ETH Market Depth and Structure

Transcript

okay good morning so before we start just today I'm we're going to try to go and dive a bit deeper into aetherium and what goes on in on marketplaces where aetherium is being traded so just a little background about where the data will come from so all the data that is being used to do to derive the insight that will be briefings that presented in this presentation come from the public API is of the exchanges and what we've been doing at keiko since 2014 is building infrastructure to be able to reliably and robustly collect normalize store and then redistribute

all this market data and so the questions we'll be answering in this talk is consists of mainly where do people trade when at what price if you want to buy a lot of eath how much would you move the markets is there any trend any pattern in terms of one people trade do people wake up early in the morning to start trading early or they're mostly BOTS trading all day long will Dell delve into the this sort of questions so before I start showing you more plots and and charts on exposing this data I'd like to give you a brief introduction about how what this data is so we'll be

looking at basically past year from January 2019 until about February 2020 and will be mostly looking at level 1 and level 2 data from topic changes and top fiat currencies so because we're in Paris also I did a bit of Euro if your own Bureau and it will be mostly USD and tether so ohlc V data is basically when you look at all the transactions that take place on an exchange it's usually convenient to group them by time periods or time windows so you can do this daily and you will like look at what what the

opening price of the day was the volume and the closing time and at the closing time - and then all the the extremes values and the average price over the day audio book snapshots are a bit more granular and it's basically we'll just have a look at what a an order book is but these are called level 2 data and so what is an order book so let's say meet Bob and others so Alice has got some eath and she would like to sell this ethan she just met Bob and Bob has got some US dollars or Euros and would like to buy something so Alice was full given

asked probably the price at which he is willing to sell and Bob will respond with the bed the price at which he is willing to buy if these prices are compatible they will agree on the price and the transaction will take place otherwise what's going on on exchanges is usually you have a lot of those limit orders which is there's a lot of Bob's there's a lot of Alice's and each of them have a different price a different amount they're willing to trade at different levels so when you stack all those limit orders by price levels you will see a graph which which looks like like this and so as we said if now somebody places a bit which is closer to

the red zone when it hits the red zone a transaction will will give birth to a transaction which we will be storing so the difference between the two best offers is called the spread and the best on the left side has the best bid and on the right side the best ask so that's the core of the data we'll be having a look at and this tells you about all the latent orders that are being placed and where people have committed volumes that to the market where transactions will occur if some if they are met and all those prices and amounts of course change quite rapidly over time

so if you plot them for Bitcoin USD over time of the last month of I think the last week of August you can so this is aggregated by various levels and you will see that there is large changes over time so what we'll be looking at is basically snapshot of this of this data that we'll be integrating to see for example now if I want to buy a hundred eath and there is not enough Eve close to the to the mid price I will have to go deeper and pay a higher cost to get all the eath that I wanted and so by how much I will be moving this mid price so these are more details about how we do the calculations and how we define these

metrics so as so slippage is an important metric when you're just buying one or two eath you might not be very wary of the moving the price but when you do a lot of trading obviously and it change into a larger volumes you want that you need to nail your execution costs and you need to - this is a core issue if you want to have a successful trading strategy so this is what we'll be looking at looking at throughout this talk and this is obviously tightly linked to what's called liquidity so what is liquidity on aetherium spot

markets the first metric we'll look at is the market depth so market depth is it gives you an idea of the volume committed at a given time on an exchange and more precisely here we'll be looking at the 2% market depth so the 2% market depth is the total amount of eath or that you can express also in the end of quote I said namely daughter or tether or euro plus or minus 2% of the mid price so basically if I have 1 million dollars on the on the bit side plus or minus 2% around the mid price that means

at any given time I know I can buy instantaneously this amount and so on the left-hand side you see this 4-bit Phoenix so you can see that on BitFenix there is lower liquidity for tether than there is for a theorem in Euro for USD and euro and you can see that actually euro the euro pair and the USD pair are actually quite close on BitFenix and similarly on coinbase you can see that well actually there is a big difference on on coinbase of where most there's a lot more with courtesy on on US dollar and now if we saw that just to show you how these work from non specific

