11–20 of 42 posts
This is OK as a final project in an intro econometrics class, but nothing else.
Earlier quoted context omitted.
minutes or seconds would be more useful than one hour intervals.
This is assuming no trading fees and highly liquid exchanges with low spreads, otherwise the profit from these micro-movements could be eaten up. I see no problem with automatic trading over periods of hours if you're getting >60% accuracy.
They a look at the sampling distribution of your projected P/L assuming each trade is an independent Bernoulli trial with success probability 60%.
This is OK as a final project in an intro econometrics class, but nothing else.
It looks like it is a class project. But so what? It's giving you enough information to reproduce the study, which is more than can be said for many published research papers. That makes it useful imo.
If this is supposed to give insight to people on how to allocate their money, the critiques would have to be much harsher.
Did it successfully predict the fall after January 2018?
As a result, it's important to develop a trading approach that can actually capitalize on the information. For that, I have found three things to work best:
1. Only trading on the highest signals of increase within a model that is a spectrum rather than binary classification. This usually doesn't increase accuracy much ironically but does increase the "average value" of buying on the increase signals. I usually set this through historical testing for prediction values and taking a top percentage of the prediction values to set the "threshold".
2. More features and feature selection tuning. Right now I'm using genetic algorithms to constantly try and test new sets of features, thresholds, "hold times" after buying, etc.
3. Work on minutes, not hours. The volatility is so high that you can actually capitalize well on the micro level in my experience.
While accuracy is important, the average trade value and trades per day are far more important to returns.
Interestingly enough, the algorithm was steadily making money until April or so, when it stagnated. Mind you, it was making money from January-March due to sheer volatility even while the price was dropping most days. I've actually shut mine down for two reasons - the plateau plus the fact that the market was too thin on GDAX to quickly trade on buy signals for the amount I was running with (ending at about $3.5K). If the market thickens, I'll likely start running it again.
Takeaway: this paper's approach may seem simple but honestly the reality is that with something so volatile it's surprisingly easy to capitalize on with algorithmic trading that learns even a few small features and trades frequently.
Interesting that even the most naive methods still have >50% accuracy. Also interesting that the best method was better able to predict downward moves than upward moves, during a bull market. Any intuitive reason for this? Is there some reason the study doesn't include the post-December 2017 bear market?
All methods that you'll ever see have >50% accuracy, because if you find a signal with 50% accuracy.
Here's a bit relevant to this conversation:
> Previous work on predicting the directionality of Bitcoin prices has shown that significant signal exists in the price of the cryptocurrency. Hegazy and Mumford (2016) compute an exponentially-smoothed Bitcoin price every eight minutes; using the first five left derivatives of this price as features in a decision-tree based algorithm, they predict thedirection of the next change in Bitcoin price with 57.11% accuracy.
> Their results substantiate earlier research done by Madan, Saluja, and Zhao (2014), who found that by using the Bitcoin price sampled every 10 minutes as the primary feature for a random-forest model, they could predict the direction of the next change in Bitcoin price with 57.4% accuracy.
> An alternative model was used by Sebastian, Katabarwa, and Li (2014), who use the Bitcoin price sampled every minute as the primary feature for a forward-feed neural network. Their results suggest that this system predicts future Bitcoin price directionality with 60% accuracy.
The most glaring evidence that this entire paper is garbage is the fact that zero time is spent on putting these numbers (57.11%, 57.4%, 60%) in context. What do I mean by context? For example, observations like the fact that for the same dataset if you use a daily resolution and your prediction is always "up", you'll beat those accuracies. Obviously, the reason why this discussion is absent is because it's a lot harder than just dumping a dataset into sklearn.
Interesting that even the most naive methods still have >50% accuracy. Also interesting that the best method was better able to predict downward moves than upward moves, during a bull market. Any intuitive reason for this? Is there some reason the study doesn't include the post-December 2017 bear market?
> Interesting that even the most naive methods still have >50% accuracy. All methods that you'll ever see have >50% accuracy, because if you find a signal with 50% accuracy. Here's a bit relevant to this conversation: > Previous work on predicting the directionality of Bitcoin prices has shown that significant signal exists in the price of the cryptocurrency. Hegazy and Mumford (2016) compute an exponentially-smoothe…
Earlier quoted context omitted.
> Interesting that even the most naive methods still have >50% accuracy. All methods that you'll ever see have >50% accuracy, because if you find a signal with 50% accuracy. Here's a bit relevant to this conversation: > Previous work on predicting the directionality of Bitcoin prices has shown that significant signal exists in the price of the cryptocurrency. Hegazy and Mumford (2016) compute an exponentially-smoothe…
I presume that the context of those accuracy numbers is prediction for the next period (i.e. 8 minutes, 10 minutes, 1 minute). To me 60% sounds good, but apparently not to people in this thread :)
That's not a context. That's the statistic they're calculating.
> To me 60% sounds good, but apparently not to people in this thread :)
What does "good" mean, and compared to what? If you can't put a number on "good" you're flying blind.
But hey, I'm just trying to warn people that the maths are garbage here; but if that looks good enough for you go ahead and trade it.