Just the opposite. I began with pages and pages of rules and subrules and exceptions and OTOHs. Now I'm down to a simple strategy statement, a simple tactic statement, and four rules, all on less than one side of an 8.5x11 sheet of paper. One indicator, no MAs, no BBs, no envelopes, no volume. It's been my experience that most beginners think the answers lie in indicators and patterns. It takes them a long time - though they often never get it - to understand that pattern and indicators begin with price, and if you don't understand price, particularly trend, support, and resistance, then it doesn't really matter what patterns you look for or what indicators you use or how you tweak them, you'll most likely fail, eventually. --Db
One pitfall with trading all day is that you often wind up looking for trades that aren't really there. And mid-day trades are so often a wash. Skipping them doesn't necessarily mean a loss. To the contrary, commission costs are less. --Db
I've also give up on indicators for intraday trading. The false signals generated can take your attention off the most important indicator of all: price action. I tested stochastic, RSI, different, MACD, simple MA's and it all comes back to price and risk management. I've seen some of those public charts at stockcharts.com and some of them are so strewn with indicators you need to squint to see the damn candlesticks.
I totally hear ya. By the time I'm ready to fire away, I'd like to have everything distilled as simply and elegantly as you both do. I think if I end up with a 12-filter system that trades by the light of the 3rd phase of Jupiter, but only at 11:47 and 3:13, I'm screwed!
I should point out that using an indicator doesn't mean taking every single trade that's implied by the indicator. If one were to take every MACD cross, he'd be bankrupted by whipsaw before he ever got off the ground. This is why automated backtesting of indicators is of little use unless one plans on having an automated system. If you're thinking of using an indicator, plot it, then scroll through the chart bar by bar from the beginning of the day, never going beyond the right edge until you've made whatever decision, if any, is called for at the time. Do not look at the whole day, then sink into the "oh, I should have taken the indicator here and here and here" trap. You're going to have to make decisions bar by bar in real time, so make them the same way when you're testing whatever it is you plan to use, even if you plan to use only price bars, with or without volume. Anything else is just kidding yourself. And I assume you're monitoring multiple timeframes rather than using only one. --Db
1. Agreed. The indicator is only there to suggest conditions may be there for a POSSIBLE trade. 2. Yes, I plan to be more meticulous than Felix Unger going through bars. (For an intraday system, how much history do you suggest?) 3. I've always thought it important to look at multiple time frames, and yes, when I do build this system, I would think that will be a part of the equation - starting out broad and telescoping down to the more narrow time frame. (In the past, that's meant monthly charts to weekly charts to daily charts...but I think the same principle holds up on a shorter time frame...)
1. Think "probability" when faced with a possible trade, i.e., what is the probability that this trade will show a profit before it hits my stop? 2. Four weeks, or 20 trading days, ought to be enough, unless they are too much alike. You'll want to include trend days, chop days, gap days, N days, etc. If there isn't enough variety in those 20 days, try 30 or 40. If there is still too much homogeneity in 40 days, try a sample of months throughout the year, e.g., November, February, May, August. But I don't think it's necessary to go through an entire year. As to timeframes, consider using the macro frame as a filter, i.e., if the 2m is saying short and the 5m is saying long, consider waiting for some sort of confirmation. --Db
Great, gracias. 1. Would the probability come purely from backtesting? (i.e., "If these conditions were met, and I entered at X and took profit or loss at Y, backtesting shows I would be profitable 57% of the time?") 2. Good to know. I was planning on going back and doing EVERY year of the ES. (I've been forward testing that Quah SVS system and it definitely trades differently from one week to another. I did notice that it seems to trade better during trending periods and not as well during choppy phases...) 3. ...Which leads to multiple time frames. It would be a good thing to have confirmation from the larger time frame, giving the green light, or a yellow light, or whatever. (I keep thinking SVS would've benefitted from that.)
Probability in this regard is not a number but an estimation of how likely the trade is to succeed. This will be based on what you learn of support and resistance, the relationship of price and volume, the degree of demand and supply. If, for example, you decide on a trending strategy, then whatever tactics you employ will have to fit into that strategy. Any trade that would be taken when you're not trending would be lower probability than one taken when you're trending. Add in a consideration of strength of trend rather than just its existence alone, and you can increase your probabilities further. Time of day, day of the week, time of the year might also be considered. --Db