Quant analytics: Are There Any Statistically Proven Charting Patterns?

(Last Updated On: June 2, 2011)

Quant analytics: Are There Any Statistically Proven Charting Patterns?

There are a million and one charting patterns out there but does anybody know of any that are statistically proven to predict correctly a high % of the time? Has anybody found any positive results using back tested simulation?


I have done a fair amount of analysis in this over the last 12 years and for the most part the answer is no. I have yet to see a charting strategy that does more than give you hints about what is happening in the markets.

(Hopefully I am not starting some kind of flame war. This is just my experience and opinion.)

Most charting strategies are variations on either mean reversion concepts or data filters. These ideas can work ok over short periods of time but you need to know when the frequency of the price movement changes or you end up getting eaten alive with bad trades.

The succesful chart traders I have had experience with have all used good judgement on top of the chart info to make solid trades.

If you look at support and resistance concepts you may have better luck. I feel that these are at least rooted in a bit of fundamental reality.

The greatest success I have had in trading has all centered around specific events. Points in time where something determines what the majority of the market will do in a symbol. These situations allow you to seperate the real trading info from the noise.

Good luck with your search and let me know if you have any luck. I would love to see some solid info in this area.

I knew a manual trader who tried to explain why I couldnt model what he did, my point was your brain is logical, I can model what you do. He said “no you cant, I have had tens of thousands of hours watching the markets, I am intimate with the market, I know how it moves across different time frames I watch the cycles and how they bounce between support and resistance, its this knowledge and 6th sense that I have, that is what you cant model”.

..it was then my suggestion to look at having a Artificial Neural Network take the 5 time frame over bough and over sold oscilators as inputs then ‘the grey area/6th sense/the network’ would work out the predictions, so its this ANN which would get the feel for the market. Possible?? I dont know but im sure there is somebody out there trying!

Obviously there is a statistically proven method somewhere out there of making the correct trades, otherwise we wouldn’t have people/funds out their consistently returning 30% year over year, the trick is in finding a strategy and keeping it a secret.

I agreed. If there are a strategy out there, it will be like a magician secret. No one will tell you. The prove to that is what Joseph mentioned. A consistent high return given that there is no insider trading and no acting in concert to corner a market is a repeatable pattern. Purely base on technical, if you permute enough technical indicators, by some astrological coincident there is a straight alignment that you can confidently say it is 99% confident that it will make you money if you exit in time after you hit profit. The problem is that many of us are too greedy and do not know when to exit.

Yes – but it’s a secret.

How about over a 20 year period of research with proven results?

Minimum targeted yields are 35%. Expected ROI is 60%. I don’t talk about anything above 60% because it just tingles my toes


Bought this book (http://www.amazon.com/Technical-Market-Indicators-Analysis-Performance/dp/0471197211) a while back, and it makes for some depressing reading for a purely technical trader. The good thing though is that it does not discount the ‘mainstream’ indicators completely. From my humble perspective, the trick is to use a combination of signal types across a large number of stocks to show you were to look (i.e. something has changed with a stock, either fundamentals, events or perception) and prompt further research, not rely on them as a crystal ball


Many thanks for this book! do you have any more gems? I have a few backtesting ideas I am trying out currently, I think patterns alone done make money but good stop losses and take profits and money management is the key 🙂 thanks again


One thing i must add is that technical analysis alone will not provide you the oracle into the profitability of trading. it must be mixed with a trading strategy e.g. fx spot trade together with fx option or future to hedge and increate leverage depending on which side the market opportunity present itself. Things like miss pricing in the “triangle arbitrage” are also a good hint in the asymmetry of information across different geographical boundary (observed in trading time zone).

Putting all these ideas together in a harmonious way will bring harmony in your technical analysis. This is what the old timer trader call it experience. This is the hardest part for me to implement it.

My hardest problem is to have an indicator that can identfy the trend change i.e form bull to bear. Simply looking at MACD crosses and MA cross is not good enough(visually it is not a problem). I am happy to heard if any other people have other ideas.


I am a strong backer of technical analysis and such, but the one thing that you cannot model, the one thing that separates an average prop trader from the founders of Citadel is that you can’t model human emotion with precision. Some people just have the extra bit of intuition of the markets to make it.


I believe the key operative words in your question was – “statistically proven”.

Then you had to use the term “charting patterns” in the same sentence. My reaction to that is similar to Lee Marvin’s reaction in the bar scene in the movie Paint Your Wagon when he found out that this Mormon kid had never drunk any booze or had sex.

My first “charting” book was titled Technical Analysis of Stock Trends by Edwards and Magee copyright 1966. I was a mere “yute” then. I rigorously read the book (and others like it in later years).

While a lot of the concepts of “technical” analysis were interesting and charts make pretty pictures, when push comes to shove they are a 2 dimensional (time/price) tool which every chart pattern know to man doesn’t even factor in volume to explain price behavior. Granted volume is often at the bottom of the chart. But last I checked, even that is not used to define the double dips and woop di doos that comprise the terminology of classical technicians. I apologize to MTA if you are offended.

A lot of the supplemental technical tools (such as Bolinger Bands, RSI, etc.) are conceptually interesting and even useful in an intuitive way. However, while each tool may have been developed using statistical analysis I know of no SYSTEM (other than one I am automating) which provides a conceptual framework into which ALL these tools can be integrated.

