Searching for edge and back-testing
This is an overview guide on how to search for an edge in the markets. It is rather a complex topic and the article will cover only basic touches and methods that trader could be using for edge discovery. Humans are curious by nature, which is required to seek the edge in the markets as without curiosity the whole premise of finding something that works as the pattern is not possible.
Many traders tend to over-simply the process of seeking an edge, they strip too many variables away from the market or pattern they are trying to find an edge in. This leads to a common problem in an industry where everyone tries to strap random indicators on random charts while trying to find an edge to such randomized methods with too few variables, which makes an edge discovery harder.
In reality, many true edges in the market are constructs that are relatively complex, constructed of many variables, which contribute to creating specific behavior and having that mindset of layered-complexity is the core outline that trader needs to use when searching for potential ideas in the market that might have a consistent edge to them.
Instead of having just one variable as an idea for a pattern, have instead the layered approach of several variables that overlap, each lined in in the same direction (bullish or bearish).
Below is a conceptual example of the difference in such approach, each of those two traders uses different complexity of how well they define their pattern (potential pattern, mind that this is still exploratory stage, seeking for edge).
Trader on left uses the very simple singular rule to define the possible pattern and seek for an edge, while trader on right uses and defines many more variables. Trader on right will have much higher chances of finding real edge versus trader on the left.
The above example is a basic guideline and very common mistake that beginner traders tend to make, thinking that having the edge in the market is just about plotting random support or resistance on the random asset class, and then starting the backtest.
Be as detailed as possible on defining the patterns, that way you exclude trust from yourself and put it more directly towards data performance itself. Below are two conceptual examples of theoretical pattern definitions around which the trader might be collecting back or forward testing data.
Left: Loosely defined pattern where the trader is putting a lot more faith towards own perception of "eye" and "logic" as the pattern is only defined trough few variables (leaving a large gap to traders own perception of what might or might not be the actual pattern)
Right: more robustly defined pattern, giving more trust and reliability towards data rather than traders' own perception.
The more defined the pattern, the more this effect is exaggerated, the better it is. Although there is a tipping point, if patterns definition is too complex, the trader might not find anything on backtesting, or the process of testing would be far too demanding. It is about balance. The balance is mostly found through trial and error.
Single variable=Too much noise in backtest
The problem of having too simplistically defined "patterns" is that there is too much noise in the backtest. A trader might have the wrong conclusion about the actual performance of a pattern because there is too much noise in data.
Traders who have been trying to find an edge in the market only by using simple technical patterns should know what is meant by that.
The majority of strictly chart traders (not using any other methods to complement strategy) will often wonder what is wrong with the pattern they trade. On the backtest they see it working perfectly, or in the cases, they have seen from other traders shown on social media sites, but once they start to trade it, the performance data is suddenly all over the place. It simply becomes 1 win, 1 loss, and it barely turns any profits. This is because there was too much noise in data with a too simplistically defined pattern, or in some cases, it is due to fact that many traders do not even create proper backtests to confirm if something has or does not have an edge.
Think of noise in back testing or forward testing this way:
By noise or hidden variable, it is meant that there are certain aspects that make two "same" patterns behave differently. What trader defines as the same pattern, is actually not, because there are certain critical differences that trader should acknowledge when defining a pattern. With too simplistic patterns, the noise will be high, and eventually, the trader will give up on the method as live results will not deliver.
The critical rule is, to be specific enough, have many variables that define your pattern. And if you do not know what are important order flow variables, then its time for more learning the basics of markets before going back to the drawing board of edge exploration and patterns. Edge exploration is the process of going back and forth between learning, exploring ideas, and going back to the drawing board again, repeating this process until one pattern really clicks.
Try not to under-estimate the difficulty
Think of it this way, trading is a highly competitive industry, you are straight on competing against elite performers and well-capitalized players no matter which market you choose. The complexity of game is high because as the difficulty increases (due to strong players) so does the deception of game, as key players make game intentionally more difficult for their own sake (to cloud you in the fog of confusion as retail trader).
The real edges in the market are a construct of many variables, and a trader needs to approach edge seeking this way. Do not be confused by experienced traders who keep telling you that KISS ("keep it simple stupid") method is the right way to approach trading, and to not over-complicate. It is very common within this industry or any other industry that with experience people tend to forget all the little details that made them profitable or successful in the first place, and thus they tend to oversimplify what really all the required variables are to perform well (because they operate on autopilot after all those years). Creating an edge in markets takes a lot of time and practice, especially if one is doing it from the ground up, using only basic foundations as building core.
