Unmasking the true accuracy of Cindicator
A recurring theme throughout my undergraduate business education stressed the notion that “there is no such thing as a free lunch”, or rather, that it is impossible to get something for nothing.
While this old adage hailing from the early 20th century rings true across nearly all facets of life, it is especially relevant to discussions concerning financial markets. There is not one speculator on this Earth that can be said only to have made winning trades — and no investor that can claim to have avoided all financial risk prior to seeing positive returns. Success in markets always comes at a cost — Whether it be in the form of capital allocated toward an investment, time spent researching various projects, or many months (even years) of patience and trust in the fundamentals — nothing is free.
Take for example the infamous case of BitConnect.
Among financial professionals, it is known to be wary of “guaranteed profits” and those who claim an ability to provide them.
Nonetheless, this did not deter individuals from expecting “a free lunch”. Some may have escaped unscathed, but ultimately the price was paid. No matter how clearly fraudulent a project may be, there will always be those enticed by the allure of riches and a charlatan’s promise.
“Hybrid Intelligence for effective asset management”
I raise the concept of a “free lunch” because I find it to be particularly relevant to Cindicator. It is no secret that among speculators there can often be witnessed a widespread expectation of earning significant returns through minimal effort. When it is suggested that traders can obtain highly accurate market predictions from a project with a published track record, it’s easy to understand how some may arrive at the conclusion that they have discovered an end-all, be-all solution. But as we know, there is no such thing.
This is why it comes as no surprise that many react poorly to “inaccurate” indicators. Reactions such as these demonstrate little more than a surface understanding of Cindicator’s usefulness as a tool. Only once the illusion of an end-all, be-all trading solution has been shattered, and one takes the time to learn how to best interpret Cindicator’s predictions, can its true potential be realized.
What I aim to demonstrate in this post is my approach toward analyzing indicators over time to develop trading strategies based on its performance, and how “inaccurate” indicators, may prove to be highly accurate indicators in disguise.
A real-time sentiment gauge
Firstly, let us consider the value in market sentiment data. In any market, there exists a vast array of participants with varying time horizons, positions, capital, strategies, and experience. While some investors are less willing to engage in frequent trading activities during volatile market cycles, there are swarms of traders concerned with the day-to-day price movements, hoping to benefit from volatility.
“Sentiment describes the opinions, emotions or views of a group of people. In investing, sentiment can be a powerful determinant of security prices, especially in the short run. Here, emotions — whether rational or irrational — can drive prices. . . If you could forecast changes in sentiment, then you should have an advantage in determining changes in the market. The problem with sentiment is that it’s really only known after-the-fact.” (Thorpe, Wayne A. “Gauging Market Sentiment.” Computerized Investing, June 2001, pp. 21.)
Published in 2001, Thorpe’s article reveals that market sentiment data has been of great value to investors for many years. Traders expecting the price of an asset to decrease, are likely to sell , or wait for lower prices to buy— and vice versa. A person with the ability to monitor, in real-time, the prevailing sentiment among investors as it relates to individual assets, industries, geopolitical events, and markets as a whole — is undoubtedly placed at a significant advantage. With the predictions of nearly 100,000 forecasters underpinning each signal, all effectively contributing their market sentiment to the predictive engine, the Cindicator platform turns this advantage into a reality.
But just how many tokens are required to truly benefit from the wisdom of the crowd?
5000.
In January of 2018, I set my sights on obtaining 200,000 CND Tokens in order to attain “Trader” level. I had hoped to attain the 200,000 tokens prior to withdrawing from the Binance exchange, thus incurring a fee. With around 40,000 tokens in my possession at the time, and prior to the introduction of the “Explorer” Tier (30,000 CND), rather than take advantage of the “Beginner” Tier afforded to owners of at least 5,000 tokens, I made the stubborn decision to withhold from initiating the CindicatorBot. My sentiment, at the time, was bullish, like many others — and while discussions were rife with expectations of an impending correction, I had no statistical basis for determining when one might occur.
Then — it happened.
