- November 22, 2019 at 3:04 pm #6468
Just when it looked like crypto was becoming boring…yeesh!
And I’m pleased to see the new Bitcoin Anomaly model doing its job so well. It kept us out of the markets as things slid, then caught a great bounce today.
It’s been live for just under 4 weeks, but it’s already up 15% while the markets are down at -13%. Trade History is updated (see the BTC Beta tab): https://crypto-ml.com/trade-history/
I’m also pleased to see our Market Index continuing to provide some good big-picture guidance.
What do you guys think of these markets? What’s next?
- November 23, 2019 at 9:58 am #6470kwParticipant
So far so good with the beta model! Just as the market starts to get boring, that’s when it gets exciting again. In my mind, BTC price may go lower from here, but I feel like we are in the process of a drawn out trend reversal, from bear to bull. I fully expect the bull marked to start coming back in Q1 2020 and into the halving. As BTC stabilizes, I expect ALTS to have a little run before BTC goes full bull. Could be a good time to make a little BTC.
My end-goal, and one which I know you guys are working on, is to automate my portfolio trading 100%. By doing that, we can gain one of our most valuable assets, time. The other part of it is, as a human, I am emotional by nature and can sometimes go against my plan, which usually doesn’t work out too well. By leaving everything to a machine, it fully takes emotion out of the equation.
So for the long-term future of Crypto ML, what I’d like to see is full automation, no thinking on my part. It could decide whether to take trades, or sit out of the market. It could decide which trades to take, and which pairs. I understand this a lofty goal, and probably couldn’t happen overnight.
I’d like to see additional support for more of the Coinbase coins (e.g XRP, LINK, REP, XTZ, EOS, ZRX, BAT, XLM) because I believe these sometimes present opportunities to gain BTC. It sounds like the 4.0 model could work well without much tuning or effort.
Something else to think about, is that during a bear market, my goal isn’t necessarily to gain value in USD, but in BTC. So it might be nice to choose whether you are trying to gain USD or BTC and optimize against that. Perhaps it roughly works out the same because they kind of go hand-in-hand, but maybe not.
- November 23, 2019 at 10:40 am #6471Rob DParticipant
That btc-beta run from the high 6000s was picture perfect. May they all be so successful!
- November 24, 2019 at 11:40 pm #6475matthew55Participant
Unfortunately the second run canceled any previous gain 🙁
- November 25, 2019 at 11:35 am #6478
@kw- This is really good information and a very good goal. I agree with everything you’re saying. Good perspective on the markets as well.
I sure hope a bull market is on its way. When I say “The markets…wow” I sure wish I was saying that due to a massive bullish move.
And during bear markets, you’re exactly right. It should be about asset protection. I can have the team look at it from a BTC value perspective and see what that nets out differently.
- November 25, 2019 at 11:40 am #6480
@matthew55, I had a feeling Murphy’s Law would kick in the moment I called attention to a nice win.
But now, it looks like BTC Beta is back in a good trade–we’ll see how it closes.
That’s the nature of the game though. And in bear markets, we really have to watch for this and manage our risk accordingly. The draw downs are so sharp.
- November 25, 2019 at 7:55 pm #6496
> And in bear markets, we really have to watch for this and manage our risk accordingly. The draw downs are so sharp.
Right. To that point, it would be awesome if the ML was able to comprehend that, sorta like @kw mentioned. The algo should know we’re in a bear market (according to the ML index thinger), so it should behave accordingly. I dunno, maybe protect profit quicker?
I mean, this could just be meddling with the machine, but one option could be a slider for how aggressive to be with profit taking. For instance, there could be a user defined thing that has the following if/else
if in long,
if no counter signal provided,
if % gain has hit n,
then issue sell signal.
though I guess that could be a whole lot of rework on the backend, cause right now it seems that the machine just blasts out the same signal to all the users.
Granted, yes, I could achieve the same thing by setting my own limit order manually… but whats the point of a bot, right?
hrm, as I type that I realize thats a who
- November 27, 2019 at 8:18 am #6562
Hey @martyf, yes–integrating the Market Index with the signals was a roadmap item. However, the Anomaly Detection model is aware of the big market picture and can adjust on the fly.
Even the 3.5 models have a bit more big-picture perspective–you’ve definitely seen them be pretty quiet recently.
But what’s interesting is we’re not really seeing stop losses or trailing stops tighten up–which you’d expect. As you said, this falls in the “don’t meddle with the machine” category, but in my opinion, there’s still work to be done here.