When seeking edge you need to understand the basic layout of what market you are seeking edge in, who is general participant (retailers, institutions, market makers...), are you seeking an edge with them or against them, how many other traders you know of that have successfully found edge in that particular market...
One can go about finding that out by reverse engineering, basically finding edge first and then deconstructing the orderflow behind that, or the other way around which requires knowing specific participation of players (market makers) first and then building or finding an edge on top of that, based upon their actions.
From my personal experience in doing countless back-tests, stress-tests, statistical analyses and ideas that either failed or were successful in finding niche areas in the market, there is one thing that you have to follow to increase your chance in finding the edge: Lean on your personal strengths.
For example, if you are a math wizard make sure that your focus on finding the edges in the market that revolves around this specific area. If you are good at behavioral observation, make sure that your focus on finding edge around price behavior, if you are good at understanding what other people are doing, then your focus could be finding edges in order flow positioning of other traders and their actions...Everyone has one specific thing they are better than other traders, lean on that and search edges in the market that allow you to use your personal strength as much as possible.
The majority of traders who find a specific edge in the market, do it by using their personal strengths as a guide for exploration.
And mind that the word "strength" is completely relative. Being very stubborn could be a strength or not, depending on what results does it provides for one individually. Certain people who are very stubborn will suffer poor performance due to this (in general life), while others will have a better track record as that might make them push through every difficulty they come against (in business, relationships, etc). Certain "strengths" might only seem that way because there are more layers of your persona tied to it, it is not just one personal attribute that defines your strengths. Make sure you dig deep enough to know what those real strengths of yours are and focus your edge exploration around that. This way it will also be more enjoyable since edge exploration can be a very frustrating expedition.
If you are very new to the whole process of searching for an edge or just trading in general, you should start very slowly. Start with very simple basic backtests before you move to difficult tasks, otherwise, results of back-tests will be too ill-formed and you will be wasting too much time.
To understand the whole process of edge searching, you need to start with basics and sadly there is no way around it....initially you will need to waste some amount of time. To start with very simplistic ideas that most likely will not work, but those ideas have to be tested in order for a trader to understand fully how properly structured testing process looks like.
As mentioned above, the majority of edges in the market can be rather complex, but as beginner trader or beginner tester, you will be unable to create complex back-test, until you gather enough testing experience to know how to structure quality backtest variables. And by using the word "complex" that does not mean for one to have an extremely good education to find an edge in the market, it simply means that you do not know which variables are essential to puzzle together in order to find a successful edge in the market.
Complexity simply comes from "you don't know what you don't know" and is not necessarily highly difficult by itself, it is just not straightforward to un-cover the real pieces that matter.
Using loosing ideas as guide to structure future tests correctly
Start with simple back-testing of ideas such as:
1.Bullish engulfing candle buy on random chart/instrument , seeing if it can be traded with edge.
Rules of strategy:
-Buy when bullish engulfing candle appears, risk under low of candle, TP the same size as stop loss
2.Bearish pinbar rejection candle in down trend, if it can be traded with positive edge.
Rules of strategy:
-risk above high of pinbar candle
-TP is the same size of stop loss or 2X the SL size
3.Stochastic indicator buy in random up trending chart when indicator turns oversold.
Rules of strategy:
-buy when Stochastic indicator drops under 30 and the price is above 200 EMA.
Important rule, above entries have to be executed on every single setup / opportunity to prove the point. No skipping of setups.
All methods above are negative performing and if trader structures test correctly the methods should all be dis-proven as profitable strategies. If you did not manage to prove the above methods as un-tradeable then your test was perhaps structured wrong, re-do the test enough times until you have well defined back-test data that the methods above do not provide positive results over 100s of repetitions.
Those 3 simple "strategies" are a solid example to get toes wet in the back-testing field and to get as quickly used to the idea that if it sounds too simple to be true it usually is.
Most of those ideas or "strategies" can be scooped up anywhere on the internet, especially common in FX markets or crypto, and by default, they do not have an edge, as it can be easily proven by any decently capable back-tester. It is just a fact that majority of traders do not bother with backtesting , they jut start to trade them straight away, and then only later on wonder why strategy is not delivering the results.
One might ask, what is the point of wasting time on testing the methods that have no chance of providing positive performance? It is simply to get used to how backtesting works, how to structure the data that will be collected around each backtest, and to practice on that such simple method is usually a good way to go about it. Creating a valid backtest with a decently structured idea requires a lot of practice as a trader needs to start asking the right questions straight away, in the initial phase of backtesting.
