Qualitative vs quantitative analysis
Updated: Dec 17, 2019
Financial markets are large field and especially trading can often be too conceptual for participants to define and test their opinions on market directions. It is often said that trading price action is marriage between art and science, as each chart is slightly different to next one and it can often be too much asymmetric information to build cohesive thesis around the traded example. This leads to huge portion of traders using too much of qualitative analysis and trying to make sense out of market without actually creating robust research methods to prove themselves right or wrong in the first place. Logical thinking is one of the most overused and over-blown methods in modern era, and especially when it comes to trading it can be very dangerous and damaging to trading capital. All of which falls under umbrella of qualitative thinking.
Basic sum of both methods and the differences:
Qualitative research studies things in their natural settings, attempting to make sense of, or interpret, phenomena in terms of the meanings people bring to them.
Quantitative research aims to establish general laws of behavior and phenomenon across different settings/contexts. Research is used to test a theory and ultimately support or reject it.
Both methods share same variables, with exception of one additional variable that quantitative research has: The test.
Many traders tend to use qualitative analysis, trying to make sense out of what they see in the market. It creates problem where initial opinions are never stress tested and trader does not even know if his opinion or strategy can perform with positive expectancy. This is especially common in Forex, where trader jumps from strategy to strategy every week without even correctly stress testing if strategy can perform or not, all the result of just logically assuming that strategy or pattern or indicator might perform rather than actually using correct quantitative approach to test it out first before trading with it. Most strategies, especially pure PA / indicator strategies can be stress tested within single day, thus it is no excuse for someone trading strategy for whole month and then switching to next better thing. And yet we all do (did) it, why? Because only trough mistakes one learns the importance of quantitative research since basic school education does not familiarize people with this concept enough.
Personally i did not have too much issues with that, because as a kid i was into biology a lot, and learnt very quickly how powerful the use of testing is. Even without much theoretical knowledge on animals (havent read much books), if you just observe, form opinion and then test it out you can quickly start to see what works and what does not and which variables matter for behavior on animals. Not only that, but as you are using quantitative approach and constantly testing your own opinion you start to notice that to have good realistic opinion about new subject that you are dealing with, it takes about 100 of failed tests of poor opinions before you land to that one which does hold. That is completely normal, it is the basic statistical success rate of "innovative opinion" in new field. And that data gives you real insight that you should never just trust and hold to opinion, just because it makes sense to you, because the overall data says it is not safe to do so.
As a trader you need to rely much more on information that is outside of your head, and rely on objective methods of testing that exist outside as well. Too many traders rely on their inner logical reasoning when it comes to defining an opinion around the market. Majority of real good knowledge exists outside of yourself. Use that to fact check yourself.
The core of quantitative analysis is testing the hypothesis, testing the strategy or the pattern if there really is consistency to it, or if trading it using strategy can bring the required performance.
The holy grail of trading
The real holy grail of trading is to use quantitative analysis and find quantifiable pattern in markets that is backed by huge amount of data. That by itself is the real best thing that trader can have. Pattern with 1000s of historical examples where large amount of statistical research can be done with, forming decent robust trading approach on it. It is not an easy task. Majority of very frequent patterns in markets are very simplistic patterns (for example RSI indicator passing into 70 overbought area) and those kind of patterns do not have any edge, patterns with edge are more complex overall and much more difficult to study or form proper research on. But trader should strive to approach it from the view of quantitative approach, using the variables listed on the image above to find few of such patterns. In reality being fully quantitative researcher means that one has to deal strictly with math, and in retail trading that is just not realistic. Thus there will always be some qualitative approach used, but the goal should be to include as much quantitative approach as possible in the mixture.
Examples of qualitative analysis approaches:
1.I have read about cryptocurrencies for about a month and i believe that this indeed is the future of money. If i think and it makes sense, thus it already is probably true.
2.I have read about small cap stocks on Nasdaq being good shorting opportunities, due to fundamentals. All articles made sense, thus i will start trading it next week.
3.I think that certain country will go into debt crisis, because it has over 120% of debt to GDP ratio. It makes sense that is over-indebted and it is impossible to sustain higher debt without yeilds rising a lot higher, thus triggering shrinkage crisis. If it makes sense, it probably is also true.
Examples of more quantitative approaches:
1.I have read or a month about cryptocurrencies and the idea sound solid, but now let me research about past innovative technologies to see if there is specific pattern that might confirm for this technology to stay around. Let me also research the history of money to see if this technology really solves certain problems around current fiat currencies which might ensure for tech to stay around (technologies that solve problems usually tend to stick around). Let the data prove my initial opinion.