Your slider comment is interesting and got me thinking…this could eventually evolve into offering different models for different risk tolerances. By moving a slider, you’d be effectively toggling between different machine learning models.
- December 2, 2019 at 8:05 am #6638Rob DParticipant
Beware the tyranny of choice. More fun when we ride or die together anyways!
- December 2, 2019 at 8:35 am #6639
Is it just me or is the trade history page broken? I cannot view the results of other symbols like ETH when i click on the different tabs, it only shows for the default BTC symbol.
- December 2, 2019 at 12:31 pm #6647jessica.cryptoModerator
Hi there @Choco, thanks for reporting this issue. It should be fixed! There was some work going on with that page and apparently some browser incompatibilities that are taken care of now.
And @Rob_D, love it! I was really pulling for the BTC Beta to close near that $7700 level but it hasn’t yet hit the point of a negative anomaly. It has been close over the last 24-48 hrs, but just hasn’t crossed the threshold. Watching closely…
- December 2, 2019 at 5:56 pm #6658
Thanks for the fix! Really appreciate the quick response 🙂
- December 14, 2019 at 10:10 pm #6820
Does BTC Beta also employ hard stop losses just like the previous model do? (E.g 3.9% hard SL)
- December 15, 2019 at 3:01 pm #6842
- December 15, 2019 at 3:15 pm #6843
And to add a note to that… 6%puts the stop loss $6948. So that appears to indicate a break below the $7K level is adverse for near term.
As long as the trade moves up, the trailing stop would move up as well. That minimizes loss territory, but it hasn’t come into play yet on the current trade.
Also, we will be putting out bi-weekly updates (market analyses) starting in January. These will help cover some of these details. We’ll be looking at our tools, the market as a whole, and some other perspectives on market direction. I hope it will be a great value add.
- December 16, 2019 at 1:25 pm #6846
woof btc beta… giving up a lot on that last one.
- December 16, 2019 at 6:51 pm #6848
Hmm I’m not sure if btc beta should follow Crypto-ML Market Index where it still says bear market. Long/buys should only be made when the index says bull market. But then again what do i know about machine learning…hopefully it learns and gets better
- December 17, 2019 at 11:28 am #6851
Yes, agreed–that trade was a real bummer.
The problem seems to me that a stop loss of ~6% in these markets is too big. The awareness of the Market Index and general market direction should lead it to setting tighter stop loss bands. We are reviewing this to understand why it kept that level loose.
That said, looking at the charts, it seems as though it was positioning for a bounce off of the $7K level, with the loss below it in case the bounce didn’t happen. In this scenario though, it opened the trade a bit too early in my opinion.
So there is plenty to dissect on this particular trade and see how we can avoid a similar situation in the future.
- December 22, 2019 at 9:46 pm #6896AnonymousInactive
It looks like the BTC beta model is decent at predicting rises in BTC price, but the triggers predicting downtrends tend to be late and if solely following this model would be associated with losses. How can you best combine the beta with the normal model? Given that I am reasonably new to Crypto, should I decide on a value I should sell at or should I wait for metrics suggesting a downtrend is upcoming?
- December 23, 2019 at 6:53 am #6898
uh oh, bots trying to lose all the gains again 😛
but who knows. I guess we’ll find out. Is this the breakout we’re looking for?
- December 24, 2019 at 9:05 am #6903
Hey guys, the Trade History page is updated with that most recent BTC Beta trade.
In the two months it has been live, the model is actually up 7% while Bitcoin is at -13%. Not bad at all, but could/should be even better.
The biggest dings against it are the -7% and -6% losses.
Since then, this tweak was implemented to help avoid losses of that size: https://crypto-ml.com/community/topic/btc-beta-stop-loss-has-tightened/
And that’s actually what kicked in on this last trade and stopped it.
Also, losses are absolutely expected. It’s pretty surprising how the numbers work out over a long enough time frame. We actually break down the specific ratios of our models here: https://crypto-ml.com/bitcoin-trading-with-machine-learning-anomaly-detection/#Wins_and_losses
@futurescientist1234 the tweak I mentioned at the top of this reply is the right answer to your question. And, ideally, you would just follow the signal (rather than deciding your own value). But definitely everyone has their own trading approach.
If you’d like to read more about how it works, please check this out: https://crypto-ml.com/bitcoin-trading-with-machine-learning-anomaly-detection/.
And don’t hesitate to ask additional questions!
The replies here from you and @martyf are what helps us keep this thing moving forward and performing its best.
- You must be logged in to reply to this topic.