To clear it up, the above methods are not per se "bad" or "don't even think of using those indicators". They are meant to outline that seeking an edge in the market is more than just about a single candle and trend selection approach, to create a winning edge. In other words, each of those methods needs to be more defined, more specific market, more specific conditions, which assets are ok and which are not, etc etc....
The above methods or strategies are the basic area where everyone should start, yes it is a waste of time more or less since no edge is as simple as that usually, but it is a required step to take for a trader to understand how robust back and the forward testing process looks like.
The usefulness of backtesting comes all from knowing what are the right questions to ask (to yourself) from the initial phase, even before you start to collect the data on backtesting. Because the questions one asks are what formulates the idea, without a decent idea, the backtest has no chance of providing a promising performance.
Backtesting and forward testing
Backtesting: Idea testing process starts with backtesting, using historical data of assets to either search for ideas or to collect them historically to see if the edge can be built around specific behavior. There are two separate stages of back-testing that trader should or could be doing:
1.The first stage, using historical charts or fundamental events, scrolling trough them and finding ideas around very specific market events (fundamental events such as specific news in a specific market, specific price structures under certain conditions....). This involves going through hundreds of historical charts before potential ideas might come about. This is the easiest aspect of it, and the majority of these ideas will eventually fail, which is normal.
2. The second stage is where an idea has proven already some consistency to it, but the trader/researcher only uses his memory to assume consistency, he or she does not have data yet, only remembers few examples where that idea did show some potential. This is where the trader now starts to collect plenty of historical data to move the research stage forward, by potentially finding certain consistencies of behavior in this method or not. In some cases, it takes a day to collect the data and to prove or discredit the idea, and in some cases, it might take weeks. This can be an often frustrating aspect of seeking an edge in markets because nobody likes to see putting plenty of time into research just to see ideas fail after-wards.
But that is how it works, if you find the majority of ideas passing trough from back-testing to forward-testing phase then you are certainly doing something wrong. Your idea/thesis testing methods are structured wrong, as those results are not normal. The "normal" results should be where the majority of ideas do not pass backtesting phase, and never see the light of passing to the forward testing phase. The majority of ideas should fail in the initial backtesting stage, either in the first or second backtesting stage.
Should one use algorithms or automated systems to backtest ideas?
Often traders ask:" Should I use algorithm / automated system to backtest my pattern or should I do it manually?"
Personally, my answer is to do it always manually. To create a reliable and well-formed backtest with the use of code and automatization, it requires a well-experienced trader who has already performed hundreds of stress-tests of ideas and is also a capable programmer and the majority of traders do not have such experience.
Doing it manually will always be a more practical way for the majority of traders as it decreases chances for errors in big time, especially in initial stages of idea testing. Besides, that one is learning more doing it by hand, than just getting a code to run through check of variables. Most of the edges in markets are not very simple, they are layered of many variables, and for that, it is already hard to enough to miss things doing it by hand manually, let alone ensuring that algorithm collects all required data on each sample to sample basis, without any data noise. It is a far more difficult task. The manual method of idea searching and backtesting should be a priority for the majority of traders. Pen and paper, chart and eye.
Only after the trader becomes experienced back tester with successfully finding few proven edges in the market should one start considering using algorithms as a cooperative method to seek for an edge on new patterns.
Should one use spreadsheets?
Common especially in small-cap stocks, is that traders tend to rely too much on just collecting spreadsheets or Excel data, using that as a backtesting guide. For the majority of cases, that is not enough to create robust trading ideas, chart is a necessary component to have along with each trading idea, along with liquidity and order flow exceptions, math by itself usually does not provide the behavioral dynamic, especially not on per minute to minute basis.
Using spreadsheets combined with charts or additional order flow information will always be the best way to approach the testing. Spreadsheet mathematical data is a great overall guide on the performance of those assets, but it should not be enough to base any backtesting or patterns just from spreadsheets alone.
Below is an example of usefully structured Excel data in small caps from a fellow trader. Data collection in this example focuses on non-compliance stocks with reverse splits:
Such data above can be good starting point for potential edge exploration, for trader to start building ideas.
Forward testing (or live testing)
Forward testing: Both back-testing and forward-testing are required methods to prove if a strategy or trading method can be traded with positive expectations. The key to forward-testing is that it possesses certain qualities that back-testing does not provide which are crucial for a trader to test.
Forward testing is performed in the live trading environment and it can be done on a demo / simulated account or real account, however, it is always best to choose a simulated testing environment at least initially and moving to real account only later on using small position sizes.
