#project-smackdown - the project reporting weekly on the top self comment voters and flagging the top 20 self voted comments every day.

This is a summary of the first two weeks of the new algorithm, which I'm calling phase 2.

What is the new algorithm?

Firstly, the algorithm runs an analysis on every day as a whole day. For every user we look at every self voted comment in that day and add the pending payout for the self voted component, i.e. how much the user gave themselves for the each self voted comment. We add those together over the day, then divide by their SP. Users with low SP are ignored.

This is how we get what we call daily comment self vote return on investment, or daily comment svROI. The terminology comes from the idea that self votes do not add any value to anyone else in the voting system, and so the action is reduced to simply asking the system to pay you out a dividend, or a return on your investment, as the dividend is based solely on your stake and choice to activate it directing to yourself. It is a daily snapshot this regarding comments.

When the return on investment from self voting comment only is quite high I think we can call this out as unsustainable, and unethical on a system which is purportedly a social network.

To be clear, we are not against profit or gain from investment. But we oppose the use of Steemit as a money printing machine. The social aspect must be the means to profit, and outward votes should be the vote direction of higher value. As this is not currently the case due to the system being out of balance by our estimation, we have run this project to contribute towards realigning those incentives by providing a strong voice against comment self voting.

Summary - Daily Top 20

We record all comment self voters but only flag the top 20 each day.

Users by total svROI in this period

What we see is the usual expected high self voters peaking first which falls off fast, but then a pretty high middle area that does not fall fast. From about 0.5% svROI we see a steady decline to 0.08%.

Daily top 20 svROI charts

Here we can see the top 20 per day in context with all days in the two week period. There seems to be a slight decline as the days go on, but there are 4 noticeable large outliers, 2 which are especially large.

I won't go in to details, but all outliers are caused by one account voting a lot for themselves in one day. These accounts were, with svROI for just these days:

• on 17-08-11, @hjk96 with 0.939%
• on 17-08-12, @bhightech with 0.396%
• on 17-08-14, @ika497 with 0.886%
• on 17-08-21, @jiahn with 0.329%

Summary stats for each day

The max and min are in blue and green, and the average and median in orange and red respectively.

The outliers are clear here too in blue, and the have some effect on the average of course. The median is steady as is the minimum.

Summary stats for each day, average and median only with linear trend

We can see that in general that most daily comment svROI values are around 0.15% and dropping to 0.1% as the weeks go on. This is not enough data to confirm this trend but it seems to be clear enough. The linear trends are shown as lines above the graph.

Number of days in daily top 20 list

Nearly half (43%) of all users that we identify as having the top 20 svROI are only identified once in this two week period. But the exact same number of users are in 2 or 3 times (27% and 16% respectively). That means that again, nearly half of the top 20 comment self voters by this svROI metric have this behavior twice or more in this period. The remaining 14% appear even more than that.

Interpretation

Since we're only looking at the top 20, we can't read much into this for the population at large, except to say this is representative of the "worst" by our svROI metric.

The cause is unknown, but there is is a downward trend in svROI over the period of time. This would correlate with what we saw in the last report, a slow but overall downward trend in comment self voting. Again, I would like to take credit or a part of this, but we can't know the real cause.

Thoughts

It's clear that because we are adjusting for SP, we are seeing more users of lower SP on the list. We do see a variety, but lower SP users are disproportionately represented. It stands to reason that the amount of SP you have affects your behavior quite a lot.

My guess is that because lower SP users get less per self vote that users with larger SP, they will do it more as they are getting a few dollars at most. Those with a lot of SP can get many tens of dollars from just one large up vote. So perhaps it is here that the change in algorithm to svROI is shown the most.

Users

Top most self voting user per day

Date	User	svROI total (annual projected)	Overall position	SP
17-08-09	@yumyumyum	0.319% (8.292%)	37th	1,205
17-08-10	@greensmile	0.43% (11.171%)	21st	11,744
17-08-11	@hjk96	1.01% (28.561%)	**1st**	4,861
17-08-12	@bhightech	0.396% (10.298%)	26th	1,466
17-08-13	@cryptokraze	0.566% (14.715%)	**8th**	1,016
17-08-14	@ika497	1.003% (26.084%)	**2nd**	1,035
17-08-15	@avilsd	0.214% (5.552%)	64th	6,184
17-08-16	@dodders007	0.456% (11.852%)	16th	2,305
17-08-17	@sanevys	0.694% (18.046%)	5th	1,358
17-08-18	@donaldtrumpfan	0.158% (4.115%)	79th	25,693
17-08-19	@gotoperson	0.278% (7.22%)	48th	13,956
17-08-20	@nigelmarkdias	0.164% (4.269%)	76th	2,152
17-08-21	@jiahn	0.329% (8.553%)	34th	1,603
17-08-22	@sikklel	0.644% (16.732%)	**7th**	10,720