Performance of: ai-scan.rb (Round 2)
Previously, I posted a script that claims to predict new post payouts based on trending posts. The last analysis can be found here: Round 1
Again, I did this report manually. I also want to emphasize that these results are experimental. I also want to highlight the fact (as I've mentioned in the past) that this script does not take the author's name into account. It's looking for other factors to make the prediction. The first link above explains what those factors are.
Here's the output from Monday, Februrary 20th, 2017 at about 7:30 PM (UTC) (note, I'm running the script locally, but if you want to see these results live, they're on Ganymede too):
$ ruby ai-scan.rb
Predicting the following payouts will rise by:
20 SBD: https://steemit.com/history/@uwelang/today-in-sports-history-13-figure-skating-funny-review-february-20-1988
20 SBD: https://steemit.com/writing/@jacobtothe/warhammer-40-000-fan-media
25 SBD: https://steemit.com/steemitclinic/@araki/vertigo-stand-still-but-room-spinning-around-oo
29 SBD: https://steemit.com/life/@hanshotfirst/what-do-brand-new-parents-really-need-besides-sleep
38 SBD: https://steemit.com/deutsch/@shortcut/es-gibt-keine-wahrheit-ausser-dass-es-keine-gibt-german
47 SBD: https://steemit.com/photography/@sykochica/zoo-part-4-inside-the-herpetarium
All of the above posts had a pending payout of less than a dollar or two when I ran the script. They all had at least 23 hours left for the first payout.
So, what we're looking for is the resulting payout for these posts, once they get to that point. How close does my script get to those numbers after 24 hours?
To actually claim a prediction as accurate, the post has to make at least as much as predicted. None of them exceeded the prediction. But these results are a lot better than Round 1.
Conclusion
Like last time, the predictions aren't accurate at predicting the price. But unlike last time, there are no zero payouts. The goal is to predict trending, not "greater than zero," so the predictions are technically inaccurate.
On the other hand, if we use the prediction as a list for voting, this allows curators to get their vote in rather early.
Note: In this round, I skipped the analysis for narrowing the results on tags. They didn't seem any more significant in the first round.
cool @inertia
i don't know about other blog but your prediction just killed my blog, it reached 23$ but fluctuated with change in STEEM price .
i won't ask you how you do that because i know, cant understand it, but its COOL
let make make a prediction now based on hunch , your blog will hit more than 40$ :)
This is an interesting program...be cool to watch as it gets refined/trained.
It would have been pretty sweet to have made 47 SD, lol.
I've never really been on many whale radars, no biggie though. Makes me appreciate steem guild. Part of why I quit writing anything on steemit, politics or religion since those are disqualified from their voting. But I'd be curious to see what your program predicts when I do.