Demo testing should be used especially if a trader is not very well familiar with software that he/she is trading in, or market in general.
It makes no sense to expose real accounts towards potential technical errors that trader might step upon, rather expose simulated account in such case until trader gets familiar enough with the overall market, trading platform and the broker itself.
Forward testing should also include certain human-specific variables into a testing environment that are not present in backtesting. Those are required to see if the trader is capable of executing the trading strategy at least somewhat optimally:
-ability to make quick decisions once already in position
-ability to re-asses situation as market conditions change quickly
-objectivity under changing conditions
-adjusting of rules
There are many factors to list but those are some of the key ones. Lets break down how each of those components is important and why it has to be put trough forward test for each trading strategy:
Under the back-testing environment, there are no emotions impacting the performance, while in forwarding tested environment emotions can have a significant impact on traders' performance.
Let's skip the phase of what those emotions are or how to control them because that is not the point of this article. Rather the point is, that each edge/pattern that you will be researching might be playing with different emotions of you as a trader, some patterns will trigger certain emotions in you if you do X, Y, Z while others might not. It is important to have an objective approach in the back or forward testing as much as possible, a trader should consider him/herself strictly a tester (not a trader!) and should separate ego completely away from trading while the testing phase of the potential edge is performed.
That means the tester should be objective when trade failed in forward testing phase if it was due to failure of pattern or it was the emotions or human factors that impacted the end result of trade. If tester fails to understand to see objective reasons, the whole forward test results will either fail or be highly skewed to the positive side (not in a good way).
For example, certain traders might be more emotionally triggered if they loose on breakout trade than others, or some might be more triggered with anger if they loose on counter trend trade, others might be less emotionally triggered if they scale into position slowly with 3 or 5 smaller positions rather than single full entry. Things like that matter, and if trader is too emotionally impacted under forward-test it might affect his conclusion on the patterns performance, especially if trader is unable to tell for himself exactly how much impact did his emotions have on performance and edge. And this issue is not to be underestimated.
Additional to that, there is a real factor of where traders simply refuse to cut loss in the forward testing phase, just because they want to see their pattern to prove a positive edge over time. They are so inclined to see future positive performance and confirmation of edge, that they are willing to skew their demo testing performance (risk management etc..) just to prove to themselves that they found something. As a beginner researcher, I have spotted myself constantly under such a situation, by stepping outside of myself and looking objectively at last few thoughts, just to see that I was trying to mask the under-performance of the previous setup to"help" the chances of edge confirmation. And such behavior was observed at many other traders that I was doing backtests with under different stages of development.
Hope is one of the key survival elements of human beings and in the edge exploration process, it shows how big of a role it can play if not taken under full objective control.
While researching and searching for the edge in markets, it is very important to understand all layers of what goes into the traded pattern / edge, and to be as objective as possible. It is far more important to be objective at this stage than 1 year later,once the pattern is actually traded on live account. Because without objectivity and understanding all layers of edge, the testing phase will either be false or fail without trader understanding it why.
Ability to make quick decisions and re-assesing situations
Under backtesting phase time is an almost irrelevant concept, a trader can take as much time to figure things out, while in real live markets time becomes a more pressuring component. To some making, quick decisions are not a problem, while for others it is.
Make sure to know if you are an adaptable person or not, if you are not make sure to forward test your patterns in time frames where you are not feeling rushed (choosing higher time frames).
Or if you are capable of fast decisions without getting nervous then use that to your advantage and expose yourself to the shortest possible time frames in the market.
Trader exposed to shorter time frames can forward test many more ideas than one who is exposed to longer time frames, but only use this approach if you are comfortable with making quick decisions. Eventually, with practice, everyone gets better at making faster decisions, especially if trading or testing long enough.
Ideal for testing ideas in the market is to use shorter time frames (1 minute, 5 minute etc) and quicker repetitions no matter if it is backtesting or forward testing. In the higher frequency of repetitions and shorter time frames, one can test many more ideas, it's simple as that. And since many of those ideas will fail, by using shorter time frames one has a better chance to find ideas relative to your age as an individual.
If you are not familiar with why this concept is so powerful, study machine learning / neural networks approach which has been used a lot over the past several years on AI and other algorithms or robots.
Objectivity under changing conditions
It is very common that traders find the back-testing stage way easier than forwarding testing or a live trading phase. One of reasons is because when doing backtest your ego is detached from the tested idea itself, while once you start forward testing or trading it, your thoughts and opinions on trading asset start to connect trough the ego, now it becomes harder to be objective, especially if price is moving against from your initial opinion or the strategy/pattern that you are testing.