exchanges but now if you want to have a look by pair this is what what you'll get so on the left hand side you can see that there has been very large increases in in market depth on BitFenix recently and the couple in the past few weeks and overall you see that it's quite volatile this is obviously a daily average so if you look at more granule data you'll see even more volatility and you can see that in general market depth has been increasing so there is more and more volume present on this order book around the mid-price okay and so next so as we said if you

place 10k order or a 50k order you will have to go further away from the mid price how does this change how much more slippage we have to pay in order to execute this order so this is what those two plots show so with the dashed line represents what you'd have to pay for a 10k order and the continuous line is for a 50k order so five times the amount and what you can see here is basically there is some some increase and on BitFenix same as previously so it's quite similar for euro and and USD it's much higher slippage if you

to tether so you might want to have a look at other exchanges for for example for tether but overall it's quite impressive how much the slippage has been decreasing on the on the tether pair on both necks and also on the euro pair for on coinbase over the past year so this means that it's there's more and more liquidity on those exchanges for those instruments and finally for the stable coin so on tether you can see that on Bobby there was a lot of liquidity already at the beginning of the year whereas on balance it's been improving at a very important rate and so finally if we look at again prepare on the on

different exchanges what you'll notice is Kraken recently added a you tether etherium tether pair and this is the adoption of this pair on the exchange where more and more people are willing to trade on the on despair and so you see the slippage going down as people have put more and more volume and the depth increases and similarly you can see that there's a general trend especially on tether with slippage improving over time so making markets more efficient so next I'd like to get into the arbitrage question first observation is long gone or the days

when you could look at the exchange prices and manually go to other exchanges and try an arbitrage out the price already a couple of years ago was difficult and sometimes it was difficult to get out of some trades but now it's just white wine oh it's not feasible anymore so for this we try to get a case which is a bit demanding so not just st. pairs but different pairs and looking at stable coins versus fiat so we look at US dollar in a three-room US dollar pair on BitFenix and on a different exchange what's going on with aetherium tether and what you'll see is this also you know shows

one more traders were there trading processes okay I move everything to crypto you and see one tether as one dollar and what you can see on this plot is right in the middle that is the correlation diagram between the the minute returns on the the minute price differences on a theorem dollar and it's a measure on the on those two exchanges and so what you can see is that as you as you might expect when the price moves on a bit Phoenix on a theorem dollar it also moves in a very similar fashion on the on the other pair and now what what happens if you shift the those price

changes by one minute so on the right you look one minute later on the left you look one minute before and what you see is the correlation vanishes and that shows you that price changes are incorporated into the very efficiently in the markets and in less than a minute it's already been reflected and just one little remark on why we look at price differences and not returns that's due to the micro structure of the markets where when you go to very small timescales returns are not as relevant as absolute price differences finally let's try and see okay well let's see in

this room when do you if you buy or sell when do you buy or sell do you do this in the morning before going to work in the at night when you can't sleep when when Donald Trump tweets on Twitter over the weekend so that's what we'll have we'll try to have a look at so over the past year on average how much was traded on each day of the week each hour of the day so let's start with the day of the week and just simple volume so how much in total was exchanged on the day the first thing you can notice is the volumes on stable cones are impressed and they were much higher than what you

get on on Fiat so look so those blots will show the total volume trait converted into the the quote pair just because it's easier to grasp intuitively and on across all the exchanges named on the title of the plot and so in general the volumes are much higher on USD T then they're only on euro or USD obviously forms are you're lower on the euro pair and most noticeably there is usually a little dip over the weekend so there is less volume being traded on weekend and that's even more not scible for fiat pairs than it is for tether so

next let's just go so from on the weekday now we go to the hour of the day so every day was what what's going so there isn't really a clear pattern that emerges you could change so we look at the whole year you could look at different months and then try and see if there is anything statistically significant that emerges out of this one observation is if you look at the ether euro there is a dip over not nighttime which might indicate that there's a little less trading activity over overnight there are also two peaks that are quite correlated on both plot so if you look at it from Euro or USD there is a peak around 4:00 p.m.

