Trading Lessons/Insights From Coin Flipping

Quote from tim888:

I don't think this is the important issue here. I think that Mike805 is trying to demostrate that the performance of some money managers and traders can be attributed to pure luck and not to some edge. Maybe it would be better for him to explain what motivated his test and I apologize if he has already done that.
More...

I'm glad these results have gotten people to ask a variety of questions, that was the motivation for me. I have answers to these questions that I'm comfortable with, but, I encourage people to formulate their own conclusions.

One thing before we continue:

Intradaybill, I appreciate you taking me off of ignore. Unfortunately you are really bad at picking your battles. You see, master_jaz is being constructive. Goochgobbler and you, well, you two kinda belong together. I wonder if somebody out there makes a couples "douche kit"? You know, for douche buddies... I'll ask around.

Anyway, back to the fun stuff.

The point of my post is what many here likely already know. I can "spin" a new set of results from the random() function by pressing "simulate" on my backtester. After a 100 "spins", I'll get a bunch of equity curves that are profitable, a bunch that are negative, a bunch that are breakeven. In the case I posted, I spun 3 times and voila, a "profitable" equity curve arose...

What does this mean? Well, my personal conclusion is that randomness plays a large role in what we do, and, many traders' success can be attributed to a lucky streak similar to that of a coin flip. The difference is always difficult to pin-point, but, there are ways to identify if a traders' results are random, and, I can get into that later...

My personal belief system is in the conceptual idea behind the trade. The conceptual idea must pass a very stringent set of tests before I run with it. Among those tests is body of academic research, the "common sense" factor, and perhaps most importantly, the idea has to show signs of profitability from the moment one codes it up for testing, i.e. no working a set of variables and doctoring a good looking equity curve (that's not to say I don't use optimization).

That last paragraph likely went a bit further than I was intending, so, lets get back to that pesky 12:55 PM entry time. As another poster mentioned, the system goes long/short only at that time if the random() value meets a criteria, meaning it can hold for a while... this was done on 5min ES data, so the number of bars is the ES day to day session +/- a few bars...

Why did I choose ES and why did I choose 12:55? Well, the product really doesn't matter that much. The time of day, as I will show later, has some interesting effects...

Anyone care to add in a "non-random" condition to the rules to see what happens?

Also, anyone want to mess around with the time of day? Any ideas?
 
Quote from shortie:

random # generator with 50/50 chance of the outcomes = coin flip

one caveat to Mike's code is that he uses a pseudo random # generator and that may have some caveats. but my limited understanding is that for most practical purposes it can be considered truly random. certainly in this case it should not be an issue.
More...

Yes, this is a caveat as there are no true random # generators out there. But, for qualitative purposes, something simple will work. As you mention, in this case, the point is qualitative, truly random versus pseudo-random won't really make a difference.
 
Quote from Mike805:

...

Why did I choose ES and why did I choose 12:55? Well, the product really doesn't matter that much. The time of day, as I will show later, has some interesting effects...

Anyone care to add in a "non-random" condition to the rules to see what happens?

Also, anyone want to mess around with the time of day? Any ideas?
More...

how would you test the outcomes under various conditions? picking better looking equity curves won't work :)
 
Quote from shortie:

how would you test the outcomes under various conditions? picking better looking equity curves won't work :)
More...

Very good question... I have a home brewed way of looking at outcomes, but, it would be neat to hear some ideas...
 
Quote from Mike805:

...I have a home brewed way...
More...

Should I assume that it is excessive use of booze that has caused the damage then?

You generated some random numbers and you selected some results with a slight winning bias. That does not prove anything Mike. Nothing at all. It is not enough to make any sort of inductions like you keep doing without any reservations. Rule-based trading is not random trading although the results may look similar to random entries. Why is it so difficult for you to understand that but instead you resort to vulgarity? What is your problem exactly?

Some guy plans to go to a bar to meet a friend and some other guy gets there by chance and meets the same friend, just by driving around, no plans. What is your point Mike? These are two different actions. One involves planning and the other random actions. The results are similar. You cannot compare these two actions. If your trading system does not perform or breaks even, it is not randomness maybe but the rules are such that it breaks even.

If you do not get the point yet, try generating a curve with a profit factor > 1.6 with your original random system. How many of those do you get? Maybe good systems are those which are as rare as high performance random systems. Try to refresh your brain Mike. You are stuck like a 7 year old who can add 1 to 3 but not 2 to 5.
 
Quote from intradaybill:

Should I assume that it is excessive use of booze that has caused the damage then?

You generated some random numbers and you selected some results with a slight winning bias. That does not prove anything Mike. Nothing at all. It is not enough to make any sort of inductions like you keep doing without any reservations. Rule-based trading is not random trading although the results may look similar to random entries. Why is it so difficult for you to understand that but instead you resort to vulgarity? What is your problem exactly?