This can have a significant impact on how reliable the performance data will be as if the trader is too un-objective the results of the forwarding test and conclusions made might be completely wrong, and it should not be underestimated how big of an issue this is!
Sometimes traders completely loose faith in pattern or strategy they are testing because they were under-performing the optimal way of executing the strategy by so much, that the results they got were completely different from their initial back-testing phase. Or in some cases, it is the other way around. In some cases, traders perform way too good relative to what their back-test suggests, simply because they keep breaking rules in forward-testing procedure just because they want to see superb results to "tap themselves on the back" and have something positive to look into future (hope).
It is utmost important that in forward-testing phase there is a check-list of rules that trader has to go trough before every taken trade to ensure that objective pattern spotting / entry / exit process was applied.
Especially for under-experienced traders on new strategy the objectivity by itself is not going to be easy to achieve. It is far easier to achieve objectivity by using external rules as your guide, rather than just trusting yourself!
Adjustment of strategy rules in forward or live trading phase
The reality is that every strategy or pattern before it is traded in a real environment, will lack certain practical aspects, it will be theoretically or paper suited, but not necessarily well practical enough to be executed.
As the trader is forward testing strategy/method there will be certain aspects that trader will find impractical and needed to change (too wide stop losses, too hard to know where to enter exactly, too confusing to know where pattern is invalidated based on initial rules written down, too conceptually defined pattern with not enough variables....).
All of those rules will slowly shift and be adjusted if the trader is always asking the right questions to oneself and questioning the strategy from the right angle. What needs to be changed, and how to achieve better results, to make the strategy more robust more straight-forward and yet still detailed enough? Or in other words, how to simplify the pattern, without overly-simplifying it where the edge starts to decrease.
Every edge that trader initially finds is like a living organism, it will constantly change, things around the strategy will change and adapt as trader collects more data, places more trades, runs more tests. The pattern or strategy that you have laid out for the backtest on day 1 should be significantly different 100 days later in the forward-testing or live trading process.
Adjusting the rules of how each pattern is traded is about having a constant curiosity to improve and find a better way to execute your strategy. Always asking yourself, is my entry ideal or can I find potentially a way to trade my setup quicker with more ideal entry. What are potential variables that would let me find such a clue?
Alright, let's go back to the initial stage now, before the backtesting or forward testing, there is an idea shaping stage.
Formation of ideas
Stage 1: Shaping idea and finding potential setup
The first stage is at the same time the easiest and the hardest part. Coming up with ideas is generally easy. Thinking of good ideas that have the potential for edge is actually not easy at all.
Building an edge in the market starts with creating a basis for an idea where the certain market process is resolving/moving around, thus for such an idea to have a chance, the process has to be there. The process involves certain variables that together construct a specific situation in a market that is pushing the price into a certain direction with a certain probability rate.
The practical rule when it comes to seeking an edge in the market, is to ask yourself is there a specific process behind the idea that I am trying to find edge around? For me personally, that was the core foundation of the majority of patterns traded, as they are mostly manipulation patterns with a specific process of order flow behind them. Starting with the question of what process are you looking for will get you in a much more realistic chance to find edge because most things in nature or society that do have the consistent edge are all due to the nature of processes that shape them. There is a common underlying reason behind them all, which is a specific process.
Examples of specific processes:
-heavy liquidation (large stop loss hunt or mass margin call liquidation by strong move)
-very clean symmetry initiated by key single player defending the lows or highs of symmetry progression
-strongly consistent bounces from same level (single player unloading into the level and getting rid of position, seen on tape)
-strong absorption of bids or offers, tape with lots of selling / buying but no progress on price movement
-pump from group of traders in low liquidity environment
For example, if one was to start seeking an edge with either of the processes from above, it would be a higher chance to find edge than by just randomly approaching the market without any structure of the process in mind.
Realistically the majority of beginners will not think this way as those processes are not beginner-friendly to understand since experience is required. For that reason, it was mentioned at the start of the article that edge exploration is a path that trader takes step by step, and it might take a while before uncovering first true pattern.
Markets are not random (or at least not always)
As edge hunter, the core belief you need to adapt is that markets are not random. Majority of order flow and price moves happen for a reason (and the underlying processes), if you accept that markets are random then you have already sabotaged yourself an there is no point in seeking for an edge in random environment (random markets theory suggests that no edge can be found in the market).
The majority of the time market will be random for you as a trader no matter what market you trade, and that is okay. Your task is to extract from randomness one specific situation that is not random to you, and you can prove to trade it with the edge a non-random.