UTC and another one at midnight for the peak at midnight you could think some people have some algorithms that are rebalancing at a particular time but again nothing nothing too sure about about this one would have to look a bit deeper so that's about the volume now still thinking about liquidity let's have a look at the slippage which is somewhat linked to it and again one the on the most liquid pairs on the most on the biggest exchanges there isn't much time dependence so whatever the the hour of the day you'll get on average about the same slippage apart here on coinbase you can see there is a little more slippage during night time in Indy

you which might indicate again that there's a bit less the market liquidity over over those timeframes although you would expect that to be mostly done by market makers with with completely automated BOTS so not necessarily too sure of where that comes from and finally the over the week so what's interesting is we said there was lower volume on the weekend and that translates in lower liquidity and sometimes in higher slippage that's not the case on big exchanges because you have a lot of automated bots which are and market makers which are here to

provide this liquidity in compensate and let's say and also for Gemini on the etherium USD pair that goes with an increase an increase in the in the spread over the over the weekend and before we wrap it up and open the floor for questions well I'd like to thank all the organizers of the HCC event and and and let you know that we are going to be all this data will be available online for the presentation and we're and hopefully we'll have soon some fact sheets that we'll be delving deeper and

this sort of data to try and give an idea of what is going on on the markets and how traders are behaving to add up thank you very much [Applause] someone want to ask you a question refer bid time thank you for this very interesting presentation just because it's fun to speculate why what do you think would be the reason there was a surge of trading at 4:00 p.m. so the the first thing I would look at is how did

that that there might have been like some big moves occurring around that time at particular times so the first thing I'd look at is okay was this the same during q1 q2 q4 is this consistent across that the other thing is it's always unclear about the interactions you could get between Asian markets us European markets so trying to have a look at the what time it is in those countries but the other thing which is a bit tricky in the crypto markets is you don't have the same structure as you have in traditional finance where you know traders on the East Coast on the west coast they start at 5 a.m. to match

the East Coast time so the hours are much more framed on on Krypton markets I mean if you're a trader and you want to wake up at 10 a.m. you can just start trading at 10 a.m. so there is not this big difference and obviously it's 24 hour markets so you don't have the statistical anomaly that you get in traditional markets where at the opening and that the closing you have especially at the clothes you have more volume going on and potentially the 4:00 p.m.

is related to some sort of soft closing time on on more European markets thank you for your presentation it's really interesting I gave a question about different type of analysis you know like on regular financial markets there was this market crash of 2008 where people were producing all of this complex you know security is and they dependent on each other on the algorithms and what ended up happening is it algorithms no one knows how they run so at some point they created like short squeezes or stocks going up and down on things so it seems like we're also like in creating these complex

algorithms now more and more complex so the question I have is okay do you see on the markets this kind of erratic behavior where you have the singularity so things jumping due to algorithms and can one use data just identify the systemic risks that people may be putting in some kind of catastrophic scenarios where the market just can melt down due to this algorithm like fighting each other or some other things so I think there's a lot of to impact in this question to impact but I feel some of it is around more very high frequency low latency trading and you know all the flash boys thing and how those impact

their markets let me just say this on crypto Marcus the there isn't asthma it's not I mean the high frequencies nowhere close to what it is on traditional markets as to how the micro structure important the micro structure impacts the broader markets that's always tough to say there's a lot of research that's been going on on the you know fractal behave yourself similarities across different scales it's it's a that's a that's a tough one I will engage on on back to the crisis to to a right quote crisis there's also the a lot has been blamed about

mathematical models and how having too much consensus on the way under way to prices instruments could lead to crushes because people were under estimating the risks and at a big component of that is how do you price things but in crypto I mean for derivatives here you can use the traditional relationships which were more due to the nature of the instrument on crypto markets traders have still been struggling to find fundamentals really so when a lot of it is speculative you have a lot of shocks going on in the system and and and it's actually quite correlated a lot of it is

correlated to make call it's a Bitcoin and that that goes into what you said about about the market so if you define them are there systemic risk if you define the market as some average of Bitcoin a lot of a lot of betas are close to one [Applause]