[00:00] Hello everyone and welcome to the Quampedia explains trading strategies video series. Today we will discuss overnight effect in gold mining stocks. Hello everybody, my name is Razan Botkan, CEO and head of research at Quampedia.
[00:15] Today I would like to discuss with you the overnight effect. And then we will take a look on the dangers of using open high roll close data, which is I mean very often available for free on a lot of the websites.
[00:28] We will especially discuss how the opening prices in these datasets are un-reliable, so it's not very good to use them in actual backlisting. Firstly, Zolio is built on the article that is called
[00:40] The dangers of relying on open-hylo-closed prices, the case of overnight drifts in GDX ETF. The article is freely available on our blog. Firstly, what's the overnight drift, or what is the overnight effect? The overnight effect is a tendency for a local class
[00:54] to increase in the price during the overnight session. so during the session when the market is really closed. That effect was found out in S&P 500 indexes and in the stocks. Here we can see a nice chart, so we can see the majority of the returns, or all of the
[01:08] returns of the S&P 500, or SPI ETFs, come up during the night. So, I mean, if you hold S&P 500 over the night, you get all the performance for S&P 500. If you hold it only during the day, you will get zero performance over the last 40 years.
[01:20] Why is it so? The reason is because when you hold some asset during the night, the market is closed, you You cannot exit with your trades, you cannot sell the trades, you cannot sell your assets and you are in the risk and the overnight effect is compensation for this risk.
[01:34] Now this overnight effect is very common, so it's not just in the S&P 500 or in the indexes or in the stocks. I mean it's very common also in the cryptocurrencies. Yeah, the cryptocurrencies are open 24-7, but I mean there are times of the day when the
[01:49] trading in the cryptocurrency is low, so there are less liquid and that's the time when you can hold the cryptocurrencies and we are compensated because the cryptocurrency is more illiquid in this time because major markets are closed. I don't mean the cryptocurrency markets but
[02:02] people are sleeping and I mean the major equity markets etc. are closed so the professionals are not in the work and they are not trading also the cryptocurrencies. So there are only the bots and only the algorithms. So that's the overnight effect. If you have this overnight effect you can
[02:16] find it in other assets. So that's what we tried and one of the most popular assets is the gold. So we downloaded data for GLD ETF, which is the ETF that holds the gold as an underlying,
[02:29] and we tried to find out the overnight effect in the gold. So what we can see is that based on the prices on open high low close prices in the gold there is overnight effect When we hold the GLD ETF during the day so between the open and the close we have the zero performance If we hold the GLD during the night so between the close and the open when the GLD trading is closed I cannot sell
[02:51] the ETF, all the performance is realized over that overnight session. Okay, so we have the same effect in the US equities, we have the same effect in US indexes, in future currencies, in high hold and also in the gold. So let's try to combine these effects, because there
[03:04] is a special type of stocks and I'm talking about the stocks that buy or extract gold from the earth's crust. I'm speaking about the gold miners. We can buy the gold mining stocks and we can hope that
[03:17] we will have the convergent effect. So this convergent asset, gold miners should give us the potential for combined overnight effect drift because there should be drift in the gold prices and also the drift in the equity prices of the gold miners. We downloaded data about the gold miners ETF
[03:32] the open high-roll and close, and we check what is the overnight effect in this asset or in this ETF. And our analysis revealed a significant overnight risk of approximately 40% per annum. And there was a substantial negative intraday risk of about 9.25% per annum.
[03:47] So what the data suggested is that if you took the GDEX ETF and you bought the ETF at the close and hold it down to the open, you could have 40% performance per annum. And if you bought the ETF at the open and hold it down to the close,
[04:01] you could have minus 25% performance per annum. Which is, I mean, those are really extreme numbers. Just looking on the numbers and just looking at the charts, we could get quick here. But I mean, for our experience, this looks too good to be true.
[04:14] I mean, this is like a very, very strong risk and very high outperformance of the 90 sessions and strong underperformance of the daily sessions. So we have to investigate the underlying problem. So what is our experience is that the problem is usually hidden
[04:27] in the opening crisis of the open, high, low, closed data set. Notably, the problem is with the opening price, because the opening price in the dataset is usually not the market on open auction results, so it's not the price you will get when you
[04:39] enter the market on open order, but the price in open high local datasets is usually the first trade that's realized after the market is open. And it can be the trade with very low volume, I mean with totally off the price you will
[04:55] get from the auction. It's not possible to get the opening prices. So the prices that are in the open, hire or close data sets are not very low, especially when we are talking about opening prices. So what we can do is we can get the data with the higher granularity.
