The network value experienced a similar 90-95%+ drop in 2016 long before linear, and also at that time it was also greatly underperforming other cryptocurrencies.
It was also at the time that broad awareness of Steem was likely close to its peak, and was certainly very high, as it was the first social network-connected blockchain or cryptocurrency and people were notable people from both within and without the cryptocurrency world were flocking to it in part due to the high and highly visible payout values. Dan and Ned were featured on many well known podcasts, etc. One can even see the many attacks on Steem as a scam or ponzi scheme as a sign of its (short-lived) success; nobody even bothers to make these attacks now. At #70+ in market cap ranking, it's not even worth attacking.
And yet, in 2016, long before linear was even considered, it not only lost a tremendous amount of its value, but it fell in value to the point that did not suggest investor believed in there being a lot of promise for it. (Nor, I might add, did apparently Dan, since he left for something which, we can all see with the benefit of hindsight, has indeed been far more promising.) The decision to switch to a linear model, however we may feel about it now, did not happen in a vacuum, it was made after seeing the superlinear model fail to perform in driving the growth and retaining value.
No one has demonstrated either a solid descriptive or predictive model nor empirical evidence as for why people would buy Steem in order to influence eyeballs. There are gaping flaws and leaps of faith in any argument I have ever seen including the original white paper.
This does not mean that Steem can't become popular or increase in value, but these are still very different question, and I don't really attribute the lack of success Steem has ever had (regardless of payout curve) in convincing investors of its promise to the 'eyeballs' issue in and of itself. I attribute it to lack of any convincing story for why Steem should succeed and accrue value. In that I largely agree with @lukestokes.
Probably the closest I've ever seen is the "secret plan for world domination" post that @lukestokes likes to reference (and he is not alone). Indeed that post also has very little to do with people buying Steem to influence eyeballs. It is more about growing a large community with users and a variety of useful apps (including smart contracts) which uses Steem as its currency platform. Unfortunately we got really distracted by this whole "paying content creators" concept which was supposed to be a means to an end. When it became largely an end in and of itself and a primary focus, things went downhill from there.
I disagree and many metrics disagree with your opinion. Superlinear reward weren't given enough time but aren't fundamentally flawed like linear. Don't get me wrong, I was super aware of the collusive voting going on to game the curation but it couldn't be perfectly gamed as the collusion would always profit some more than others.
I bought Steem under superlinear to influence the platform knowing other people would compete to do the same. Currencies are just that, store of influence, store of energy. Whether or not people understand this doesn't make it less true.
We do indeed disagree on superlinear (as implemented in Steem) not being fundamentally flawed or even any less flawed than linear (I would say in some ways more flawed, but yes both are definitely flawed). It was just not fundamentally flawed in exactly the same way as linear.
Name one. There is no evidence in terms of actual metrics of Steem being on a success trajectory in 2017 after about a year of superlinear. It was cratering in terms of user base growth, retention, web ranking of the main site, etc. and there is no data to support that more time would have done anything but continue the trend.
I'm not saying there is a causation in regard to those 2 metrics but these 2 fundamental trends were up before the linear rewards were introduced.
Linear rewards were introduced in June 2017.
What you are missing is that for several weeks prior to the linear hard fork, linearity (or at least a very strong attenuation of n^2) was largely in effect already due to the 'whale experiment'. Whatever gains you are seeing there, to the extent they have any tie to the payouts, would correspond with getting rid of n^2, first by whale voting changes and later by the fork itself.
Also, IIRC one of the forks (I don't remember which one) botched the payout pool logic and resulting in there being hardly any payouts for a month or so. That may also have helped the price, but can hardly an endorsement of n^2, or even paying rewards at all.
These are your assumptions.
The whale experiment isn't linear rewards. These 2 are different.
Yes, those are my assumptions because that was a premise upon which the whale experiement was based (abit did some math and determined how low votes needed to be to largely stay out of the most superlinear part of the curve). The overall effect, while not perfect, was to dramatically reduce the vshares in the pool and therefore increase the weighting on smaller payouts. It is very similar to the effect of linear or convergent linear.
The biggest problem with the whale experiment was the high cost of the downvotes, and indeed that is probably the biggest flaw in the original white paper design, and likewise also in the assumptions that went into switching to linear (that people would downvote, retaining what is effectively a non-linear curve, and avoiding the self-rewarding issue with linear raised by the white paper). Since downvotes are too expensive, that didn't happen (more then 0.01% of the time at least). Everywhere you look, expensive downvotes are a problem.
With cheaper downvotes, any curve works a lot better. Without them, any curve works poorly.
Whether the upcoming changes will make downvotes cheap enough remains to be seen. Social stigma, retaliation, and uncompensated effort may still impose too high a cost.
It was very different. The day the fork happened the change in the payout made this very clear.