If you look at trade outcome strictly from the standpoint of success/failure there are just 2 outcomes. Assuming market is random that is 50% success rate. If you add price behavior during the trade, there are 3 possible behaviors: 1. price rotation above entry 2. price rotation below entry 3. price rotation around entry Based on the above there are 5 outcomes. Given a win rate of about 80%, it seems like every one of the above outcome is randomly distributed. If the profit factor is 1.2, where is the edge coming from most likely (if there is any)? These stats are net fees/commissions I understand the limited sampling size, but just wanted to get the discussion going to see if somebody can provide insight into this without really knowing the strategy.
%% WELL good thing for us, market is not random; trends + profits+ prove that. Actually can be defined in 2 ways, profit or loss.[Average trade duration = >2 hours] I proved that first year year\did far worse, far worse, than random\LOL
I'm not sure if you got this, but you chose to create a scenario where 4/5 times (or 80% of the time) your trade breaks even or is profitable. That is what is driving your result (selection bias). This not an edge. What you should do is go out in the real world and see how many times you are able to identify each pattern. And then you can run a test on that. E.g. your ability to spot pattern 1 may be 15%, pattern 2 5%, pattern 3 40%, pattern 4 10%, and pattern 5 25%. You'd still want to figure out if recognizing such patterns is profitable for your time horizon/trading. To do that, calculate the average subsequent return (e.g. 30 min pnl after you open a trade based upon the pattern signal). Then you can figure out how generally good you are by multiplying your ability to spot patterns by the average returns to give you a probabilistic profit on any general trade (and average profit per pattern -- you may find that you can make more money by not trading low edge stuff).
First off, I would argue that the top pic and middle pic in column two, based on your 3 behavior defines, are price rotation above -or- below entry. You have differentiated range from directional trend, but the behavior is the same price rotation above or below. Therefore the number of "outcomes" is 3, not 5. Also worth mention is that while you are looking at price activity, you are also including pattern recognition. So that's 2 additional variables, price and pattern. As for any edge, it comes from the inputs used to determine and trigger entry in the anticipated/expected direction.
Your entry price is not a factor. The market doesn't care about you. What you are showing are possible topping/bottoming/reversal/continuation patterns that matter based on the price action that preceded it. You already have the innate ability to perceive it. Don't over think it. Peel the onion further as there are many layers/contexts beyond the price chart.
First off the number of trades is like first couple of hours for short term traders, I use bare minimum of 10,000 and making sure over several years. There been systems I have designed had 100% win percentage but 150 trades is nothing to risk money. You need to break down each signal, break down when not to take trades, when to double up.
This is a simple beginners mistake. You only account for probability of success but not for outcome when losing vs outcome when winning. You confuse W/L ratio with profit factor. Market is random when it comes to where it goes but it's not random when you find trades where your risk is 80cts on the dollar. Finding these pockets is where the money is. I would not touch a system with a daily sharpe below 3 with a 10 foot pole
some information on the strategy: trading 1 pattern on the ES conditional stop and reverse scaling skewed risk to reward i'll get more trades in to get more meaningful ratios (sharpe, sortino, calmar..) . These are forward testing by the way, so it may take a while. Be right back lol
Your average trade duration is one hour, which tells me this is an intraday strategy which is very prone to failure given the extremely small sample size.
what are you guys' thoughts on extensive back-testing? I feel like what @Handle123 says makes sense from a sampling perspective, but what if you're getting on that strategy at the tip of market change that renders your strategy a loser right when you think you have significant evidence that it worked in the past? Unless of course you believe that a strategy can work forever without tweaks as time goes. If you believe in continuous improvement over time, do you constantly testing on 10k data points every tweak? Feels like you're always operating on a lag.