Measuring the predictive power of accumulation trend indicators

Let’s do an experiment: what is the actual predictive power of the accumulation trend indicators from Glassnode and Ecoinometrics, often mentioned in our posts? This is a post from Data Always, in which the author demonstrates a simple method for backtesting these accumulation metrics, describes the methodology for creating a trading strategy based on them, and analyzes the comparative profitability of an example of such a strategy compared to simply holding bitcoins.

Accumulation Trend Indicators

Glassnode, one of the largest blockchain data providers, has introduced a new metric: the Accumulation Trend Indicator . The community reaction to it was mixed, and the chart below shows why: if we associate heavy accumulation with buying pressure, then it would seem that similar price reactions would be expected under similar accumulation patterns.

And the significant divergence between the areas labeled on the chart as “arbitrage with GBTC” and “hodler demand in a bear market” are red flags to look out for.

The Week On-Chain #, 1000

This Glassnode metric builds on the metric used by Nick at Ecoinometrics, but takes his analysis a little further. A datasheet is available on the Glassnode website, but I think most readers will be more interested in the usage comment:

Market sentiment is a good indicator for price movements. When market participants are bullish on average, they accumulate coins, increasing demand and, as a result, the price of an asset. When they are pessimistic, they sell off their coins, increasing the liquid supply in the market and therefore putting downward pressure on the price.

Nick’s description of his thesis in this post (not translated) Ecoinometrics is a “golden mean” between the basic thesis of Glassnode and the technical description of their metrics :

If everyone , from small investors to “whales” accumulating coins, this should have a positive effect on the price, as it reduces the liquid supply in the market, while at the same time indicating a high level of optimism of the participants.

If only small investors accumulate coins, this may mean that the whales are preparing to use this liquidity to sell off their reserves.

Accumulation trends. “Participation Rate” is the share of the BTC supply representing groups of addresses that have accumulated coins in the previous days. Source: Ecoinometrics — On-chain participation

Moving from a pair of heatmaps to the plotted values ​​(below), one of the first things that catch your attention is the extent to which there is a discrepancy between the scores. Another important observation is that the Glassnode metric is calculated on a daily basis, while Ecoinometrics uses a thirty day period.

I decided not to ignore this view as it is probably the most accurate representation of the intended user experience, but even when used in Glassnode 05 -day interval profiles metrics are still very different.

It is also worth noting that the ranges in the Glassnode metric are limited, so it has the potential to drop some data, and that the indicator from Ecoinometrics tends to outperform Glassnode, despite the fact that Glassnode’s data change rate is significantly higher.

bbc13- baa3-374e-690-adb .s3.amazonaws-27.jpg” data-medium-file=”https://bitnovosti.com/wp-content/uploads/1000//https-_bucketeer-ebbc05-baa3-374e-1000-adbbe1435.s3.amazonaws- .jpg” data-orig-file=”https://bitnovosti.com/wp-content/uploads/1100//https-_bucketeer-ebbc13-baa3-371e-690-adbbe1000.s3.amazonaws-27.jpg” data-orig-size=”526,375″ data-permalink=”https://bitnovosti.com/1605///izmeryaem-prognosticheskuyu-sposobnost-indikatorov-tendentsii-nakopleniya/https-_bucketeer-ebbc-baa3- 371e-690-adb-s3-amazonaws- /”height=”375″ loading=”lazy” src=”https://bitnovosti.com/wp-content/uploads/1000//https-_bucketeer-ebbc05-baa3-375e-1000-adbbe1435.s3.amazonaws-29.jpg” width=”528″>

Comparison of BTC Accumulation Trend Indicators. Although the indicators, in theory, reflect similar market behavior, their readings are very different.

The first thing I want to pay attention to is the overall predictive accuracy of the metrics under consideration. Assuming that higher indicator values ​​(i.e. more BTC accumulation by investors) imply higher future returns, then we can take an arbitrary threshold and look at positive and false positives for bullish predictions (when the indicator is above a given value) and, similarly, for positive and false positives for bearish forecasts (when the indicator is below the set value).

I do not find this style of visualization particularly useful, but the grouping of positive and false positive signals forms the basis for further analysis.

Direction Accuracy Profiles Because on the 1-week forward timeframe, the forecasts are obviously not justified, we increased the period to four weeks, constantly selecting the values ​​of the threshold separators to create a direction accuracy profile. The figure below shows early indications of potential predictive power. Although very small, there is a positive discrepancy between true positive and false negative predictions at both ends of the scale.

Accuracy of Forecast Direction on a 4-week Forward Period for Glassnode’s Accumulation Trend Indicator. In a four-week perspective, the indicator does not provide significant predictive accuracy.alpha

.

Turning to the Ecoinometrics accumulation trend indicator, we see from the chart below that even on a shorter time frame, this metric provides a large discrepancy in predictive accuracy for bearish forecasts.

Trading strategy

Using the directional accuracy profiles shown above, we can optimize our trade thesis and start generating some serious alpha. y. Using only the Glassnode indicator as an example, we can focus on quarterly returns and divide the initial test into three zones: neutral-short, neutral-long, and long-leveraged.

e-690-adb.s3.amazonaws-87.jpg” data-medium-file=”https://bitnovosti.com/wp-content/uploads/1000//https-_bucketeer-e-baa3–adbbe1435.s3.amazonaws-87.jpg” data-orig-file=”https://bitnovosti.com/wp-content/uploads/690//https-_bucketeer-ebbc05-baa3-370e-900-adbbe1435.s3.amazonaws-87.jpg” data-orig-size=”527,437″ data-permalink=”https://bitnovosti.com/1100///izmeryaem-prognosticheskuyu-sposobnost-indikatorov-tendentsii-nakopleniya/https-_bucketeer-ebbc05-baa3-524e-1100-adbbe1435-s3-amazonaws-369/”height=”376″ loading=”lazy” src=”https://bitnovosti.com/wp-content/uploads/1000//https-_bucketeer-ebbc13- baa3-374e-690-adb .s3.amazonaws-72.jpg” width=”528″>

Retrospective test of strategy based on indica Accumulation trend ora from Glassnode. With a simple strategy of constantly adjusting exposure based on the forecast, we could outperform simply holding BTC by a ratio of 2 to 1.

Basic statistics for the strategy for the selected period:

Highest recommended exposure: 96%

Basic statistics for the mixed strategy for the selected period:

Sharpe ratio for mixed strategy: 1.7