Gustav and Xardas,
I understand your point of view and I think it actually makes sense. In a sense, anyone who publicly picks stocks can alter preferences, and due to that, could potentially increase demand/prices of recommended stocks. The point is valid. That said, the formula should be altered repetitively (and regressions rerun repetitively) in order to adjust for new historical data. If I do post the regression results, it will be in an academic journal.
But, getting back to posting results and boosting demand, I suppose that's an additional reason why this "could" work. But, what I was aiming to do was really test the theory here. Nothing more. I wasn't willing to toss too much money into this, and as the guy who created the model saying that I wouldn't bet on it (yet),... it would be crazy to invest your own money. Thus far, some days have worked, others not so much. But, on net, I think it's working to a degree, as winners are up. To give a bit of insight, I feel my formula is pretty much based on economic theory and swings. When I say swings, I mean it's not surprising that stocks (or in this case altcoins) that have huge swings would revert back to a mean. Since, it's a panel regression model, this is particularly true because we can measure "relative" swings from coin vs coin in variables. I use the variables on coinmarketcap.com to construct the model.
Like I've said, I'd love to figure out how to run the model consistently and do it in an easier fashion. Meaning, I'm a statistics/econ guy, not a programmer. The regressions are easy. But, I'm manually pulling the data off coinmarketcap.com. Then, I'm adjusting the data, which is labor intensive. Then, I'm plugging in the forecasted model to create a price change for each coin. The process is a pain in the ass and it's becoming more complicated as new coins enter the market. This is one reason I haven't been able to really keep up with it as much as I would have liked.
Ideally, I'd be the regression guy. That's it. That's where my experience comes from. The data pulling, the programming, etc... I'm not efficient with that. There's probably a way to do this in an easy manner if the community really wanted to get behind it. I mean, the truth is it could be automated. But, who knows if that will ever happen. Time will tell there.
please link us the articel here as soon as it is published. Since i think pricechange depends on a lot of factors of which not every one is a quantitative one. Since prices are multicausal the formula should become better the more influences are captchered in it. For example: Buy/sell depth on the books and how exactly it is distributed and changes is one thing. History data is another thing, then the overall mood of the market, then the blockrewarddecrease, then the change in networkhashrate and other thing. Then when you got it running and are giving out predictions regularly and gain attention you would have to factor in also the influence of your predictions (maybe deriven from views per day of your website) - there is many factors to consider. If you can catch the bigger picture it should be possible to get something useful from it. I will be watching this thread and waiting for the published first formula. Good luck with this.