Current location | Thread information | |
![]() ![]() ![]() ![]() ![]() ![]() |
Last Activity 4/21/2021 9:48 AM 15 replies, 2274 viewings |
|
|
Printer friendly version |
^ Top | |||
jpb![]() Veteran ![]() ![]() ![]() Posts: 168 Joined: 5/11/2005 Location: Brown Deer, WI ![]() |
Ever since I was a kid, I've had this feeling that trading at different times of the day (not day trading, just seeing what's happening in the market and placing orders) would give you different results. But I've never taken the time to explore that thought until a few weeks ago. I set up a simple test using ATM, Auto Trade, and 8 GXT Paper accounts. The basic process of this test is to set Auto Trade to run at different times of the day and submit orders immediately after it is done. So, this isn't running an EOD run after hours and then submitting orders at different times of the day, this is running an EOD run during market hours and placing orders that result from those runs. For this test I'm using 2 custom OmniScan lists. Both lists start with "All Optionable U.S. Stocks & ETFs", sorted descending by AVG(C*V, 10), selecting the top 300. The second list is the same but uses a different formula for the Sort value -- same starting population. These lists are then analyzed together in the To Do list using "All Lists in this profile". Port Sim is set to use All Analyzed Symbols as its source list. So, because of this, I would expect to see that there is a slight difference in symbols traded in the later part of the day than what might be traded at the beginning of the day. Times of the runs (all in Central Time): 0800 (This is intended to be a representation of an EOD run, finishing before 0830). 0900 1000 1100 1200 1300 1400 1430 (This is intended to be as close to MOC as possible, finishing within the last 10 minutes of the market). All runs are typically 20 - 25 minutes in duration. Data has the 20 minute delay of the OmniData EOD data feed. Each time period is run within a unique instance of OT in a unique VM (actually a linked clone off one base VM so that everything between the 8 VM's was identical). 7 of the 8 GXT paper accounts started with $10,000 with the exception of the 0800 account which I repurposed. As of this time, you can't reset the value of an account. Since I repurposed it, I needed to make an adjusting calculation in the spreadsheet to normalize it to a $10,000 beginning equity. There is also a quirk with the broker connection to the GXT paper accounts in that OT doesn't remember which account you were connected to like it does with the live GXT account. Nor does it remember that you were connected to the paper account the last time you used the profile. These two items mean that when you start OT each day or if you lose a network connection, the GXT paper account defaults back to the first one. I had one event over the 2 weeks where trades from a different time VM ended up in the 0800 account. I closed the trades and moved them over to the proper account, introducing an adjustment to the 0800 account. After a little more than 2 weeks of running, here's what things look like: From this short time period, I was surprised to see 1000 in the lead while 0900 bringing up the rear. 1400 still has a slight edge over 1430. I seem to be getting slightly better prices 30 minutes before market close than right at close. We'll see if a few more weeks of execution continues to prove this out. The focus here was impact on the ending equity if you traded at different times of the day and watching how an OmniScan list changes over the course of a day. The specific trades weren't interesting, but the symbols selected and ending equity is. The attached spreadsheet has the journal of events. It is interesting to see that 0800 and 0900 tend to act similar to what you might expect out of an EOD run. Not terribly shocking since the 0900 run has only 10 minutes of market activity to adjust its symbol list. However, when you start to get into the 1000 and 1100 runs, you begin to see trades that show up in the next day's 0800 & 0900 run. There certainly are a few phantoms that pop-up in one hour's run and trade that don't make it to the subsequent hour's. Finally, I forced all accounts to go flat on 1/25/19 about 5 minutes before the end of trading. This is simply because I'm headed on a 2 week vacation on Monday and want to be able to pick this up again when I get back. So, a question to the forum: Have you seen any studies performed that offer any insight into time of day trading for the EOD trader other than MOO or MOC? [Edited by jpb on 1/25/2019 4:10 PM] ![]() ![]() | ||
