RE: STEEMIT || AN EXPONENTIAL SMOOTHING PEEP INTO THE FUTURE

Repository

e.g. https://github.com/utopian-io/utopian.io

Introduction

Sometime last year I did an analysis using exponential smoothing to try to predict the future of Steemit. I covered areas such as new accounts that would be created, number of posts and distinct authors, number of votes and distinct voters and payout forecast. The potency of any prophesy or prediction is to check the degree of accuracy of the predicted indices.

In case you missed the initial post you can read about it here

What is Exponential smoothing?
It is a statistical technique used for time series forecasting and it is based on past observations. The past observations are weighted and used to predict the coming months

Outline

• Scope
• Results
• Conclusion
• Tools and Scripts

Scope

Date of the analysis: 10th April 2019

The time frame of the analyzed data: August 2018 to December 2018

Components of the analysis: checking data from the database to compare with predicted data.

Results

ACCOUNT CREATED.

DATE	ACCOUNT FORECAST	ACTUAL ACCTS CREATED	PERCENTAGE DIFFERENCE
Aug-18	48,775	22,325	-54%
Sep-18	41,822	25,442	-39%
Oct-18	56,763	21,421	-62%
Nov-18	50,541	14,671	-71%
Dec-18	63,324	3,866	-94%
TOTAL	261,224	87,725	-66%

In 2017, the number of new accounts created monthly was always fluctuating. People were just joining from recommendations of existing users. The initial evangelist of Steemit always used lines like social media that pay.

Steemit came with a lot of hype, a hype that didn't last long. With the price of cryptos going down, it was now clear why many joined the platform. They didn't join because they wanted to change the world, the joined because they wanted to make some money for themselves.

There were insinuations among crypto enthusiasts that the price of the cryptocurrency would bounce back to their former glory. Everyone believed it; including me. But it was not to be, the price of steem and other cryptos never recovered.

On this forecast, it fell far from the actual figures. The prediction deviated by about 66% of the actual values.

NUMBER OF POST

DATE	POST FORECAST	ACTUAL NO OF POST	PERCENTAGE DIFFERENCE
Aug-18	933,766	495,921	-47%
Sep-18	878,518	357,923	-59%
Oct-18	1,007,437	299,233	-70%
Nov-18	1,149,119	429,758	-63%
Dec-18	1,093,870	671,330	-39%
TOTAL	5,062,710	2,254,165	-55%

One of the ways of checking how active or interactive a social media platform is, is by checking the number of posts made by users. The number of posts for August to December 2018, saw fluctuations in varying degrees.

My observation was that as the price of steem dropped, users were reluctant to make a new post as they seem not to get value for their content.

The prediction deviated by about 55%

DISTINCT AUTHORS

DATE	DISTINCT AUTHORS FORECAST	ACTUAL NO OF DISTINCT AUTHORS	PERCENTAGE DIFFERENCE
Aug-18	79,044	47,364	-40%
Sep-18	79,670	36,534	-54%
Oct-18	89,698	27,254	-70%
Nov-18	86,228	42,655	-51%
Dec-18	90,024	69,721	-23%
TOTAL	424,665	223,528	-47%

With the number of posts reducing drastically, it is only logical to find out that the number of distinct authors also dropped.

The forecasted values deviated by 47%

VOTES

DATE	FORECAST VOTES	ACTUAL NUMBER OF VOTES	PERCENTAGE DIFFERENCE
Aug-18	20,275,735	21,311,964	5%
Sep-18	20,592,975	17,022,727	-17%
Oct-18	21,131,906	17,523,597	-17%
Nov-18	21,359,495	15,930,248	-25%
Dec-18	24,360,111	18,721,406	-23%
TOTAL	107,720,223	90,509,942	-16%

This was the best performing prediction. The prediction deviated by merely 16%. Voting on Steemit is the easiest activity a user can carry out, it takes less than a second. People vote more than any other thing on the platform.

DISTINCT VOTERS

DATE	DISTINCT VOTERS FORECAST	ACTUAL NO OF DISTINCT VOTERS	PERCENTAGE DIFFERENCE
Aug-18	125,214	107,666	-14%
Sep-18	124,987	99,830	-20%
Oct-18	139,010	86,679	-38%
Nov-18	135,134	74,949	-45%
Dec-18	141,552	93,032	-34%
TOTAL	665,897	462,156	-31%

Distinct voter means the number of different voters that voted within a period of time. Just like we saw in the number of votes made in the period under review, the number of distinct voters was not also far from the predicted values. The prediction deviated by about 31% of the true values.

TOTAL PAYOUT

DATE	TOTAL PAYOUT FORECAST	ACTUAL TOTAL FORECAST	PERCENTAGE DIFFERENCE
Aug-18	906,644	880,937	-3%
Sep-18	2,837,349	752,166	-73%
Oct-18	3,652,936	582,088	-84%
Nov-18	3,597,132	1,326,118	-63%
Dec-18	3,001,022	1,358,370	-55%
TOTAL	13,995,083	4,899,678	-65%

For me, this is the backbone of Steemit. There is nothing that affects activities on Steemit like the price of Steem. Over time it has been proven that there is a direct correlation between the price of Steem and the amount of activity on Steemit.

Cryptocurrency is among the most volatile markets we have, a steep increase today can be followed by another steep decrease tomorrow. No one really knows the tomorrow of cryptocurrency and as such

Conclusion

The prediction didn't turn out as we all would have loved. The numbers are no longer encouraging. Users have deserted the platform and other similar social media websites have emerged, creating a competition. These platforms have come with different approaches and most try to solve the problems that people have always complained about on Steemit.

Steemit is now facing competition from other platforms in terms of revenue and sustainability. I believe it is this pressure that has led Steemit to now accept ads, something we never thought would happen.

Can Steemit return to its glory days? I don't think so, but with some changes, the right people can return and find their way around.

The fall of crypto currency did not also help the cause of Steemit. In the future, predictions would still be difficult because of the many factors involved in making the platform a very volatile one. One which is prone to unpredictable variables having a big say.

Tools and Scripts

Microsoft SQL Server Management Studio 17 was used to access the data from STEEMSQL ( a publicly managed database by @arcange )

Microsoft Excel 2019 was used to plot graphs and Charts.

Scripts

For number of votes

SELECT count (txvotes.voter) AS Votes, count (distinct txvotes.voter) AS [distinctVoters] FROM TXVOTES WHERE timestamp>= CONVERT(datetime,'08/01/2018') AND timestamp< CONVERT(datetime,'12/31/2018') GROUP BY MONTH (txvotes.timestamp)

For accounts created

SELECT count (accounts.created) AS [ACCTS] FROM ACCOUNTS WHERE (Convert(date, accounts.created) >= '2018-08-01' AND Convert(date, accounts.created) <= '2018-12-31') GROUP BY MONTH(accounts.created)

For number of post, Distinct Authors And Payout

SELECT count(comments.created) AS [POST], Count(distinct Comments.author) AS [DistinctAuthor], sum(CONVERT(REAL,Comments.total_payout_value)) AS [TotalPayoutValue], sum(CONVERT(REAL,Comments.curator_payout_value)) AS [CuratorPayoutValue] FROM COMMENTS WHERE (Convert(date, comments.created) >= '2018-08-01' AND Convert(date, comments.created) <= '2018-12-31') and DEPTH = 0 GROUP BY MONTH(comments.created)

Proof of Authorship

https://github.com/jingis07