For the last 10 days, the total market capitalization of all cryptocurrencies is down from $800B to around $500B. Will the total capitalization of cryptocurrencies manage to recover and climb above $800B by February 20, 2018? (results will be according to coinmarketcap.com)
On January 18th, the above alert went out to all users of the CindicatorBot indicating an 83.11% probability that the market would recover by February 20th. While the indicator ultimately expired inaccurate, in the right hands — this information may have signaled the worst had yet to come. Again, referring to Thorpe’s words in 2001,
“history shows us that more times than not the market will go against the majority. Extremely bullish levels of sentiment often come after strong market run-ups when investors are fully invested in the market. Even if they are bullish about the future, they have limited additional resources to invest.” (21)
Had my tokens been staking, and I received the January 18th indicator, would I have interpreted the high probability — undoubtedly the result of extreme bullish sentiment among forecasters — as a sign of a continued downtrend? Most likely not. My sentiment was similarly bullish. But despite the inaccurate result, the indicator was useful in that it provided a snapshot of the prevailing sentiment at that moment in time — and as suggested by Thorpe,
“If you see the survey reaching all-time levels for either bearish or bullish, you should expect that a market is probably nearing a bottom or top, respectively.” (22)
Shifting market conditions
Six days after the January 18th Indicator, another Market Events Probability indicator concerning the cumulative cryptocurrency market capitalization was sent to Expert Tier users.
The cryptocurrency Market Capitalization has fallen by USD 55.7B to USD 506.8B in the previous 24 hours until 6 AM UTC January 23. In your opinion, will the Market cap fall below USD 400B at any time before February 28?
Indicator:
65.25%How do I interpret the results?
[65%...80%]- a high probability indicator; the probability of the event happening is high
Notably less extreme, this indicator clearly communicates the shift in sentiment among forecasters in response to the continued sell off in the cryptocurrency markets. However, only after examining the following several indicators regarding the cumulative market capitalization does an intriguing truth reveal itself.
The following indicators were sent to all CindicatorBot users on February 5th, and 6th respectively.
The cryptocurrency Market Capitalization has fallen by almost USD 50B in the previous 24 hours and settled at USD 400B at 8 AM UTC on Monday, February 5. In your opinion, will the Market cap drop below USD 325B at any time before February 15?
Indicator:
72.00%How do I interpret the results?
[65%...80%]- a high probability indicator; the probability of the event happening is high— — — — — — — -
The cryptocurrency Market Capitalization has fallen by almost USD 50B in the previous 24 hours and settled at USD 400B at 8 AM UTC on Monday, February 5. In your opinion, will the Market cap climb above USD 500B at any time before February 27?
Indicator:
75.00%How do I interpret the results?
[65%...80%]- a high probability indicator; the probability of the event happening is high
While seemingly contradictory, the eventual occurrence of both events lends validity to the assumption that the change in market conditions paved the way for prevailing sentiment to once again drive price action within the markets.
Revealing “true” accuracy, and evolving predictions
At this point in time it is important to recall precisely what the Cindicator platform offers. As clearly, and eloquently phrased in the FAQ section of the Cindicator Website, “each indicator or index is not an unambiguous trading signal, but only an additional metric in the market that helps in the analysis of an investment decision.”
As such is the case, by paying close attention to these additional metrics over time, it stands to reason that one can learn how to more effectively interpret indicators, and subsequently utilize them in their decision making.
This became my goal. To learn how to best interpret indicators, and when to act upon them.
Having attempted to accumulate tokens in the 20–30 cent range, I understood just how quickly one could get priced out of the higher tiers (as the above chart reveals), and after noticing Cindicator had experienced a significant pullback to the 6–7 cent range, I began working toward accumulating as much as possible — selling my mining equipment in the process.
Now, I don’t profess to be an expert trader myself, nor do I claim to be an expert in utilizing indicators in my trading — quite simply my endeavor began as the result of a desire to track the accuracy of predictions in real-time. As I started to focus more attention on the results of expired indicators, I began to wonder just exactly what “accurate” meant.
Examining all 13 indicators concerning the cumulative market capitalization included in the released 2018Q1 data reveals that 11 (85%) indicators were accurate, with only two “inaccurate” predictions. The January 18th indicator, and the following Beginner Tier Indicator sent on March 15th, 2018.
The cryptocurrency Market Capitalization settled at USD 312B at 6 AM UTC on March 15. In your opinion, will the Market cap rise above 360B at any time before March 22?
Indicator:
65.52%How do I interpret the results?
[65%...80%]- a high probability indicator; the probability of the event happening is high
At this point, I had begun keeping track of the accuracy of indicators in real-time, creating a prediction log to monitor outcomes of the “Explorer” Tier— and I had started to notice something. Certain high probability indicators, while the events themselves did not occur, came very close to the targeted values. I decided as part of my accuracy analysis to quantify in percentage terms how close certain “inaccurate” indicators came toward reaching their targets. In the case of the above indicator, according to coinmarketcap.com the cumulative cryptocurrency market capitalization peaked at approximately $3.56B on March 21st — one day prior to the expiration of the indicator.
Using a simple percent change calculation, where V1 is the actual value, and V0 is the target value, we can quantify how close the actual value came to the target.