Some guy plans to go to a bar to meet a friend and some other guy gets there by chance and meets the same friend, just by driving around, no plans. What is your point Mike? These are two different actions. One involves planning and the other random actions. The results are similar. You cannot compare these two actions. If your trading system does not perform or breaks even, it is not randomness maybe but the rules are such that it breaks even.

If you do not get the point yet, try generating a curve with a profit factor > 1.6 with your original random system. How many of those do you get? Maybe good systems are those which are as rare as high performance random systems. Try to refresh your brain Mike. You are stuck like a 7 year old who can add 1 to 3 but not 2 to 5.
More...

Intraday,
We don't have a beef so maybe I can moderate...maybe...so you said, and I quote "Rule-based trading is not random trading although the results may look similar to random entries. " Now, the question that Mike and I are begging to answer, can you PROVE the results are not random??? That is all we are asking...
I get that this "thought experiment" is a bit chicken-and-egg, but the fact remains that I, personally and without hesitation, am confident in saying I have been deceived by profitable-looking results that I believe are now random (or part of a random distribution--key difference, see..).

So, please help us, by showing how you determine a random equity curve from a non-random equity curve. That would prove far more productive than attacking Mike...

Masterjaz
 
Quote from shortie:

how would you test the outcomes under various conditions? picking better looking equity curves won't work :)
More...

shortie,

Like Mike, I also use a home-brewed approach via Excel, but there is also an on-line calculator. I would suggest that everyone build their own simulator before using the on-line one such that they realize the variables going into the calculations and what the results truely represent. For the online, go to: http://www.hquotes.com/tradehard/simulator.html


I would also add, that I prefer to use avg win/avg loss rather than total win/total loss for the win ratio as it better filters outliers and is just related by the number of trades n.


Now, this is not the only tool I use, but it provides a basis for looking at possible equity curves using basic trading statistics GIVEN those statistics are liable to change with changing market dynamics...


Keep thinking, we are all getting somewhere with this exercise (well, most of us...)

masterjaz
 
Quote from intradaybill:

Should I assume that it is excessive use of booze that has caused the damage then?

You generated some random numbers and you selected some results with a slight winning bias. That does not prove anything Mike. Nothing at all. It is not enough to make any sort of inductions like you keep doing without any reservations. Rule-based trading is not random trading although the results may look similar to random entries. Why is it so difficult for you to understand that but instead you resort to vulgarity? What is your problem exactly?

Some guy plans to go to a bar to meet a friend and some other guy gets there by chance and meets the same friend, just by driving around, no plans. What is your point Mike? These are two different actions. One involves planning and the other random actions. The results are similar. You cannot compare these two actions. If your trading system does not perform or breaks even, it is not randomness maybe but the rules are such that it breaks even.

If you do not get the point yet, try generating a curve with a profit factor > 1.6 with your original random system. How many of those do you get? Maybe good systems are those which are as rare as high performance random systems. Try to refresh your brain Mike. You are stuck like a 7 year old who can add 1 to 3 but not 2 to 5.
More...

intradaybill,

Since you've responded with something somewhat relevant, I've decided to address you by your ET handle rather than your temporary handle: intraday-douche.

You have this amazing ability to use soooo many words to say very little. People with that ability bore the crap out me. If you can't make your point concisely and clearly, TRY AGAIN!

Ok, Where did I claim my results are anything profound? They're not... Inductions? Huh?

Got the bar analogy... seems trivial to me... like your end goal is more important than the process. That's BS. Process is everything. I guess that's why you dislike my "inductions", you're not very good at the process.

That said, *I think* your point is differentiating random from rule-based. Yep... got it. Been there since page 1. It's Kinda the title of the thread. Coin-flip -> Random... trading ?= random... seems to be what we're discussing here... yep.. (sarcasm here)

I think this statement should be your next *gift* to trading: "Maybe good systems are those which are as rare as high performance random systems". Stunningly profund bill. Stunningly... I'm still in awe... the light emanating from your brilliant post is almost too much to handle.

Hmmm... what constitutes a "high performance random system" bill? Would "rules" be involved?

Geez, you're a tough one bill. Get over yourself.
 
Quote from tim888:

I don't think this is the important issue here. I think that Mike805 is trying to demostrate that the performance of some money managers and traders can be attributed to pure luck and not to some edge. Maybe it would be better for him to explain what motivated his test and I apologize if he has already done that.
More...

BINGO!!! That is exactly what I, and only I (maybe mike...), am trying to suggest. What is random and what is an 'edge' and HOW do you determine between the two??? Are trading statistics sufficient, equity curves? How many trades are representative 20, 100, 200, 1000? How does one demonstrate an edge and how does one adapt to market changes? All in light of random results...That's what I'm trying to determine and apply to my trading systems...

Masterjaz
 
Quote from masterjaz_99:

How does one demonstrate an edge and how does one adapt to market changes? All in light of random results...That's what I'm trying to determine and apply to my trading systems...

Masterjaz
More...
Isn't that impossible? How will you be able to know if your edge is not due to randomness?
 
You must log in or register to reply here.
Back
Top