Mike, thanks for the lead to the thread. Here is a chart, with four completely random lines drawn blindly on a three minute chart of the TF 6-15 contract.
Look at the lines as they could serve as S/R . Pretty amazing to me that we could actually take these random lines and apply them as trading tools. If a person did not know they were drawn randomly, one would suspect the lines would be used as some important turning points in the market. I have more charts, which I experimented with, which are just as astounding. I just felt posting one is enough to prove the point.
Many thanks to Adam for a great presentation (I'm your new greatest fan!) and to Big Mike for organizing it!
The stuff about markets being random most of the time was very well presented in terms of the balance between Momentum and Reversion to the Mean, and has caused me to take a break and reflect on what I am doing - trading "noise", effectively..
He has motivated me to embark on a quantitative study of my trading signals using his Event Study Methodology.... however the devil is in the details as usual..
I was wondering if anyone who watched (or maybe even the Guru himself!) could enlighten me as to how to deal with the situation when the cumulative baseline returns are NEGATIVE for following bars N+1, N+2..
We see this in bear markets such as CL lately, also GC. It does not make sense to subtract a negative baseline return, so I am thinking that if baseline is less than zero, set it to zero (basically ignore it).
Also, the baseline values I am getting further out (at N+2, N+3 ...) are simply multiples of the first value (N+1), or close enough.
i.e. Value(N+x) = Value(N+1) * x (approximately)
I have looked at daily data (1000 days) for CL,ES,GC so far and I always see the same pattern emerge, but it seems kind of artificial. Anyone else seeing the same thing, or have I got my code logic wrong? (latter is the most likely possibility!)
This is one way to think about it, don't know if Adam would do the same: in the case you mention, crude, you would have been short-only if you had a position over the last year or so, so that would be your baseline drift--from the perspective of what you could have earned being in the market and short. That is one way to look at the baseline drift (and hard to beat in crude, with such a precipitous decline)--it has been south, and heavily so, and hard to beat the short-and-hold in this case.
The following user says Thank You to josh for this post:
Thanks for your input. I see where you are coming from with the contextual nature of the baseline drift. Also, I think part of the issue is that my 1000 calendar day dataset for daily CL bars is insufficient as it contains no real bull market such as the pre-2007 ramp up in prices. I will have to hunt around for some more data as Kinetick wont go back further than 2009 for daily data.
In any case, the goal is to get some stats for intra-day bars (Minute or Volume) since trading CL off a daily chart is beyond my means..
EDIT: I have just managed to get back to 2007 by using 1440 Minute bars to replicate Daily bars. The baseline is very definitely positive over 2070 days data!
Last edited by ourssolaire; March 31st, 2015 at 07:55 PM.
Reason: Added information
The following user says Thank You to ourssolaire for this post:
Yes, but the baseline drift is to measure against some alternative to buy-and-hold, or sell-and-hold as the case may be. The purpose (again, as I understand it, Adam may differ in his thinking process) is to say, "how well would I have done trading this on a smaller time frame than X days, versus a simple buy/sell-and-hold approach over the same X days?"
Thinking of it another way: a buy-and-hold would have lost 50% from a year ago in crude. But a methodology that merely loses 20% is not really an acceptable strategy either. Neither would it be really accurate to look at a 10-year period in a stock where the stock went from $50 to $100, and back down to $50, and say the drift is zero. Compared to this, any method which does not lose money over that 10 year period would be above the "baseline drift," but that's not really the baseline drift--the drift was up 100%, and back down 50%, to where it started. So, the point is to look at a particular "leg" of movement, something that is a reasonable buy/sell-and-hold period, where an investor would essentially do nothing, versus an active approach, doing something, and comparing whether there is an advantage to doing something, versus nothing.
An example is FB--it was in a pretty tight range from Aug 2014 through Feb 2015, essentially eeking out a few percentage points gain. When looking at that period, the question is, does my methodology produce more than a few pct points gain, again, looking at that period? That period will be different from looking at the drift from July 2013 to March 2014, where it almost tripled in value--did the methodology more than triple the value over that period? If not, it did no better than the baseline drift, though perhaps it produced superior returns than it did over the range mentioned earlier.
Last edited by josh; March 31st, 2015 at 11:05 PM.
The webinar was truly brilliant! I started following Adam's work when Big Mike had posted some of his links in the random line thread. Glad the webinar followed shortly thereafter. Thanks to all involved.
After watching the webinar and the first half of the course, I have started to read Adam's book. Much more in-depth than the videos, though it's covering the same topics. Despite what he says, I don't see it as an advanced book. It's very accessible if you know the basics of trading.