A crypto-trader’s diary — week 5
The Hacker Noon folks have been kind enough to give me my very own page for this series: hackernoon.com/crypto-traders-diary
I’m basically a movie star.
This week, like the one before, is a tale of two trains of thought:
- Risk management
- More resistance/support lines, but on the lower time frame charts
This week I’ve watched quite a few interviews with professional traders.
It’s cool to hear what smart people had to learn the hard way. So when I make the same mistake I can say hmmm, as it turns out I wasn’t smarter than that expert and really should have listened.
Many of these experts say that the vast majority of people who try trading will fail. 90% seems to be a common figure. And the most commonly-done-wrong thing is risk management.
This set of interviews with Francis Hunt is particularly great. The videos are, of course, brought to you by a company that has something to sell — but there’s no pitch woven into the interviews.
So my take on risk management is this: if I were going to the horse races for a bit of fun, with $100 burning a hole in my pocket, how much would I bet on the athletic abilities of any one horse?
I reckon $5 sounds about right (as a percentage, that’s roughly 5%). Not so much that I’ll be out of money in half an hour, but not so small that any win will seem more pointless than betting on medium-sized colourful mammals riding around on the backs of large brown mammals.
Another piece of advice I liked is to completely define an entire trade before I ‘place a bet’. E.g. “Buy bitcoin at $9,800. Sell if it goes as high as $10,780 or as low as $8,820”.
It’s all locked in, no emotion involved when it comes time to exit.
This is my kind of trading, because as much I like to pretend I’m not susceptible to human emotions, I’m just like everyone else — I cry when I see a tree with a broken branch, I laugh when I see a child fall over, and I will get excited about a trade and probably want to ‘hold on just a little longer’ or ‘cut my losses’ when the going gets tough.
So if I’m making the above trade with my entire pocket-contents of $100 (I still have it — I broke even at horse track) then I stand to lose $10 because my stop loss is 10% down. So actually, to adhere to my own risk rules I can only put $50 of my capital into this trade, thus only risking a loss of $5.
And because I suck at maths (or ‘math’ in some weird countries), I will try and make trades where I can put my stop loss at 5% just so I can use all my capital.
Although 5% seems low to me, I’m forcing myself to admit that on any given day I will be less skilled than I will be the next day. I need to do Future David a solid and not spend all his money while I’m a bit inexperienced.
In other words, a dollar in the hand of Future David is worth more than it is in my stupid fat hand, so it’s a good idea to leave him some cash so he can use his mad trading skillz to make some wicked fast cash.
If I’ve learnt one thing (and I have) it’s that no single pattern universally works on any chart.
I was really pleased last week with the predictable behaviour of the price breaking through a resistance line. But, this was all done on the one day chart (where every candlestick represents one day), so the question is: will it behave just as well with other time frames?
This week, I’ve been looking for that exact same pattern on the 5 minute chart.
Just like last week, I’ll do the before/after screen shots for each instance.
I wonder about all these charts people use to show some pattern that indicates a buy point, but has that buy point in the middle of the chart. It makes sense to show what happens next of course, but I worry that if all the examples of a particular pattern I ever see show the buy point in the middle, it fries the wrong image into my brain.
I want the image that sets off a ‘buy’ alarm to be the same as what I would actually see when the time comes — that is, a chart where I can’t see what happens after the buy point.
So, here’s the first one…
The price only tested the line once, but there’s five candlesticks with their tops right on the line, and the last bar steps over the white line quite decidedly.
What about this one? The price hit a peak an hour previously, then twice more 15 minutes back, but one of those was a bit over.
Ah, this one was more what I was hoping for:
From the opening of the next bar, it jumped up 1.8%. A big jump relative to the typical movement, but obviously trading fees are going to matter a lot if I’m biting off these little chunks.
This is a pretty solid one with lots of touching up against that white line:
So clearly I’ll be betting the farm on this one.
Oops. I never really liked that farm anyway.
This next one touched the line 8 hours ago, then 2 hours ago, then 1 hour ago, then blew past it with gusto.
Bitcoin says no…
So already it’s becoming clear that this 5 minute chart is not as reliable as the 1 day chart.
I no longer have confidence in these signals, but here’s another one anyway. Six bars hit the line, then a decent sized bar breaking through:
One last one before I give up on this pattern on the 5 minute time frame for good.
