(By Michael Harris) As the title indicates, in this blog we are dealing with probabilities and not with any indubitable rules. In the markets, there are no such rules, at least not any that I know of, and only those who can properly calculate probabilities within the context of strict but also realistic risk management survive at the end. One approach involves identifying patterns in price action with a significant bias in either direction, long or short. But remember this does not suffice for profiting in the short-term but only in the longer-term.
The reason that even the identification of a directional bias is not enough to succeed at the end of the day is because even with the tossing of a coin that is biased to generate more heads than tails, a streak of losers just enough to produce a drawdown of any size has finite probability. Of course, as the required streak gets longer the probability becomes extremely small but it is always finite as I have already explained via the use of the Bernoulli equation in my book "Profitability and Systematic Trading", which is out of print now. My point in the book was that in the same way that there people lucky enough to win the lottery, there are also traders lucky enough to get ruined by a long streak of losers. The sooner a trader understands this, the longer the stay in the market will be. The solution to this problem is risking a small percentage of bankroll on each trade, around 0.5% to a maximum of 2%. But this requires that the trader is well-capitalized to make this activity worthwhile. I guess my book was not good for business…
The pattern
The pattern I am talking about refers to the closes of the last 6 days and it is enclosed in the shaded rectangle on the daily SPY chart above. The main statement of the pattern is that the most recent close, that of last Friday, was greater than the close of 5 days ago, not counting the last. (Actually, the close of Friday was also greater than the close of the last 6 days).
Pattern identification methodology
This pattern was identified by a Price Action Lab scan using non-adjusted SPY data since inception with 2% profit target and stop-loss. This is how the workspace was set up:
After running the workspace, the following reproducible result was obtained (sorted for highest win rate P):
The three patterns on the Price Action lab scan output above are similar and we will only deal with the first one with the highest win rate of 72.41%. One question is: is this a statistical fluke or a pattern with some predictive ability of timing long entries? One way of minimizing the probability that this patterns is random is via portfolio testing. This may be done with Price Action Lab by generating code and back-testing the pattern in some popular analysis platform but recently a new feature was added to the program that does exactly that with click of a mouse to save valuable analysis time. This is how the beta version of this feature looks like now (after some more improvements from last time) when the pattern is tested on 11 popular ETFs:
The portfolio backtest above shows that this pattern has been profitable in 9 out of 11 popular ETFs. The win rate of 58.69% and the profit factor of 1.56 of the portfolio backtest indicate a low probability that this is a random pattern but there is one thing to always remember: Probabilities say nothing about the next event like when tossing a coin you can get streaks of heads and tails. Thus, even if you overcome the randomness issue, the probability issue is still there. However, based on the above analysis, the expectation is positive. Specifically, the expectation of the first pattern (average trade) is equal to $0.3515 while the average loss is $1.5266 based on the portfolio backtest. In my opinion, these are not spectacular numbers but still way better than average.
Disclosure: no relevant position at the time of this post.