VecViz provides “Support and Resistance Based Investment Analytics”. Support and Resistance are long standing, popular concepts in technical analysis. However, the definition of neither “Support” nor “Resistance” has ever been well settled, let alone the methodology for identifying and measuring it.
In the table below we summarize four papers from academia that focus on Support and Resistance: two older, commonly cited papers and two more recent papers. We review how their findings relate to VecViz’s methodology, and end by discussing how the definition of support and resistance has evolved over time and may continue to evolve.
| Paper | #Citations | Support & Resistance Source / Definition | Price Data Utilized | Findings |
| A (Jul 2000)1 | 210 | Published support and resistance levels by 6 fx market broker dealers and analytics services, but describes their methods to include local max and min, trend lines, Fibonacci retracement levels. | USDJPY, GBPUSD, DMKUSD for 1996-1998 | Bounces occurred at the published support and resistance levels with more frequency than chance. Predictive power of the published levels was found to last five days after the levels were communicated to market participants. However, attempts by firms to score the strength of support and resistance levels were found to have no predictive power. |
| B (Mar 2014)2 | 37 | Assumes S&R reflects limit order book depth. Local max and min prices are reviewed and scanned for subsequent “bounces”. | 9 large cap stocks on the London Stock Exchange, for the year 2002. Focus on shorter, intraday windows. | The probability of a bounce occurring at an identified support or resistance level was found to increase with the number of prior bounces that occurred there, but also to decay with time (bounce likelihood more than 180 seconds out was not statistically significant). |
| C (Jan 2021)3 | 2 | Assumes S&R reflects limit order book depth. Local max and min prices, more explicitly defined than B. | Brent, EUR/USD and LLOY for the year 2018. Focus on short term, intraday windows. | Confirms findings of paper B (Garzarelli, et al), that the probability of a bounce increases with the count of prior bounces, but decays with time |
| D (Oct 2022)4 | 1 | Assumes S&R reflects “popular agreement”. Utilizes “ZigZag” technical analysis pattern to identify tops and bottoms that are likely to receive “popular agreement”. Imposes minimums on touch count and maximums on lookback period. | EURUSD, GBPUSD, JPYUSD, AUDUSD from August 2017 to January 2021. Uses a lookback window and forward time horizon of 2.5 days | Neural network based trading systems incorporating Support and Resistance related features are more consistently profitable than those that do not incorporate such features. |
Study A highlights include channel oriented techniques, applicability to longer time horizons
The lookback periods used to formulate the support and resistance levels published in the reports is not disclosed, but their predictive power was determined to hold for up to five days beyond the data they were published. VecViz certainly shares a reliance on channels and Fibonacci levels with the publishers of the reports that contained the Support and Resistance levels studied.
Key findings of studies B and C are consistent with important elements of VecViz’s Vector Strength calculation.
The only input to any VecViz analytics at present is daily closing prices. Thus, it would be easy to disregard B and C as irrelevant to VecViz, as they focus only on intraday lookback periods and forward time horizons. However, Mandelbrot argues that financial time series data have fractal properties (i.e. that they are similar whether scaled in minutes, days, weeks, months, or years). Perhaps that explains how B and C possess findings that are consistent with key elements of the Vector Strength calculation. Specifically, their findings that the more bounces a price zone gets the more support or resistance resides there, and that the influence of a zone of support or resistance tends to decay over time are directionally consistent with the treatment of “Touch Count” and time proximity in the calculation of Vector Strength.
Study D confirms that support and resistance related features can add value to a broader, neural net based trading system
Study D speculates that image processing techniques such as those we explored in a prior VecViz blog5 are capturing support and resistance. It derives support and resistance levels from algorithmically derived “ZigZag”6 patterns commonly deployed in Elliot Wave related technical analysis. When distances from entry price to the resistance and support prices are included as features along with other common technical analysis trading rules in a neural network they improve the performance obtained. These distance metrics are somewhat similar to VecViz’s “EUB/FV Put” criteria for the V-Score of expected forward price performance (one of 13 such criteria). The definition of EUB/FV Put from VecViz’s V-Score Closest Comparable Dashboard is provide below.
Conclusion
We looked at 4 studies of Support and Resistance spanning the nearly 30 years we have been involved in the markets. Each adopts an approach that somewhat reflects the techniques “du jour” of its time. Study “A”, based primarily on published broker reports from late 1990’s that, arguably by necessity of the way they were distributed and consumed, were based on visually compelling charts and offered support and resistance forecasts intended to last more than a fraction of a minute. Conversely, studies “B” and “C” were written during or about the post crisis years where high frequency trading was of great focus, and are thus not surprisingly take a very order book composition/ short term oriented perspective on Support and Resistance. Lastly, study “D”, written in late 2022, considers whether Support and Resistance has relevance in the era of neural net based trading systems, that can incorporate vast amounts of information. In acknowledgement of the emerging research on chart image processing techniques, “D” relies more on the visually pronounced patterns of “A” to define Support and Resistance, moreso than those of B and C.
Though studies A, B/C, and D differ substantially, each has some elements in common with aspects of VecViz’s Vector Strength metric and /or V-Score. That gives us some incremental confidence in these important VecViz metrics. However, none of these studies attempt to link the concept and resistance with narrative or cognition of risk and opportunity more broadly, and so we continue with this endeavor, counting upon narrative and cognition becoming “du jour” in the years ahead.
- “Support for Resistance: Technical Analysis and Intraday Exchange Rates”, Carol Osler, FRBNY Economic Policy Review / July 2000 ↩︎
- “Memory effects in stock price dynamics: evidences of technical trading.”, Garzarelli, F., Cristelli, M., Pompa, G. et al. , Sci Rep 4, 4487 (2014). ↩︎
- “Evidence and Behaviour of Support and Resistance Levels in Financial Time Series”,Chung,K., Bellotti, A. , January 2021, ↩︎
- “Support Resistance Levels towards Profitability in Intelligent Algorithmic Trading Models.”, Chan, J.Y.-L.; Phoong, S.W.; Cheng, W.K.; Chen, Y.-L. Mathematics 2022, 10, 3888. ↩︎
- Comparing VecViz’s V-Score to a Chart Image Recognition Based model recently Featured in the Journal of Finance, which discusses ““(Re-)Imag(in)ing Price Trends”, published in the December 2023 issue of the Journal of Finance, Jingwen Jiang of University of Chicago, Bryan Kelly of Yale, and Dacheng Xiu, also of University of Chicago. “(Re-)Imag(in)ing Price Trends” is not cited in paper D, but builds upon papers D cites. “(Re-)Imag(in)ing Price Trends” has 47 citations according to Google Scholar and was published in December 2023. ↩︎
- https://www.investopedia.com/terms/z/zig_zag_indicator.asp ↩︎