This is pretty solid. 6 candles coming very close to the line, a breakthrough, and the breakout candle closes right at the top — there is no upper ‘wick’. (A hint I just learned that makes sense to me.)
Another lost bet.
This is a super-valuable lesson for me. If someone says to me “hey here’s an indicator and when it shows this it means you should buy”, I will know with great certainty that I’d be a fool to accept this wisdom without first going through the particular chart I wanted to trade on and finding examples of success.
It’s actually kinda rare in life that you can take some piece of information and very easily go and verify it before trying it out.
Here’s an interesting chart, says me:
I’ve taken three years of daily bitcoin price data and calculated the price movement for each day. E.g. if the lowest price for a day was $100 and the highest price was $103.20, that goes in the 3% — 3.5% bucket.
For those unfamiliar with histograms, the y axis is the number of days that fell into that bracket.
Blame Excel for the awful axis labels, all you need to know is that each bar represents a 0.5% range. So you can see that fluctuations up to 5% in a day are pretty common, and even a movement of 10% is not exactly weird.
This means, that with the right indicator it should be possible to find instances where I can grab 10% in a day or two.
Now, the same chart, but for this I’ve calculated the price variation for every hour.
You can see for most hours (20,000 of the 35,000 hours that have bombarded us all for the last three years) the price fluctuates by less than 1 percent. It’s not unheard of to get a juicy move of 3% or more, but those hours are rare.
The reason I did these charts is that I wanted to get a feel for what sort of time frame is needed to have a reasonable chance of finding a movement of a certain percentage. I mean, because of fees, there’s not much point in me trying to predict movements of half a percent.
And obviously, fewer trades means less work; lots of 1% trades don’t bode well for my dream of working only a few hours a day.
Which brings me to the five minute chart.
You can see that the price rarely moves more than 1% in any 5 minute period. Not surprising in principle but interesting to quantify (interesting for me, at least).
So what all this means is, if I’m going to buy bitcoin then sell it again, and want to make at least, say, 2% on a trade, then I’m probably going to want to hold that trade for at least several hours.
So if I’m looking at the five minute chart, I need to be confident that I can predict what’s going to be happening over the next few dozen bars, which is way beyond my abilities (my ego wants to doubt that it’s even possible).
And because I’m spending 22 hours a day not sitting in front of a chart, I’m missing out on eleven twelfths of opportunities (although I’m still to look into the world of alerts).
But on the daily chart it’s enough to look once a day and make a decision.
It’s taken me a while to realise this, and now it feels like I should have realised it a lot sooner, but I’m pretty sure I need to look at no shorter period than the daily charts.
And if I want to get a return of, say, 10% a month (otherwise I might as well just buy and hold bitcoin) then I’m going to need some more assets (altcoins, forex currency pairs, stocks, etc.).
Is ‘assets’ the right word? Or is it ‘instruments’? It’s a mystery.
So one of the next things I need to work out is this:
- Which patterns work,
- and for which assets do they work.
- How often do the patterns appear,
- how often are they right,
- and what return can normally be taken from such a pattern.
This will probably call for a spreadsheet (hooray!).
I already know one of these:
- The very basic resistance-breakthrough pattern,
- for bitcoin,
- happens a few times a year,
- works 80% of the time,
- and is good for 10%.
So one focus should be: how does this pattern behave on different assets/pairs, and another focus must be on learning more patterns.
Good, I’m pleased with this direction (that I’m working out as I type). Now I’m off to read a book about chart patterns.
Oh actually, I just thought of another thing. The risk/reward ratio. If a pattern is good for a 10% gain, and by its nature I can put a stop loss at 3% — then that’s a good ratio. If I expect volatility on my way to profit, and want a stop loss way out at 15%, but can only expect 10%, that’s not a great risk/reward ratio, and it means I can only put one third of my capital down.
Back to resistance lines — now on the 4 hour chart
I know I’ve just said that I want to focus only on daily charts. But what if this one pattern happens really often and very predictably on the 4-hour time frame?
Lemme just check real quick. I’ll find five instances and see how they look. Yes, I’m going to put all of the charts in here. I feel apologetic about this for some reason; feelings are so stupid!
4 February 2016
That looks promising. The same line was support for a bit, then resistance for a bit, then a whopping breakthrough.