^ Top | |||
LearnerGuru![]() Member ![]() Posts: 29 Joined: 12/3/2009 ![]() |
Try working with Volatility, you'll get some pretty good insights. Two weeks a bit short ... consider a study length of a spectrum 3 months or more, better one year worth of data collected - seasonal of market participants during Winter, Spring (sell in May and walkaway), Summer (volume drops), and Fall months (how market picks up after Labor Day and school starts) .... should see how market makers are busy around what times, and when they go for lunch, sluggish times, you should even see Algo spikes at certain times too, etc. etc... | ||
^ Top | |||
LearnerGuru![]() Member ![]() Posts: 29 Joined: 12/3/2009 ![]() |
I have seen 5-minutes works best, but can test various other time frames of your interest. | ||
^ Top | |||
jpb![]() Veteran ![]() ![]() ![]() Posts: 168 Joined: 5/11/2005 Location: Brown Deer, WI ![]() |
Thanks for the thoughts. I agree that 2 weeks is too short. I'll pick it back up after vacation. Since I have to manually connect each VM to each account each day, I can't just let it run without software changes to the GXT Paper connection in OT. 5-minute time period sounds like real-time trading. Have you seen any studies about time of day? I've seen the lunch periods and seasonal discussions, but I've never seen anything that studies an EOD methodology, using partial/evolving day's data, and placing trades once a day. It is kind of like asking if MOO or MOC or place trades over night or just prior to market open is beneficial to small traders or will they get better results if they delayed their trading (decisions and order entry) to a different window of the day. I think that's the interesting question since I can set Auto Trader to do this for me. | ||
^ Top | |||
LSJ![]() Legend ![]() Posts: 515 Joined: 8/17/2006 Location: Citrus Springs, FL ![]() |
There are times of day that when I was day-trading were important to me. However, for EOD trading I do look for entry and exits with the TOD in mind. I do watch for a mid-morning switch around 10:30. The Morning Reversal- 9:50am to 10:10am Low Risk Trading – 10:10am to 10:25am Decision Time – 10:25am to 10:30am Final Move of the Morning – 10:30am to 11:15am Go Eat your Lunch!! – 11:15am – 2:15pm Back to Business – 2:15pm – 3:00pm It’s GO Time – 3:00pm – 3:10pm Use Caution & Stay with the Trend – 3:10pm – 3:25p Funds and institutions rebalancing their portfolios) – 3:30pm – 4:00pm The above is from https://tradingsim.com/blog/day-trading-time-zones/ | ||
^ Top | |||
jpb![]() Veteran ![]() ![]() ![]() Posts: 168 Joined: 5/11/2005 Location: Brown Deer, WI ![]() |
Awesome! Using time of day that are have some significance to day traders. I'll look at this further and perhaps overlay these times with what I'm seeing (adjusting for timezone and data delay). Thanks LSJ for providing this data. | ||
^ Top | |||
LSJ![]() Legend ![]() Posts: 515 Joined: 8/17/2006 Location: Citrus Springs, FL ![]() |
Thanks for sharing your research. As I think about I recall that Larry Williams did some work on trading time-of-day, day-of-week, etc. He even had seasonal intraday data. LW also came up with the "OOPS" system: "a) If you are watching daily charts, the first condition is that there has to be a SUSTAINED downtrend for a few trading sessions. I mean few Red Candles on daily Charts b)On the last day of downtrend when the Oops buy occurs, there is a gap down, which opens well below the previous day’s low. c)During the course of trading the stock rises and goes above the previous day’s low, and also the previous day’s close. The above 3 steps if occurs generates an OOPS buy, with a stop loss of that day’s low." All of these ideas could be incorporated into an EOD system and the testing could go on ad infinitum. At the moment I am playing with GA and I will try putting the "OOPS" structure in a GA and see if it adds anything. [Edited by LSJ on 1/26/2019 8:52 AM] | ||
^ Top | |||
LearnerGuru![]() Member ![]() Posts: 29 Joined: 12/3/2009 ![]() |
From my notes: RTH Notable Reversal Times 9:30 - 9:35 a.m. 9:50 - 10:10 a.m. 10:25 - 10:35 a.m. 11:15 a.m. 12:00 p.m. 12:45 p.m. Major Reversal Times ---> memorize 9:30 - 9:35 a.m. 9:50 - 10:10 a.m. 1:30 p.m. 2:15 p.m. 3:00 p.m. 4:01 p.m. 1:30 p.m., 2:15 p.m., 3:00 p.m., 3:30 p.m., 4:01 p.m. - - usually every 45 minutes after 12p.m. ========================== Key Market Open / Close AUD - 6:00p.m. - 12:00a.m. Japan - 7:30p.m. - 1:00a.m. China/HK - 8:30p.m. - 2:00a.m. Germany - 2:00a.m. - 11:30a.m. London - 2:30a.m. - 11:30a.m. New York - 9:30a.m. - 4:00p.m. ========================== [Edited by LearnerGuru on 1/26/2019 6:40 PM] | ||