Meaning, markets are to myself or the majority of traders:
-90% random (you dont know the reason or processes)
-1 to 3% not random (you know the process causing it)
Those 1 to 3% is what you are hunting for as an edge seeker. And mind, the only reason why the majority of the time market seems random is not that dynamics of the market are random, it is just because you have not found a specific pattern or order flow behavior that helps you define edge across any minute of the market, any time any day.
Anything that happens in nature or universe has a reason, if you don't know the reason it does not mean that there is no reason or process behind it, it just means you don't know what the reason is.
Think of an example from physics such as gravity. One might come with an idea that gravity is constant process where the object has a certain direction of movement after it is thrown into the air and that process has a certain probability to resolve in a specific direction of the move. The process of why gravity happens or works is a construct of many variables such as rotation of Earth, the density of air particles, wind (could influence), how far from the center of Earth the object is being thrown into the air ....etc.
All those defined variables sorted and placed into just the right order at just the right scale and one might get a very reliable process with 99%+ reliability rate of what the result on objects direction will be (fall to earth with a certain timing, relative to variables). This is an example of how a trader should as well approach the whole edge seeking process. Finding which variables matter and stacking them right to find consistent behavior. Obviously there is no such thing as 99% performing pattern in trading, but coming as close to that as possible is the objective.
From forming an idea to constructing the valid variables behind the idea, sorting those variables into right ratios/situations so that once those variables truly are found just at decent ratio there is the potentially certain market process that might happen along with price direction, giving trader edge to extract.
The reasons and outcomes of certain processes are certain variables that are behind each process. The same applies to trade. The more variables you expand the more precise trader can be on timing the moves in the market, the fewer variables included, potentially the more difficult or more guesswork would need to be involved.
Before shaping an idea about the potential researched pattern, a trader needs to be curious. Without this stage there is no edge seeking process, searching for edge as an individual by yourself means to step into areas that many do not dare to step in, or have never even thought about it.
You need to be curious about what else is out there and never be discouraged if someone tells you that something is impossible, at least within some reasonable sense. This requires digging into many books, articles, seminars, whatever triggers your thirst for learning and research because you need some basic pointers before you can allow yourself to start building an idea in the first place.
Learn, research and study as much as you can, the more you do so the more chance you have to come up with the idea for potential setup. But the key is in research and not learning or studying. Edge exploring process is the process of seeking in dark corners of the market where most do not look or at least you do not see them look at, most of your ideas will come from your own research.
Formation of the idea comes first from getting some knowledge around the topic itself and then doing some basic research and brainstorming about potential ideas for an edge. Without the required basic knowledge one cannot seek an edge in the the market. The process looks somewhat like this:
1.Acquire basic knowledge around specific niche in market you have interest in
2.Brainstorm what makes sense to you for potential pattern in market
3.Define that pattern
4.Get access to historical data
5.Backtest large sample base of patterns
6.Re-define pattern better
7.Forward test initiation
8.Edge confirmation and start of trading on live account
As mentioned on article above, if you did everything right in most cases you will fail at stage 5 to progress further. Only few ideas will actually prove themselves worthy at backtest phase to be pushed forward.
Two examples of gathering knowledge and idea formation:
-maybe you read a few books and watched many videos about how SP500 performs so well over the past decade and decided to do some of your own research into it. Researching several top-performing stocks, and index in general what makes it perform so well, or what technical clues do index or main stock give along with potential further bull trend. Are there any clues ahead of bounces or are bounces on the upside of the trend just random? The point when you have the basic form of an idea and you cast a shadow of doubt on it, that is the birth point where the idea has come about. Now stage 1 begins. "Searching for an edge on bounce trading SP500 along with bull trend".
-maybe you read about how the crypto market has performed extremely well over the last decade, you asked many experts with a proven track record on their knowledge, watched many hours of videos and decided that there is worth in this to dig deeper. You started to research technology, the performance of bull and bear trends in crypto and then came up with some basic idea that there might be some way to spot where long term bear trend might rotate into a bull and it could be an edge to extract. Perhaps you have some indicator/duration of bear trend/volume idea that together makes sense, but you are not sure if there really is an edge behind it. There it is, the idea has been born, now it is time to take it through all three Stages to confirm or disprove if there is an edge to it.
Just two basic examples of how ideas are formed for the potential testing phase, in order to confirm or disprove the edge within the thesis. The above two examples were used, because it is very easy for traders to get un-practical on both of the themes above. As a trader you will not find consistency by having a belief that SP500 will drop next year because it is too over-extended or having a belief that Bitcoin will go 100k in the next 2 years, none of those beliefs are consistent trading methods. The true method is something that is test-able! If you cannot even structure proper back and forward test around your thesis, it might not serve you in creating a consistently profitable strategy.