It was a disappointment.
I forgot to record the date for this one
I’m clutching at straws here, these tops aren’t aligned very well. Let’s see…
12 April 2016
I’ve been looking back at all the charts I did last week (on the daily time frame) and it seems like when the price took a good run at the line, then backed off decidedly, then took another run and backed off again, they were the instances when the breakthrough really had some legs.
That would lead me to believe that this has a good chance of…
The four-hour chart is not blowing my socks off.
11 June 2016
I was tap tap tapping the right arrow, inching through the candles, and they were getting shorter and shorter and forming a pretty solid top and I thought, I bet I’m about to see either a bit ol’ green bar or a big ol’ red bar.
And I did! And then…
A healthy climb of 15%. I’m like Yoda for stock prices.
Did Yoda see the future? I’ve only ever seen the Bad Lip Reading version of that show.
But here’s the problem, that happened at 6am on a Saturday morning, Sydney time. That’s my special time for lying awake thinking of all the great comebacks I could have had to logic-sparse workplace discussions from the prior week.
One more, because I promised five…
3 September 2016
I’d be a bit concerned with this one. The resistance line wasn’t super well defined, with many of those candles going just above or below the line. And also that’s a really big spike, and there’s a really tall line at the top (cool kids call that a ‘wick’. One day I hope to be one of them).
The big wick, I am reliably informed, signifies that the movement ran out of steam before the period ended, coming down 1.5% before closing.
The long bar also means I would want my stop loss (on the white line) quite a long way down, thus increasing my risk exposure. Or changing my rule about where the stop loss should go.
Well, it didn’t fall back, but it did take three whole days to even go up by another 2%. Recall that on the daily chart, when this pattern occurred it typically went up by 5-10% within that same time period, and much more reliably.
OK, so there’s five examples of these on the four hourly chart. They were spread over 8 months. One was great with 15%, one was lame with 2%, and the other three would have been a loss.
Colour me unimpressed.
So I won’t be primarily watching this time frame. (I think professional traders will look for general entry points on the daily chart and then look at the four-hour and less to fine tune that entry point. So there’s that to think about.)
I have been resisting going down the not-bitcoin route, simply becuase there were other things I wanted to learn first. But now I want to both expand the number of patterns I can recognise, and try them out on different asset pairs and time periods.
Here’s some Ethereum/Bitcoin charts on the one-day scale, looking at the resistance pattern.
14 February 2017
A very borderline example. The price hit the white line a few days before, and also over to the left of the chart, but went above that in the meantime.
I’m really only capturing it because I’m already up to February 2017 and there hasn’t been anything else remotely resembling a ceiling.
And, it was nothing.
27 April 2017
Here’s another kinda weak example.
Not a big break. Oh, actually, charts are so deceiving, that one bar right at the end is a 15% increase in the price. These are colossal variations. But anyway, not a big break relative to the size of the average daily price range.
That’s more aligned with what I’d been seeing with bitcoin. It’s not quite as pronounced, but the numbers are bigger: that jumped up a further 20% in three days.
24 November 2017
I’d call this borderline too…
There were four prior days hitting that line, in two pronounced upswings, but breakout is half/half straddling the line.
And down she goes.
28 January 2018
This was not-really a top from months ago. So let’s see…
So that was underwhelming. I disregard this pattern when it comes to trading ether vs bitcoin on the daily chart.
What should I do next, the same thing for GBP/USD?
OK, I’ll spare you the charts. I just went through three years and could only see four clear instances of resistance. In three of those cases I failed to see the price behaving like it ‘broke through’ that barrier and proceeded to move up at a greater pace. In one instance, the price at least jumped up more than average the day after, but that was for a grand total of 0.85%. Before fees.
In fact it’s rare for any day to move more than one percent. So I can also forget this approach for the pound vs dollar daily chart.
Clearly, this old one-trick pony-dog needs to learn some new tricks.
Hey, I saw that glance at the scroll bar. Yes this post is not done yet.
The below is for the now-familiar bitcoin/USD pair, daily chart — but this time I’m looking for support patterns.
The figures are assuming I’m ‘shorting’ (making money from the price going down via some sort of witchcraft).
I’ve had a thought: charts should go sideways. I think my brain has expectations of gravity built in, so naturally down feels like a path of less resistance than going up.