2.I have read many articles about shorting small cap stocks on Nasdaq and it does make sense due to fundamentals. Let me now research 500 historical tickers, their long term charts and their intraday charts when there were big moves. Are there specific anomalies on the moves, are certain tickers moving more aligned and if so why? What are the common variables of the ones that do move aligned and the ones that do not. Is it in the fillings, the float, the orderflow, where? Once i found the variables, how would or should i sort the setups and how can i know ahead on each live setup under which pattern to sort it? If i cant (in some cases you wont know) should i still trade the ticker knowing that i do not have specific setup or should i avoid it?
3.I have information that certain country has over 120% of debt to GDP ratio, which is very high, should that be cause of concern? How can i find that out. Studying history how many countries with such outstanding debt had liquidity issues following? And if so, how soon? Were certain cases where issues were in next year, or next 5 or 10 years? In trading, timing is everything. I must not over-simply. And if there is significant difference in timing, what separates some countries away from others, why do certain countries go into problems that much sooner, there has to be reason for it. What variables are the ones that matter and tell that. If every behavior has a reason for it, then i have to find out those variables that define a reason. Just because it makes sense that 120% debt to GDP might be a problem, it might turn out not to be short term problem for this specific country i am tracking, due to X, XY, XZ reasons.
The more variables the better
Most common issue that traders have is that they dont even know where to begin when it comes to testing the data or how to even collect the data. And that is often the case due to not defining deep enough all the variables that "setup" has. The more variables that are outlined the more chances are that hypothesis can be tested and the results might be reliable.
For example when defining the setup, dont just have a setup of "break of high of the day". That is too simplistic , too qualitative. If such setups are collected the data will be noisy. Instead define it more and only focus on specific type of assets that share more than just that one variable (to align it more with quantitative approach). For example:
-is low float stock
-is up on a day 50%
-is breaking high of the day
-is heavily shorted (half of float is short)
-is very liquid (tight spreads)
-is trading on high volume relative to what it trades over last 50 days
To use an analogy from biology, if you had to define animal well, would it be enough to just use 1 variable? Of course not, data would be noisy. Instead the more variables you use the more precise could definition be so that you avoid data-overlapping, basically bundling together animals that are not from same species just because they share one common thing.
Know when not to have an opinion
Best traders know when to not have an opinion in market. Quantitative traders that trade markets over long time know to wait for specific situation that they have research and edge in order to start forming opinion on asset. People on CNBC or Youtube tend to have opinion every second and every tick of price action (especially high time frame traders / analysts), but those opinions are never tested and often tend to be complete guess.
As trader one needs to learn how to be patient and stay on sidelines with opinions, it is completely okay to say "i dont know", there is no shame in saying that. Trader should wait for specific pattern that he / she has edge on and only then form opinion.
And there is a balance on how solid should traders opinion be. If opinion is too strong, trader will be unwilling to change it, if price goes against him, if opinion is too weak or soft, trader will be potentially cutting trade too quick in panic reactions. Delicate balance of neutrality with slight tilt to one side is where good balance is.
Using too much of qualitative thinking, trying to logically make sense out of everything will eventually push trader into having too many opinions at random times of market with low yielding performance. Quantitative approach teaches trader to research and really form solid opinion or formation of setup and then wait for it to develop. No pattern, no opinion.
"If you wish to see the truth, then hold no opinion for or against."
Lets take an example, how a quantitative approach works and starts in trading. Lets give a conceptual example in equity markets, small cap stock that is running up high % move on a day and sets consolidated structural pattern. As a trader you should recognize which way to thinking approach you are using based from those two examples. In reality we all use both methods, the only difference is by what extent do you use either of the method, and how much are you trying your best to not over-use qualitative approach (doing the homework and proper playbook).
Bellow is example of how qualitative researcher might approach it:
Bellow is example of how quantitative researcher might approach it:
Basic premise is that quantitative trader always second guesses him/herself. Always seeking for some sort of data to confirm the initial thesis, while fully qualitative trader will do a lot less second guessing and just initiating on whatever the opinion on trading ticker is. But the real key difference is not about the data. Both analysis approaches are data base. The key difference is in consistency of data, quantitative approach focuses only on executing on ideas that prove historically to have certain consistency of behaving in certain way, while qualitative approach says that you dont need consistency, one can just rationalize every unique situation and make trade out of it.