Be practical when forming potential idea to seek edge in. Make sure that your ideas are not too wide on time horizon, or too loosely defined. They have to be robust enough so that you can actually create backtest around them and deliver some objective test results.
Define your ideas well by following those key sets of variables:
-time horizon (ask yourself, what is the realistic time horizon for the idea to play out, 1 month, 5 months, 1 week If the play-out time horizon is too wide, it might not be practical enough to use in trading)
-frequency (ask yourself, how frequently have you spotted that idea later, as it once every week, or once every year? The practical trading idea should be as frequent as possible, without frequency there is no consistency)
-complexity (if the idea is too complex, it might not be practical to replicate it many times)
-subjectivity/objectivity (is an idea built fully on objective information and research, or are you building it on suggestions or personal views of some other individual, or perhaps yourself?)
-your knowledge rate on that topic (are you a beginner on the topic around which you are trying to build an idea on, or are you well informed overall?)
It is all about practicality, the more practical one defines ideas in the first place, the higher the chance that there might be some edge for a trader to fin and extract. Or to keep above variables as short as possible: Form ideas that have the short time horizon to potentially play out, are very frequent in their appearance, are not to complex, are based fully from objective data and no personal vies and are based on the topics that you have very solid knowledge on.
People are by nature dreamers, practicality is something that is built into a human, not born with. That is why most traders dream about 100k Bitcoin trade ideas as those are much easier to define, and require no work to put in place. Trading ideas without tests are just thin branches of tree that one is hanging on by only using hope as a method.
Structure your ideas well
To have better chance in forming idea that can actually be sent trough stress test it is usually best that idea is properly structured. By that it is meant, that idea should follow certain hierarchical structure from top down (pyramid) , starting with core, expanding into broader core, then into broad base.
Bellow is example case of my own structuring of idea for Clearout pattern from several years ago:
Example of Type 1 clearout setup bellow:
The above clearout pattern example is how to use three cases of Pyramid to set your ideas, starting with the top where the main core of the idea is. Then the second lower case is where the main characteristics and variables present are outlined. The last stage is questioning how those variables should be shifted/positioned/scaled at what magnitude, so that when those variables are just at the right scale that specific setup might occur.
A quick roundup of ideas for a potential edge in markets might be:
-long trap, short trap
-manipulations of pumpers, manipulations of market makers, manipulation of investment banks, manipulations of central banks
-up trending stocks under easing central bank conditions
-down-trending stocks under-tightening central bank conditions and soft recessions
-shorting emerging market currencies for over-night swaps (USDTRY, USDMXN...)
-investing in assets and hedging short against them in margin markets
-trading against central banks, trading with the central bank
-trading against a parabolic moving asset (air pocket)
-investing in very hot extended assets and margin shorting at the same time against
Above outlined are just some ideas, nothing more. The whole point of using devil's advocate tactic and doing back-testing is that many ideas sound good in theory but on paper (and testing) they fail to deliver positive edge expectancy.
As a trader, if you are constantly seeking the edge by yourself be ready to hit the wall almost every time with the idea that you just came up with. Out of every 100 ideas perhaps only 1 will see a light and finish Stage 3 with actually proving it can be traded with an edge. That is normal. If you are not a very patient person get ready for some big frustration ahead.
After an idea has been formed, and trader judges it to be well structured (not too conceptual, but just enough robust following the principle of the pyramid above so that it can be placed into the historical test) then the next stage is where the fun begins. That is the most enjoyable part of the whole process, it is to find the actual single historical example that matches the idea very well.
Just one single example, that is all needed for a start, no more than that. The setup that one is looking for is the one where market conditions match all the variables as being part of a potential idea/setup. The more variables the better, if there are too few variables defining the setup, one is exposed to data- overfitting (you see what you want to see).
Hunt for the single sample
To start with research it is best to focus on a single market for a start, and research around liquid conditions where price and order flow is cleaner.
The important thing to add: At this stage, you are not looking for this single example on live market trading, you are looking for this setup on historical charts / orderflow!
Why? Because defining ideas trough live charts is far harder than doing it historically since you will be impacted by all sorts of emotions and non-objectivity thus doing it on historical charts is a MUST.