Like riding a bike south.
But if time was on the y axis, and fluctuations in price went left and right, maybe that bias would be removed and my predictions could be more accurate.
Moving right along. Starting three and a bit years ago…
13 August 2014
It’s that familiar pattern again, the price bumps up against the white line, then steps right over it with some high volume (even though I’ve hardly been mentioning volume).
Ah that’s a good one. The price dropped almost 20%. But I’ll imagine I had an automatic sell after 10% and say…
Result: I would have made 10%
17 September 2014
Just a month later…
This one looks pretty good. Although it doesn’t step over the line by a huge amount, and there’s a big bottom wick, and the volume is equally not huge. And then…
Niiiice. 17% in two days.
Result: 10% win
16 December 2014
3 months later. These happen more frequently than I have good days.
Fun fact, even though I’ve done, like, 400 of these screenshots, I continually forget to put the stupid arrow in. So I screenshot, paste, swear, add arrow, screenshot, paste.
I wanted you to know this.
This time I would have had to wait a few weeks for it to move by 10%, but it would have eventually.
Result: +10% (but it would have tied up all my capital for a few weeks)
13 April 2015
More of the same, no words required, right?
Sod it. It made a move of a few percent but then gave up.
Result: would have lost 4%
8 August 2015
This only really fell back to the support line once, but hung out there for a few days. But it’s a nice hearty break through and some solid volume. I have high hopes…
Awesome, it came within 0.1 of the stop loss (white line), but then continued in the desired direction. It headed on for almost 40% but I would have been long gone by then.
Result: up by 10% (and struggling not to say if only I’d hung on longer…)
31 July 2016
I’m not super keen on this one because that support doesn’t seem very well established. But, unestablished as it may be, that big red bar smashed through the white line with conviction.
If I’d had an automatic sell at 10% it would have come infuriatingly close (9.69%), then hung around between my automatic sell on either side before, 70 days later, triggering my stop loss.
I probably want to think about how long I should allow my capital to be tied up before exiting a trade. If I look at all the charts I’ve made so far, the 10% mark, if reached, always happened in 9 days or fewer. So I’m adding a sell rule: after 10 days, sell no matter what the price.
Wait this is interesting, there’s ten days missing in this chart data. What happened between the 2nd of August and the 10th of August in 2016?
Ah, Bitfinex got hacked, that’s what. I’d forgotten this is the wild-frikken-west.
OK, since the purpose of this exercise is to assess the performance of a particular pattern (which I will ultimately be comparing to other patterns), I’m not going to skew the results by taking into account a hacked exchange.
I just looked at the chart for the Coinbase exchange. But of course the price plummeted on the news that a bitcoin exchange had $72 million stolen. So I’m going to write this example off as dirty data (although for the record it of course dropped 10% on Coinbase).
13 September 2017
We all know bitcoin was going nuts in September ’17, so this one would be betting against the trend, but I’m still going to point it out.
Even if I didn’t know the uptrend still had a few months left, I would be dubious about this one because the red bar doesn’t go far past the support line, meaning a really close stop loss.
It did make a hefty 16% lunge, but it also tripped just across what would have been my stop loss line. Looking at the four-hour chart, it seems that the price went up before it went down again, so the stop-loss would have been triggered.
This stop-loss triggering has happened once out of over a dozen times, so I’m not going to adjust my rules just yet.
Result: I would have lost 2.8%
16 January 2018
I’m getting a bit fast and loose here, those bottoms don’t align very well with the white line.
Oh this is interesting. See that thin little green line dipping down the next day? That’s 16% below the open price. So if I’d had a stop at 10%, I would have got out just fine.
Result: I would have made 10%
Discounting the occurrence of the pattern that was interrupted by the Bitfinex hack, the pattern occurred 7 times. 5 times it would have returned 10%, and the other two lost 4% and 2.8%.
So, this ‘support line’ pattern has an average return of about 6%, and happens a couple of times a year.
Not far from the ‘resistance line’ pattern which also happens a few times a year and averages a return of about 7%.
So it seems there has been 25% a year to be had from these patterns.
Compared to the bigger, but riskier, but lower-taxed return of 300% from just buying and holding (rough average since 2011).