To trade any pattern in market with edge, the behavior of it should be somewhat normalized. How normalized it should be, will depend on many specific factors, but the basic principle is, if pattern is not clustered with normalized behavior - it is random and if something is random it is impossible to trade it with positive edge. With other words it is highly likely to burn trough whole trading capital if one is trading non-normalized pattern.
Generally the rule is, the more complex the pattern the harder and bigger the stress test will be. For example collecting data on price action pattern is hundreds of times easier and less time consuming than collecting the data on financial crisis (fundamental) pattern. Retail trader should put more focus on simpler patterns that are frequent, because the less frequent and more complex patterns are not worth the time and capital input if trader is not substantially capitalized.
Bellow is conceptual example of normalized data chunk of specific behavior. This is how solid pattern should behave if one wants to trade it with edge. Some data is scattered around (random) but large cluster is chunked which shows similar behavior across number of setups.
What one needs to do is stress test by collecting: -recording screenshots of every setup to find if there is behavioral symmetry of price
-observing live patterns and writing any important variables (uniqueness)
-carefully construct the data that will be collected around pattern. For example, should trader collect the time dates on setups, should % gain or move be considered for data collection? Should short interest be collected? Basically trader should collect data that matters and exclude the rest in order to keep it as robust as possible. Test the variables so that you see which variables do not matter so that the future tests are more robust and simplified in right direction. Some variables will be present consistently across specific setup, but they do not provide any edge data overall.
-getting as much historical data as possible to determine if pattern is new or how old it is (year, decade, century) to fully understand what is likely to expect and doing data normalization graph to see if there are any anomalies within specific years and if so why.
This might help you avoid trading pattern around specific situations or time zones, if there is larger chance to under-perform.
-It is very important to collect all data objectively, if data is collected with subjective bias it will give trader false expectations. It is very common to see subjective bias in data collection of traders and It happens for a specific reason, explained bellow.
As data collector, you want the test (data collected around pattern) to be meaningful, you want to find an edge, therefore you have sub-concious incline to collect only positive data and ignore negative or random data. Therefore the way you decide to collect data has to be strictly specified ahead with set of many rules, so that each pattern that you collect goes trough full "flight check" to ensure that collected data is honest. Cognitive bias has significant impact, personally i can say for myself especially in first 2 years of trading, as well as hundreds of other traders who did not even realize the impact of subjective cognitive bias on their data collection around the patterns. Basically you are lying to yourself just because you want to find meaningful pattern in the market. This is also one of main reasons why traders often point out in hindsight the patterns that did work out, and the ones that do not, they hide them in "closet of skeletons". As a trader you need to practice being neutral and objective, letting your inner monkey stay behind the cage and using outside honest data to confirm the quality of play.
No matter how much quantitative approach one is using on research / analysis, if there is no objectivity in data collection and research, then it will be meaningless. Practical example, when you are trying to find a new pattern and you are collecting winning and loosing examples you will be inclined much more to just save winning patterns and discard the loosing ones. That is a fact that will by default be present to every trader. You have to fully recognize it that you are doing it, and form a proper data collection plan or mindfulness that prevents you doing it.
My usual method of collecting the data is by using binary counting with simple 1 - 0 or Yes / No collection on specific behavior of price. There is no need to over-complicate quantitative data collection, as long as it is objective is good enough.
If there is substential normalized data around specific behavior (for example 70% of 1 / Yes) then chances are there is some process pushing that price to behave the way it is.
Anything in universe that is consistent is consistent for a reason, its not just random.
Example of binary count on heavy short squeeze pattern in small cap equities (with float under 5 million shares):
-Is short interest in last 4 weeks increased to over 3 million shares ( 12 Yes, 37 No)
Example above is normalized on non-confirming side which is not good, retail trader should be focused on researching the patterns that conform on Yes side with normalization because those are going to behave more consistently and it will be easier to extract edge. But by default those kind of patterns are much harder to find, because by default majority of patterns will be non-confirming (no edge, no consistency of behaviour, random).
To sum it up, as trader one should use quantitative analysis as much as possible and always test the potential opinion or pattern. Use qualitative analysis only when you cannot afford quantitative due to time or capital cost. But mind that you are not in this game to take short cuts, trying to make gains as quick as possible with as little time input as possible. This just does not work. This is innovative sector, in any such sector to increase the chances of long term survival the quantitative approach will always be a better choice, the only question remains how much of quantitative research should you do as retailer. Hedge fund might be able to afford proper quants that do research every minute of a day to provide the data to trader, as retail trader you need to be able to weight it realistically on how much you can afford to slack, and on where you need to have quantitative approach.