Randomness revisited: random levels?

AdamAdam Grimes's blogs, General Comments, Trading Theory14 Comments

This is a followup to the series of posts I already did on randomness in markets. I have received quite a few questions, both from blog readers and from traders I mentor, about specific kinds of levels. I have been trading a long time and have seen so many different ways of calculating support and resistance. I have studied a lot of these things in some depth… everything from the standard Gann stuff to “floor trader” pivots, to Drummond geometry trendlines, to Fibonacci levels, Hurst cycles, and then you have the “mathemagical secrets” like the Camarilla Equation (Google it if you’re curious, but don’t waste more than 2 minutes thinking about it.) I have also run into many traders who are passionate about the merits of one method or another, and their justification is always (pay attention here… this is important) that “of course these are good levels. look at price action around the level. look at this example. look at this one…”

I was one of those traders until one of my mentors dropped a real bomb on me. (He was a former floor trader in the grains with a strong mathematical background.) He gave me a series of levels and asked me to take a look at them and let him know what I thought. Assuming these were some kind of ratio/cycle method I spent a day digging into them and came back with multiple examples of how amazing the levels were. Turns out… you probably already guessed the punchline… they were completely randomly generated.  I was in complete disbelief until he invited me to draw some random lines on charts and repeat the analysis… then I was shocked and disillusioned as I saw these random levels looked just as good as my prized pivot levels I had generated.  Everything we think we see in our “real” levels we also find in random levels:  perfect tests to the tick, multiple tests, support failing and becoming resistance, gaps through the level that hold on a retest… the list goes on and on.  What is happening here is that price action is random enough that any defined level will set some expectations for the trader, and then our highly evolved pattern recognition abilities kick in… and, in the absence of real patterns, the brain is happy to create patterns for itself…. and so we find ourselves back at one of the fundamental issues in trading–the difficulty in distinguishing between actual pattern and random data.

Now, please be clear what I’m saying.  I absolutely am not saying that support and resistance are useless or that they are not a real concept.  They are and they are. ( I do use support and resistance in my trading, but they are usually pretty far down my checklist of what I consider to be important.)  What I am saying is that most of the levels you are using may not be as valuable as you think they are… even if they “look good” to you on charts, I would encourage you to honestly ask yourself if they are better than random levels… and you have no way to answer that question if you don’t have some experience looking at random levels as if they are real.

Please take a look at the short video below that I recorded.  It is not particularly well done or well planned, and I apologize for that, but I wanted to give you the first shot I did at recording this.  Otherwise, I could have recorded it 20 times and picked the one that looked the best!  I think the method I use to generate random levels is kind of interesting:  I basically hid the price bars on a daily chart, drew horizontal lines, and then looked at where those lines would have hit a 5 minute chart intraday.  (As you repeat this exercise, consider other ways to generate random levels on charts.)  As we go through the 5 minute chart, pause the video and look at how the day’s price action interacted with these “levels”… assume that you believed each of these red lines represented an extremely significant price level before the market got there.  In how many cases would the price action have supported that assumption?  Once you have watched this video and thought about it, I would invite you to repeat the exercise yourself.  Please don’t take anything I tell you at face value, but use it as a departure point for your own learning and analysis.  Draw random levels on charts and check out price action around them… then look at price action around the levels you are using (especially if you think you have great levels)… and then honestly ask yourself how much better your levels are than these random lines on charts.

Probably not one in a thousand traders is aware of issues like this.  If you know that you can draw any line on a chart and find it working as support and resistance, how would that change your approach to the markets?  How will it change how you use support and resistance?  How does it change how you evaluate your trading results around these levels?  I think these are profoundly important questions, and it’s worth some time to carefully consider the answers here at the end of another year.

Follow AdamG_SMB on Twitter

14 Comments on “Randomness revisited: random levels?”

  1. Thanks for the excellent video. I draw S/R levels on my chart but just to show me where the last significant stand of bulls or bears took place. I draw these levels on my main timeframe and their purpose is to indicate that something has changed, through them I am able to recognize that we might be forming a box and that a transitional period may be underway. Other then that I only use them to identify long term major levels that acted as major turning points.

  2. I can basically directly copy the previous comment for myself, plus I’ve always wandered how people can actually use these magical levels coming out of nowhere. It would be great though if you could comment on your use of previous support and resistance and the time decay you expect (i.e. in many ways an intra day level is possibly a real buyer/seller whereas a level on the dayly chart must mean something different if it is to have any relevance for the current price action. When would you consider a level obsolete if it has not been tested yet?
    This is of course assuming you do not suggest that chart derived support /resistance levels are also practically useless. That would really be disconcerting for me since they are along with trend recognition almost the only information I really take at face value from a chart.
    On a similar note I have the same question on trend lines: While they are useful for visualizing the trend I dont see any reason why the price increase/decrease needs to correlate linearly with time. Any thoughts on that?

  3. This is a good example of why Risk Management is the single most important factor to trading profitably. You can trade random lines (support/resistance) and still make money if you have a stop loss point and profit target.

    Where real money is made is when you find linearity and continuation in a chart.

  4. Hi Adam,

    As a relatively new trader- I took SMB remote trading course last year- I am totally confused and dis hearted to say the least.

    The key thing I got from the course ( it did helped to simplify trading for me) is that ability to identify S/R is the key and everything else is secondary. My tutor and all training videos I watched talked about s/r over and over again.