As I get better, this return of 25% should improve along three fronts:
- I refine my understanding of these patterns
- I learn more patterns
- I apply the patterns to more assets (instruments?)
Idea: trade reviews
In the world of writing code, it is well accepted that it doesn’t matter how smart you are, your code should be reviewed by a second person before being incorporated into the main application.
It’s called a ‘code review’, and it’s a great way to avoid making buggy software.
I see no reason why it shouldn’t be the same with trading. If I can’t define a trade and explain to a trade-buddy why I’m making the trade, then I shouldn’t be making that trade.
It would be cool to have trading teams of four or five people, where you would have the chance to get familiar to the approaches of your team mates and vice versa. They could then question why you’re deviating from your previously-successful pattern, or potentially repeating a mistake you made earlier. It protects against emotion-driven decisions, because the other person has no skin in your game.
Now I sit and wait until someone tells me there’s an app for that.
Bonus feature, ‘psychological barriers’
You’re about two feet from the bottom of the article.
There’s a theory that a an asset’s price will behave differently as it approaches big round numbers. For example bitcoin might have ‘trouble’ breaking through the $1,000 ‘barrier’.
I was initially going to write here that I think that theory is perhaps a bit of armchair psychology and is lacking the evidence to back it up. Certainly, on the daily bitcoin chart, $1,000 and $10,000 do not stand out as I would have expected.
But as I went through with markers at every multiple of one thousand, I saw that both $3,000 and $5,000 form pretty solid support and resistance lines. Can you spot which two horizontal lines are three and five thousand?
And it seems that 8k and 13k were also support/resistance levels. Surely there’s not that many Fibonacci nerds out there. I guess we’ll see when we get to $21,000.
Regardless, the predictive power of a resistance line based on a Big Round Number is not so useful. For instance I wouldn’t be able to pick the $4,000 line in the above screenshot, and wouldn’t know what bet to place as the price approached that line.
Maybe it’s worth investigating if a support or resistance line is stronger if it is also a Big Round Number.
Did I tell you (dear diary) that I finished Cryptoassets? Great book, read it.
Now I’m half way through reading the novel Reminiscences of a Stock Operator. It’s great too. I couldn’t find the person who recommended it to me. If it was you, thanks!
It’s based on a true story of a guy that trades, way back around 1900 (and I thought 6am Saturday was early!). It’s wonderfully written and an interesting story; the fact that it has some wisdom about stock trading is a bonus.
I still have those nagging thoughts: where are all the chartist billionaires? Where are all the hedge funds consistently returning 100% pa? Why would any decent trader bother charging $29/month for a newsletter?
(To answer my own questions, the trader billionaires are on their yachts, the hedge funds are, like, 60% boring gold so have necessarily low returns, and I guess for the newsletter if you get 10,000 subscribers, that’s three-and-a-half mil a year — so good on ya for selling out.)
On the optimism front, I like hearing all the things the experts have to say that 90% of people get wrong. I see it as not-impossible that, in time, I could be one of the 10% who do make money. Particularly if taking it slow and thinking about boring things like risk management gives me an edge.
On the topic of ‘is technical analysis a real thing’, I think it’s a lot like meditation. Trying to find out if there’s any real science behind it is tricky, because there’s no end of people spurting pseudo-scientific nonsense.
If you’re an evidence-requiring type, but all you see is the hand waving, it’s easy to write the whole thing off.
My current thinking is the same for both technical analysis and meditation. There are plenty of purveyors of preposterousness, but at the core, there’s something that looks quite promising.
More patterns. Probably trend lines. TREND LINES!
Also, something I’m thinking about is whether I want to abandon the idea of trading between a fiat currency and bitcoin. Firstly, I’m in Australia, so I’m not sure if that means an extra step is required to get to USD, or if I trade AUD/BTC I’ve probably got a lot less volume and I wonder if that means less predictability.
But also it seems that transfer fees are more for fiat/crypto than they are for crypto/crypto.
The question is, are the patterns weaker, and therefore profits after fees are lower than with fiat/crypto.
And lastly, as I go from bitcoin back to a fiat currency, capital gains tax applies. But if I use bitcoin as my base, and then trade between different cryptoassets, does the tax man care? (Dammit, yes crypto-to-crypto exchange is a ‘capital gains tax event’ — in Australia.)
OK, thanks for reading this giant wall of text. See ya next week!