backtesting hype a few years ago. Read this article about backtesting and since you say you understand statistics it may help you and save you a lot of time and money. Thank the author, not me, he is a great guy with some real experience in this area.

 

Yes I am well aware of overfitting. In fact I have a Ph.D. in engineering and I am a professional industrial scientist with over 15 years of experience in computational modeling so I am well aware of this issue. That is why I never optimize blindly to maximize PL, but rather try to find an algorithm that relies on as little parameters as possible, and has a physical basis or mechanism that is likely to make it work. Having said that, it is impossible to prove the statistical significance of a PL trend if it's positive. I agree, it is more of a process of elimination than of discovery so thanks for clarifying. In fact, after testing literally hundreds of strategies, I can immediately say that money-losing strategies are very clear -- they lose money consistently. And it is rare to find day-trading strategies that actually don't lose money consistently -- and even usually those that don't lose consistently are not profitable consistently given enough years of testing of data. However, I have seen strategies that are profitable within certain regimes that correlate with other market factors, such as when the VIX is elevated, or when the market is generally bullish, or bearish, etc. Those trends appear to be significant -- again, I cannot prove it but I am willing to call them potentially profitable opportunities.

 

I'll take the word of a rich, successful trader over some no-name blogger anytime:

"I know of no way to validate conjectures concerning technical trading without backtesting"
-- William Eckhardt

 

i have heard somewhere:

if you put two traders together, you get the worst of each other.

Every trader must follow his own light.

Think about it. That is so true.

 

Trading is all about backtesting working strategies and convert them to real traded strategies, so that they become historical trades as well.

Its just adding real made trades to your historical trades statistic, wheater you have only backtested them or traded them.

If your friend says he doesnt believe in backtesting he is on the wrong path.

The only different thing is, to know how the markets look, before the trade starts and how after it is made. Because in backtesting you only can see the after.