    After the course, I decided to trade /es and /cl only instated of stocks using s/r. I realize you said s/r are important but on your list of things to watch they are further down. What i am missing in terms of things which are on top of your list to watch for trading?

    Thanks.

    Tariq Malik,
    Hopkinton, MA

  5. I’ll share an interesting exercise that helped me understand randomness and edge in the market. I come from a background of being a professional gambler so developing edge is something I’m always thinking about.

    Stack two piles of poker chips and use a coin. Then choose heads or tails for this game. Whenever you flip a coin and win you get two chips. Whenever you lose you lose one chip. See how long it takes you to win the entire stack.

    This helped me learn about randomness, edge, patience and grinding out a profit. Hope it works for some of the readers.

  6. That would certainly be an overwhelming statistical edge… enough to be completely unrealistic in terms of anything I have ever experienced as a trader… but the core idea is very similar to some things I use in teaching and thinking about markets. A better way might be to structure the game so you lose 10 if you’re wrong and get 11 if you’re right. Then the questions start to get interesting… you have an advantage, but how long does it take you to get the whole stack? How big does your starting stack have to be to be fairly sure you will not run out of chips (go broke) before you get the other stack? What kind of drawdown can you expect? These are interesting questions for systems traders and a good place to start is with the Kelly Criterion paper if you don’t mind crunching through some math.

  7. I see two questions here. First is related to the validity of using S/R in trading. I absolutely do use them, but my point is that I think most traders have a kind of naive approach to levels so they are trading something that, in the end, is fairly random. The question you really need to think about is how can you be sure your levels you are using are really better than random lines on charts? To answer that question, you first need to look at random lines on charts. Very few people have done this and nearly everyone is shocked to find that random lines on charts look and behave so much like levels. Be clear that this was my intent in writing this post and I think it’s a difficult question you must address.

    As for what my list is… I have already shared a lot of that in some of my other blog posts but I’ll write another with a few thoughts along those lines.

    Some of this comes down to personal preference / style… there is ultimately no right answer to a lot of these questions but I think it is very very important to be sure you are trading with something that actually has an edge in the market. Very few traders do this work.

  8. Chart derived S/R levels can be very useful but you have to make sure you have truly important levels. In my opinion, most trading days most stocks do not engage truly significant levels where you can have a high expectation of non-random price action. However… when they do… THAT’S really time to wake up and pay attention!

    Trendlines… I also use them… I think even longer-term trendlines can be useful because they show what the market has accepted as a growth rate for a stock/industry/market… I think there is even some validity to trendlines on log scaled charts because most things in finance are best represented as continuously compounded (log) values. (Ignore that last if it doesn’t make sense to you.)… but, again, HOW you use them is the key question I think.

  9. Very interesting as I believe one should spend as much time on perfecting risk management as they do on learning setups.

  10. Read Van Tharp’s “Trade Your Way To Financial Freedom” for more insight on randomness, and how one should properly design a trading system for positive expectancy.

  11. Just for the sake of being the devil´s advocate I would propose:
    -1. Support and Resistance levels exist and are significant.
    -2 Randomly generated lines will sometimes coincide with the above.
    -3 Number 2 does not invalidate the validity of number 1 above.

    Yours are great posts. They make me think and are very usefull. Thkxs for takinf the time.
    João Pedro Oliveira

  12. I, too, was one of those traders who would look at fibonacci or Camarilla pivots and point out how amazingly they “exposed” support and resistance until one day few years ago when I was taking a nap on my lunch in the parking structure at work.

    This parking structure doesn’t have concrete walls on its perimeter, it has 6 or 7 thick steel cables running horizontally from ground level to about chest height, maybe 10 inches apart. So, there I was, reclined in my seat looking out the windshield through the cables to a line of trees a couple of hundred feet away. I noticed that many of the tops or branches of these trees stopped almost exactly at the heights of these cables–it looked exactly like support and resistance on a chart. I’ll never forget the realization that came over me. From that point I promised myself that I would have to present evidence, or at least a very good argument, that what my mind was seeing as support or resistance was the real deal.

    That said, I think there are two important points to be considered here (more, actually, but I’m just gonna make two): 1. support and resistance are generally only real because traders believe they are–there is no magic in, say, fibonacci retracements other than the magic we ascribe to them. If we believe in them then we trade as if we believe in them, which fulfills the belief. 2. It’s important to run the numbers to validate what you believe to be true–for example, Dr. Brett Steenbarger calculated the probability of price exceeding R1, S1, R2, S2, R3, S3 and the previous day’s high and low, and this, along with the other data he gathered, allowed him to make informed decisions rather than just trusting a line drawn on a chart.

    And, BTW, you got it exactly right, @flowtastical–the perpetually non-sexy risk management (position sizing, stop loss, profit target, etc.) is absolutely the most important and most often overlooked success factor in trading.

    Mark Bray
    Cypress, CA

  13. Here is something ive thought about for some time:

    Say we are exposed to a chart for the first time ever.
    We have no idea, nor have ever heard of “S/R levels, fibs, trendlines, pivots, and all the other funky stuff”.
    Now imagine we only have to variables:

    Price and Time.

    We also know that markets, bar per bar, never change and never will change.
    They all have HI/LO – Open/Close. This NEVER changes…

    So we could say that the only aspect that changes is the volatility, bar per bar.

    Based on this assumption, could we build a solid method that exploits volatility change based on the time/price variable?

Leave a Reply