Plugging in the 360B target and the peak market capitalization of 3.56207B (3/21, 15:32 UTC) between March 15th and March 22nd, the formula is as follows.
The math indicates that the peak market capitalization during the time-frame of the indicator came within less than 1.1% of the predicted target, and while inaccurate on a binary scale, the close proximity to the targeted value is striking. As far as Cindicator’s published accuracy statistics, proximity to the predicted values is not considered. The only truly inaccurate indicator concerning the cumulative market capitalization, was the extremely bullish indicator preceded by Bitcoin’s price falling below $10,000 for the first time since early December, when sentiment was near all time highs.
As I gathered more data points in my prediction log, I began to perform a similar accuracy analysis on Weekly Support & Resistance indicators. This time, I wished to see how far off the predicted targets were from the actual high and low trading prices of the week. Between March 19th, the day I began actively tracking indicators, and April 19th there have been 30 Weekly Support & Resistance Indicators, each consisting of a high and low price prediction for the week, for a total sample size of 60 predictions. On Average, these predictions have come within 7.99% of the actual values, a relatively low number where cryptocurrency markets are concerned.
Using this data set, I captured a number of metrics pertaining to the relative accuracy of Weekly S/R indicators, with the goal of potentially finding ways to further improve the predictions.
Depending on whether historically, the majority of resistance predictions for a particular asset were under or over-estimated, I thought to adjust future resistance levels by the average under/over-estimation, and perform a similar adjustments to support predictions. Testing this methodology on previously expired indicators resulted in an increase in average accuracy, but this was not unexpected, as the improvement metrics were sourced from accuracy statistics pertaining to those same indicators. Real proof of success would have to come from improving the accuracy of open indicators.
Using the above metrics, I came up with the following six basic adjustments and tested them on the open Weekly Support & Resistance predictions at the time.
[#WEEKLY]. The cryptocurrency Litecoin settled at $139.28 at Bitfinex exchange at 9:00 PM UTC on Wednesday, April 18. What will be the minimum and the maximum price of LTC/USD from 12:01 AM UTC on Thursday, April 19 until 11:59 PM UTC on Wednesday, April 25?Resistance level:
150.74937 * (1 + 13.38%) = 170.91963, ACTUAL = 165.94
Support level:129.63429 * (1 + 6.69%) = 138.30682, ACTUAL = 138.96
[#WEEKLY]. The cryptocurrency NEO settled at $72.55 at Bitfinex exchange at 9:00 PM UTC on Wednesday, April 18. What will be the minimum and the maximum price of NEO/USD from 12:01 AM UTC on Thursday, April 19 until 11:59 PM UTC on Wednesday, April 25?
Resistance level:76.71 * (1 + 6.57%) = 81.74984, ACTUAL = 84.44
Support level:68.5 * (1 + 7.69%) = 73.76765, ACTUAL = 69.03
[#WEEKLY]. The cryptocurrency Cardano settled at 0.2596 USDT at Bittrex exchange at 9:00 PM UTC on Wednesday, April 18. What will be the minimum and the maximum price of ADA/USDT from 12:01 AM UTC on Thursday, April 19 until 11:59 PM UTC on Wednesday, April 25?
Resistance level:0.2975 * (1 + 10.14%) = 0.32766, ACTUAL = 0.31869
Support level:0.24 * (1 + 9.20%) = 0.26208, ACTUAL = 0.26081
On their own the indicators were accurate within 6.74% on average, whereas the adjusted predictions achieved accuracy within 2.80% on average, with both the LTC and ADA adjusted support predictions accurate within less than one half of one percent. Even in the event that buy/sell orders are not triggered based on predicted prices, the above data suggests that over time, with more data, users can be reasonably certain that Cindicator’s price estimates will typically be accurate within a specified range. This will enable users to stay one step ahead, particularly at the “Trader” Tier which adds BTC, ETH, and XRP Weekly Support & Resistance Indicators, along with Weekly S/R indicators for traditional assets such as Gold Futures, Crude Oil, and E-Mini S&P 500 Futures.
With such impressive results at such an early stage, I’m confident the best has yet to come. With the neural network (which correctly predicted the outcome of the January 18th Market Capitalization indicator) switched to production as recently as late March, and the frequency of indicators increasing by the day, there is much to look forward to. Ultimately, indicators are useful only to the degree that a user enables them to be, however there is undeniably value inherent to a “real-time sentiment gauge” and the provision of quantitative market data. So while there’s “no such thing as a free lunch”, Cindicator might just be the next best thing.
Hopefully this post has provided some level of depth into how indicators may be interpreted in less obvious ways, and I thank you all for reading.
Access a copy of my Prediction Log here (I’ve removed any open indicators, which may impact some of the calculations).