You might need to go through hundreds and hundreds of charts or Level 2 / tape before you find that single setup that matches your idea. It might take you day, a week maybe even more, but technically it really should not take you more than a week, especially not if you input at least 2 hours per day into research. If it takes you more than that, you might be doing it wrong, or just looking for a setup that is way too in-frequent to even day trade it in the first place. Remember you want ideas and setups that are relatively frequent so that you can stack your playbook with frequency to play of each day.
Get historical data
A necessary step is first to get software that provides historical data. Those are my top picks for markets that I research and trade:
-Forex (MT4, download historical M1 data from their center or just import it from external sites in .cvs format for the currency pair that you want to research on)
-Crypto (Binance has relatively enough of historical data if you just scroll to the left side with the mouse, but for accessing historical tape data you will need external data source)
-Equities (ThinkOrSwim has plenty of free available historical data, but for equities, there are really many other options (Tradestation is also great), however, most of them are paid, you will have to pay for historical data), TOS and Tradestation are the only one free if I am not mistaken.
-futures (Ninjatrader has plenty of tick or M1 data, usually randing with few months)
Many charting platforms are by default limited with intraday historical data, one might get an only day or few days of data maximum. There are also many paid sites that provide intraday data for any market, but that data can often reach very costly numbers. Some sites charge 500 or even more than 1000 USD just for single assets, for most cases there is no need to purchase that as free data from listed sources above is just good enough. The only reason for purchasing the expensive historical data might be for algorithmic traders who need very thin tick data (1 tick) with months of years of it.
Moving forward into searching for larger sample base of setups
Once you found your first setup the light bulb should turn on in your head, make sure that you really save that one, and study it into details as much as possible, tear down every candle, every order inside the structure if you have access to such historical data.
Then next step is to find more such examples, use that one set up as a reference and then try to find as many more as you can.
Excluding overlapping non-important variables and keeping unique variables (noise reduction)
The next part is to seek which variables are not very important and do not contribute. Each actual setup in the market is a construct of variables that contribute and those that do not. And the only way you can find that out is to form variable-exclusion tests.
At the minimum one should collect at least 100 historical samples, anything under that might create un-reliable results when moving forward to stage 2 or 3. Mind that statistics is one of the key components of defining traders' edge and without enough samples, there is no statistical probability.
One of the very common mistakes in trading research is that traders under-use the sample size of collected setups. If the sample base is too small it can cause complete unrealistic expectations of how the setup should actually perform (for example only having 5 historical examples).
To give simple analogy of why sample size matters:
Let's use an example that anyone can get behind, using dogs' behavioral patterns.
Imagine that you are an owner of a dog and you know your pet well very well. Knowing the usual behavioral patterns because the dog has been with you for many years, you didn't necessarily "study" its behavior in details but you observed the key eating/sleeping/barking patterns so many times that you just know the overall pattern rate of success or failure under specific conditions.
And let's use a theoretical example that this dog barks on almost every person that walks through the door. Now if outside observer came into house and the dog is not barking on that individual but that individual suddenly starts to base opinion that this is a really welcoming and well-behaved dog, that would be example of making a wrong/unrealistic conclusion based on only using 1 single sample, while in reality, the conclusion would be totally wrong. It is just that certain specific conditions did not trigger dog under that situation, but that is not the average behavior pattern for this specific dog.
That is basically how too weak a sample base, can leave you with wrong conclusions if you start to form an opinion too quickly without having a strong enough sample base. My general rule is 100 is the minimum sample base, the more the better.
This by itself is one of the most common mistakes that traders make. They see a single picture-perfect setup that worked well along with their expectations, and they start to base that this is a common behavior for such a setup. They get stuck in the assumption phase. The whole purpose of testing is to prove that assumption=behavior or to disprove it. In trading, you cannot make consistent profits on assumptions but one can make it on consistent or relatively consistent behavior (large sample size).
Stage 2: Finding potential edge on setup
Finding potential edge on setup/ idea that you have outlined is about finding the behavioral data that can be extracted with edge (data that repeats with consistency towards certain direction). In nutshell what it means is to prove whether the setup is completely random, or if there actually is some common behavior to it. If setup and all of it behavior is very random then edge cannot be extracted from such setup, it would make trading just way too difficult and too in-consistent.
Bellow is conceptual example of random pattern:
Bellow is example of non-random pattern with specific consistency of behavior:
Seeking for edge is an exploratory process that requires a lot of patience and passion for markets. It is not for everyone, and many traders who will try to find edges might not make it. There are two ways to having the edge in the market, one is to find it by yourself or second to learn it from someone who already has found an edge, but it it is very beneficial to find it by yourself as that way one can fit trading approach strictly to owns unique personality.