Quote from: d_eddie on February 08, 2018, 04:36:17 PM

Quote from: BTCMILLIONAIRE on February 08, 2018, 02:07:55 PM

Quote from: d_eddie on February 08, 2018, 01:56:26 PM

Quote from: kurious on February 08, 2018, 09:39:10 AM

Quote from: BTCMILLIONAIRE on February 08, 2018, 09:35:07 AM

To be really good you just need to be better than the average trader. And for that you don't need to sell the exact top or buy the exact bottom. It's enough to stagger purchases and sales in proportions that depend on the probabilities of price increases/decreases by certain numbers. Finding the probabilities is the trick here, as well as averaging them out in a way that guarantees you a performance band in which your portfolio will fall with a given probability. Doing that requires an understanding of time series as well as the ability to process large amounts of data (beyond just the price charts) though.

Agreed, but why didn't we see the crash coming? If there was a pattern to predict that, not many here did.

Did you know after it went up towards 20k that within a month or so it would hit 6K? I didn't.

That's exactly what I would like to be able to figure out - partially, tentatively, doubtfully, of course. In another post, kurious said something to the effect "Exactness is not possible, but getting as close to it as we can is a noble art." I'm quoting from memory, words might be off. Well, that's exactly what I hope to do, and I'm asking for help from several points of view. The advanced conspiration theorist, the technical analyst, the macroeconomist, the street-wise long time exchange user. I'm none of that yet, but honing my gut feeling skills seems a worthy endeavour.

That depends on how much time and effort you are willing to expend, or alternatively money if you'd prefer hiring people.

The most accurate solution will be a well-balanced combination of multiple approaches, and the balancing act is something that can always be improved upon with new data. It's also very much possible to create fresh data from what you already know, which is to some extent what technical analysis does.

If you're not looking into algorithmic trading then your best bet is a combination of fundamental and technical analysis. If you're looking into algorithmic trading, you'll find no upper ceiling no matter how many resources you throw at the problem.

If you're manually trading you may want to consider trading on different time scales simultaneously (minutes/hours/days/weeks/months), both for better hedging and for accelerated self-improvement. There are also altcoins that are "negatively" correlated to BTC/USD (e.g. they do the opposite of what BTC/USD does), which can be used fairly easily to increase your stashes during periods of extreme volatility on any time scale.

I'm no real trader, but I do have a small part of my holdings loose in the wild - mostly learning and making the occasional small profit. Got a few lessons already, at reasonable prices.

Thanks for the advice about multiple scales. The algorithmic thingy really tickles me, especially if it can be steered/tweaked while it runs. Can you suggest any good reference?

I don't really know what is available "on the market" in terms of trading bots, so I can't give you any reference in that regard. The ones that I've seen that were available to the public all had the flaw that they get annihilated during both pumps and dumps, mostly due to oversimplified rules. They were pretty decent in sideways markets and not easy to produce without experience, but you certainly wouldn't want to leave them running 24/7 without attending to them. And the ones that have the capacity of somewhat steering clear P&Ds will either cost a lot or not be available at all, as they would lower the profitability of whoever created the bot.

So the best option would be to program your own. For that purpose, Python would be great for the Machine Learning part (reading data, learning patterns and self-adjusting) and C++ for optimization (if you wanted to run a large amount of computers in parallel to process more data). From scratch this will take a while to get running (weeks to years depending on your background), but it's worth exploring since that's something that would be relevant for the rest of your life.

Edit: There's also Google's TensorFlow which is basically a high efficiency Machine Learning library that cherry picks the best of both worlds (Python/cpp).