[05:10] So minutes by minutes, seconds by seconds, or I don't know, tick by tick data. And we can check the execution assumption So we can check what is the first price we can get relatively after the market opens So that what we tried So we switched to the Quant Environment The Quant Environment gives us no just open high or closed prices but we can
[05:29] also receive the time of change data, peak by peak data, or mini by mini, or second by second data. And we run four scenarios together. So, scenario number one for SPI was buy SPI on market and close, and sell it on market
[05:42] open. And scenario number two was buy SPI market on close and sell it on 9.31 for one minute after the open. Why we are using the market on close? So the close is usually reliable. So the close prices in open high or close data, they are fairly available, I would say, is okay to find.
[05:58] In a plaza mining you will be able to get these prices when you enter the market on close action. That's usually not the problem, the problem is with the open. Then we tested the GDX trading strategy, so buying and selling the gold mining for stock CTS.
[06:10] So in scenario number one we bought on market on hold and sold market on open and scenario number two we bought market on hold and sold 9.41, so one minute after the open. Let's get to SPI.
[06:22] This is scenario number one in SPI and here is scenario number two in SPI. So we can see that the total return of the scenario number one and scenario number two is plus or minus the same here. I mean when you try to trade the overnight effect in SPI and instead of selling on open
[06:38] price which is unavailable you will try to sell it at 9.31 price your return is little lower so the total return is nearly 400k when you are trying to sell it on open prices you will get
[06:50] over 400k and in total returns difference is really only the small one so the overnight effect is well analyzed in SPI the other is a slight decrease in performance if we ever should sell orders at 9.31 compared to the hypothetical execution at 9.13 which we said before it's
[07:07] unreliable, but the decrease is small. So, the effect is to present. It's still true that the significant part of SPI that's returned over the last year is registered over the night session. And it doesn't matter if that night session ends at 9.30 or 9.31, it is under the price that you
[07:23] could actually get. Now, let's get to GDX, which is the gold miners' ETF. So, scenario number one, we basically didn't want to connect, we didn't open the price, it is once again, I mean, we got our portfolio from 100k to 20 million in i don't know 15 years so i mean we could get rich
[07:38] based on these numbers but i mean one again these are not correct numbers opening prices is the first trade of the day and you are not able to exit with this price but in the second scenario
[07:50] we sold overnight position at 9.41 here we can see that there is still the drift so the prices are going up i mean it not 40 percent running the overnight drift on the other hand the second scenario yields significantly more realistic outcome So we have a strategy that has around 8 per annum performance 40 maximum lowdown
[08:11] 16% volatility ETFs. The risk overnight drift in GDX prices, so in gold mining stocks ETF, is definitely there. However, the magnitude of the effect is not as high as using the open high low close
[08:24] data. So, I mean, as the first analyze is hinted. So, what's the result and what should we take out of this analysis? So, firstly, yeah, there is overnight risk in gold mining stocks. There is definitely. This is strong and this is light.
[08:36] So, it aligns with our observation in addresses, so in equities, in cryptocurrencies, in high orders etc. What is important to understand is that when you are trying to run an analysis, it's important to understand how your execution looks like.
[08:50] I mean, it's really, really a problem when you rely on open prices from open high or close data. When you want to test the strategy that sells on the open, it's really better to move to the data with a better granularity.
[09:03] Test it on minute by minute or second by second data and check out your execution. So, I mean, what are the prices at which you will be able to execute? I mean, really execute. So, pay attention when you develop the strategies that expect execution at open prices.
[09:19] It is impossible to get the opening price from open, have a close data. So in this case we moved to 9.31, we got more reliable, better results and we see that there is a drift, but it is not so strong as expected.
[09:33] I mean the strategies can be still used, it will have probably the high trading fees and high split pitch, but I mean it can be used as an overlay or it can be part of another strategy or it can be used as a filter etc. I mean in the future I will show you how to use this finding about overnight drift to
[09:48] involved mining stocks how to build a strategy that doesn't trade every night possibly there is a lower yield etc etc but we will get into that in the future so what is the most important at the moment what is important to take from this video is do not rely on the opening
[10:00] slices from open high or cold theta so thank you very much i hope that you liked this video all of the articles that i discussed today are freely available on our blog so feel free to take a look on them it can be overnight effect in high yield overnight effects in bitcoin overnight effect in
[10:14] SPI or NFA team gold and gold mining. Please give us a like, subscribe to our channel and if you are interested in something feel free to contact us via our social media channel. Thank you very much and I hope that you will join me in the next video.
[10:29] Are you interested? Then pick another video to learn more or subscribe to QuantpediaPro and try how our analytics and reporting significantly save time spent on quantitative research.