^ Top | |||
LearnerGuru![]() Member ![]() Posts: 29 Joined: 12/3/2009 ![]() |
GLOBEX STRATEGIC POINTS 06:00 pm Beginning of the Trading day 08:00 pm Opening of the Trading Session in Sydney 08:00 pm Beginning of the Tokyo Trading Session Trading gets more dynamic. It is worth to check if there are any current corporate relations between Asian investors. Keep an eye on news and headlines. 09:00 pm the 2nd hour of the Tokyo session. Singapore opens Within these 30 minutes time frame or right before the Chinese session starts, Euro usually declines until the very moment of the opening of Shanghai and Hong Kong. 10:30 pm The 2nd hour of the Chinese session. Euro grows If China is selling some part of its U.S. Dollars, it makes Euro grow. Pay attention to the actions of the Chinese Markets. If the Chinese buy Euro, then everything is predictable. But if they buy dollars, the whole day might be like a “casino”. 01:00 am The last hour of Tokyo session 02:00 am End of the Sydney Session 02:00 am End of the Tokyo Session The end of the high volatility and liquidity period on Asian-Pacific markets 03:00 am End of the Shanghai Session If Euro has been actively sinking till this hour, China will usually drive Euro up after the Tokyo closes. 02:00 am – 03:00 am is in a manner of speaking the “Chinese hour” on the market. Hong Kong, Shanghai and Singapore are still open and take matters into their hands causing the growth of Euro. 03:00 am European Session starts London Stock Exchange, FBW Frankfurt, SSE Zurich enter the market and the European - Asian overlap starts. According to my observations Euro selling normally prevails over the first hours of the trade. Forex trends begin during the trading hours of the Europe and will usually continue until the beginning of the U.S. session. Liquidity goes down closer to lunchtime. ------------------------------ 05:00 am End of the Singapore and the Asian Trading Sessions 07:00 am Trading is drawing to a close on the Middle Eastern Oil Markets 08:00 am End of the Middle Eastern session 08:30 am U.S. Core Session starts in 1 Hour 09:30 am Core trading session in the U.S. and Canada starts 10:30 am 2nd hour of the U.S. Session. Last hour of the European Session 11:35 am End of the European Session 03:00 pm The Last Working Hour of the U.S. Session 04:00 pm End of the U.S. Trading Session 05:00 pm End of the current Trading Day | ||
^ Top | |||
jpb![]() Veteran ![]() ![]() ![]() Posts: 168 Joined: 5/11/2005 Location: Brown Deer, WI ![]() |
Adding OOPS to GA, that's an interesting idea! | ||
^ Top | |||
jpb![]() Veteran ![]() ![]() ![]() Posts: 168 Joined: 5/11/2005 Location: Brown Deer, WI ![]() |
LearnerGuru, I appreciate all the times. I'm assuming those are all in EST? Thanks for the data. | ||
^ Top | |||
John J![]() Veteran ![]() ![]() ![]() ![]() Posts: 272 Joined: 8/3/2010 Location: Leduc, AB ![]() |
LW came up with the OOPS strategy more than 20 years ago. I wonder how well it performs in today's market manipulated with HF trading and bots..? | ||
^ Top | |||
LSJ![]() Legend ![]() Posts: 515 Joined: 8/17/2006 Location: Citrus Springs, FL ![]() |
Incorrect results removed [Edited by LSJ on 2/1/2019 2:37 AM] | ||
^ Top | |||
jpb![]() Veteran ![]() ![]() ![]() Posts: 168 Joined: 5/11/2005 Location: Brown Deer, WI ![]() |
LSJ, Those are some very interesting results! I haven't looked enough at GA. Did you grab that block from an existing strategy and then add to it or did you build it from scratch? Thanks for sharing! | ||
^ Top | |||
LSJ![]() Legend ![]() Posts: 515 Joined: 8/17/2006 Location: Citrus Springs, FL ![]() |
Just a GA block and trade plan block with just a relative minimum stop. Here's the genes used. I remember seeing LW live where he showed how he tested simple bar relationships in Tradestation. Compared to that I'd say the GA approach is more of a loose interpretation of his strict "OOPS" definition. Room for lots of playing here. [Edited by LSJ on 1/27/2019 5:09 PM] ![]() | ||
^ Top | |||
jpb![]() Veteran ![]() ![]() ![]() Posts: 168 Joined: 5/11/2005 Location: Brown Deer, WI ![]() |
Very nice. Thanks for sharing. My wife's going to kill me when I start playing with this on vacation! |
|
|
Legend | Action | Notification | |||
Administrator
Forum Moderator |
Registered User
Unregistered User |
![]() |
Toggle e-mail notification |