Your basic research will likely lead to a trading framework (a business plan), which I believe is necessary for consistent profitability. I had plenty of good ideas as a beginner but without a well-researched plan, I failed each time. Once I compiled the MAE/MFE stats surrounding certain patterns under various price environments leading to the patterns (pattern + context=setup), I had a framework for trading. Technical analysis gives you full view of context (price behavior in the minutes, hours, days, weeks, months, and/or years to the left of right now) and patterns (a price action footprint such as engulfing price bar or a 1-2-3 or a narrow range consolidation flag, etc.) that lead to significant (by your definition) price moves more often than not over each series of N appearances of the setup. There are many setups that produce positive expectancy for a favorable price move of at least X ticks before an adverse price move of Y ticks. What confuses aspiring traders and causes them to deduce that markets are random, all edges are transitory, TA is worthless, etc. is the experience of having tried a lot of TA and indicator-based ideas and failed. Trading based on individual price bars, candle patterns, and indicator readings result in failure because the extensive research necessary to determine the context under which these "trade bites" convince enough traders to join one side and overpower the other side was never completed. The other issue is that specific price bar patterns (footprints) that indicate the likelihood of "herd follow-through" in a given direction under certain conditions don't look exactly the same every time. Those looking for certainty or perfection will likely find themselves unable to trade their own plans for this reason. Their risk-averse brains will offer up one reason after another why the setup doesn't meet exact plan criteria (despite the fact that close is close enough).
This is strictly from my own experience, but if I could have back all the money I lost from trading my beliefs based on fundamentals, logic, analyst recommendations, I'd be able to buy a house. I was shorting a stock in early 2009 because it was rising on something that was fundamentally unsustainable (one-trick pony sort of thing) and was trading at a P/E ridiculously beyond its peers. I finally threw in the towel about a dollar from the high. Shortly after that it became apparent that I was totally correct on all fronts and it proceeded to make new lows. If I'd simply traded the price action instead of my beliefs, my large loss would've been a gain and I'd have been able to profit again from the reversal once the price footprints signaled that the herd was turning around.
So, you are saying that it boils down to a blend of intuition and a framework both developed through research?
In the Hall of Fame thread section there is "Article by Pabst" . He is a widely respected trader that provides some insightful persective to TA vs. News and how markets work.
For me, no. My intuition has a poor track record. I'd be consistently profitable fading my own intuition. I needed a well-defined framework to keep me out of trouble. In all honesty, I'd never dream of putting on the trades my plan tells me to without the plan based on thorough research. It feels like I'm doing the opposite of what I should be doing, even after all this time. I know it's the right thing, but my intuition tells me it's not.
I wasted a few years searching for a trading edge in a million dead-ends. Along the way, I decided to focus on automated trading, in the beginning so that I could "quickly" backtest systems on large number of trades. In 2011, I created my first profitable system (that is, profitable once traded in my own account), from a market inefficiency that jumped at me one day, after seeing the same phenomenon develop about 10 times in the same month. I did some quick statistical analysis of the phenomenon over the limited historical data I had back then, found it to be potentially profitable, and developed a system around it. Around the same time, I was trading a specific reversal pattern in Crude-Oil (manually, but with the assistance of a custom indicator spotting that pattern). I slowly evolved the indicator into an automated system - slowly because I tried to incorporate a "million" nuances in the software. I made some money with that system, lost it all + some, stopped trading it & designed a V2 with some real improvement in the core technology, made some money again, but very slowly, and I could see that this pattern was losing its frequency & effectiveness altogether. Beginning 2013, I decided to seriously look into what was going on for this system - I instrumented the system to log a fairly large number of datapoints at specific times (in this case, each time the core trend indicator signals a trend-change), and started the statistical analysis of over 11,000 trend-changes between 2007 & 2012. The 1st "pattern" jumped at me in less than 1h after I had collected the data - it certainly didn't have a big edge, but I had found my way of searching for patterns in the market. The attached are the raw performance of "pattern 151" ; over 3053 trades in 2007..2013, with 2013 out of sample. By raw performance, I mean no filtering whatsoever, no trade management either - the system gets in on a trend-change, and gets out at the next trend-change - no stops no targets (but, commissions & 1-tick systematic slippage on entry + same on exit). Directly out of statistical analysis of those 11,700 trend-changes. The final system (CL151) adds filters, and some very basic trade management (entry-stop & target, an optional BE stop, an optional trailing stop, optional exit before the session close). The filters are the riskiest part, as they deal with relatively small sample-sets, and are easy to over-fit. This is why I don't even care showing here the final system performance, but rather the raw pattern performance, which can be seen as a minimum performance level for the final system, even if some filter turn-out to be somewhat over-fit.
It's the behavior of price that is important to me, not the "beauty" (perfection) of the pattern. As dbphoenix has said many times, price is continuous; the only reason we see it broken up into bars on a chart is because we choose to display the price action that way. So it makes no sense to limit oneself to a textbook-perfect pattern based on price bars. For example, one of my favorite patterns is the 1-2-3. The textbook version of this is price dips to a low, rises, comes back to test the low and rises again off a higher low (HL) to (hopefully) break the last swing high and start a new trending move (reversal off lows/high) or to continue the previous trending move (trend continuation off a key level such as a trend line, moving average, VWAP, etc.). The same pattern occurs to the downside off a lower high (LH). However, in the real world the pattern sets up off double bottoms and double tops and also off failed breaks of the previous low (or high in the case of short setups). In fact the strongest moves can come off the failed breaks because that version of the 1-2-3 traps a lot of traders and their stops will help fuel the move. A recent example of this version of the 1-2-3 can be found on Friday's 1-min chart of CLJ4 (crude oil futures) between 9:01 and 9:04 am (eastern time). Price is falling, prints a low at 9:01, bounces at 9:02, and lures in more sellers on the break of the 9:02 bar low and even more on the break of the 9:01 bar low (the key low that's in play). Anyone who didn't have the lower trend line (demand line) drawn was likely to jump in short. Those who had the line drawn were ready to scratch the short trade quickly and reverse long off the not-at-all-textbook 1-2-3 setup if the line held as support (which it did). There weren't even multiple price bars one way or the other, just a low, a bounce, and very weak break of the previous low. But the behavior is close enough (and on a smaller tick chart it's quite clear) and the trade is on.
I hate to "optimize", except for trade-management aspects (initial stop, target if any, trailing stop parameters), which is an area where a broad brush is enough anyway - the end performance doesn't change much if you pick these parameters in the correct range (usually, pretty wide). As for filters, I only use statistical analysis, I try to keep them simple, and grounded into market dynamics, and I rarely revisit them. For example, I have for P151 a filter that prevents trades entries between 8:30:00am & 8:34:59am (avoid news-induce "noise"), a solid 5min window. The 1min immediately following that 5min-filter is currently net-negative, but I didn't see the rationale for using a 6min filter, instead of a "round" 5min filter.