If anybody can suggest a reference on the subject which proposes such a systems level conceptual framework (on how to pick and trade stocks) I sure would be interested.


wouldn’t the goal of the statistical research to determine which of the indicators/patterns/whatever *did* work and which didn’t? Doing so would allow you to use only those that proved their worth (assuming such a thing is possible).

That being said, I have some doubts as to whether successful trading can be a completely non-discretionary pursuit, at least given the current level of technological sophistication. Markets are an awfully complex system to accurately model. It’s much like why economics is a social science, not a physical science.

There is at least one more book I know of and some academic papers on the subject. I know of this paper (haven’t read it yet):http://papers.ssrn.com/sol3/papers.cfm?abstract_id=722264 .If I remember any more sources I’ll post them.

From my experience, most strong patterns work 3-4 times, then they get detected by enough people and the patterns “evaporate”. New patterns arise every day though. Old ones come back too.

What is a systems level conceptual framework?

Let’s assume that each of the technical methods (RSI, Bolinger, charting (?), etc.) by itself is based on statistically significant research and sound in it’s own right within the parameters of that indicator..

The question remains – How does (should) one relate, tie together, and reconcile each of the different recommendations from the separate indicators into ONE coherent buy/sell/hold and at what price? That is what I mean by conceptual framework.

A system is comprised of various functional components which describes a process which produces a given result. In the case of stock picking and trading, presumably that desired result (objectives) are maximum ROI, minimum risk, and maximum (or a specific level of) liquidity (unless somebody has a better set).

Anybody solved that problem?

have done a very thorough research on this area about 20 years ago. We had several markets but concentrated on the SP500 futures with 25 years of data. We use an old book that read like an almanac of all technical metrics ever used and there intended purposes.
What we found is that in the confusion matrix (down, market lowers, up, market increases, down market increase (false negative) and up, market lowers (false positive)) is that all the indicators do not have a sustained year-in-out consistent statistics.
Indicators with averages are laggers, so we invented indicators with median which works somewhat better.

The fundamental problem we found is that they are all causal filters, they are telling what happened and only coincidentally the future will be discovered. you cannot defeat this property of filters.

We then let the computer take combinations of indicators at various time frames and hundreds of format, and using machine intelligence, select the best combination for prediction. We had immediate success, but the better the system the more complex the machine the shorter it lasted. There was a compromise there.

So we turn to a completely different measurement of the market filtering out everything that was not a max or min of the market. Exactly the opposite of filtering. The system worked a lot better with unbelievable sharpe and modified sharpe ratios, the proverbial out of the park all season hit. The problem now was to approach realtime fast trading, sometimes the system had a hard time generating the signals versus the historical testing. Our results are on the 5:1 deleverage contract, returns of 60% a.a. on a max ddwn of 5% for all periods.

Our conclusion is that none of the indicator work is satisfying by itself and in combination. To solve the problem of indicators, and the problem of realtime fast signaling, machine Intelligence needs to come to play with a person who knows what they are doing. Remember that we are asking the computer to design a money machine. And that is what we do today.

Interesting articles from Larry Connors… do you have the exact link for Dr. Andrew Lo MIT to his articles?

You can find them pretty easily on google by andrew lo technical analysis
or the others. Make sure you read also the 4th entry which is a critical review. The test that we have done include most of the know indicators which numbers on the hundreds, including sentiment, volume, number of transactions, seasonal, besides the lagging filters.

My conclusion remains.

If you are interested in an international competition using non linear models for prediction: INNFC – First International Non Linear Forecast Competition, 1996,

Tenorio*, M. F. , Nonlinear Financial Forecasting, R. Caldwell, ed., Finance & Technology Press, 1997, ISBN 0-9651332-1-4. Also contains articles from the First International Nonlinear Financial Forecasting Competition, 312 pages.

which has some very interesting conclusions, even if you decide to create a homemade indicator using machine intelligence with the brightest on the field. The main conclusion is that prediction and making money are not the same thing and the prediction requires a post-production step. All outlined there.



NOTE I now post my TRADING ALERTS into my personal FACEBOOK ACCOUNT and TWITTER. Don't worry as I don't post stupid cat videos or what I eat!
This entry was posted in Uncategorized and tagged , , on by .

About caustic

Hi i there My name is Bryan Downing. I am part of a company called QuantLabs.Net This is specifically a company with a high profile blog about technology, trading, financial, investment, quant, etc. It posts things on how to do job interviews with large companies like Morgan Stanley, Bloomberg, Citibank, and IBM. It also posts different unique tips and tricks on Java, C++, or C programming. It posts about different techniques in learning about Matlab and building models or strategies. There is a lot here if you are into venturing into the financial world like quant or technical analysis. It also discusses the future generation of trading and programming Specialties: C++, Java, C#, Matlab, quant, models, strategies, technical analysis, linux, windows P.S. I have been known to be the worst typist. Do not be offended by it as I like to bang stuff out and put priorty of what I do over typing. Maybe one day I can get a full time copy editor to help out. Do note I prefer videos as they are much easier to produce so check out my many video at youtube.com/quantlabs