Seasonal Trading of goldView all attachments
01/15/2018 at 12:02 PM #59521
I just wanted to share an idea for seasonal trading on Gold. Since it’s just and idea and not a finished strategy I though it would be better to post it here rather than in the library.1234567891011121314151617maxsize = 1If month = 1 and countoflongshares < maxsize thenbuy 1 contract at marketendifif month = 3 thensell at marketendifIf month = 9 and dayofweek = 1 and countofshortshares < maxsize thensellshort 1 contract at marketendifif month = 11 thenexitshort at marketendif
You must be logged in to access attached files.01/15/2018 at 12:30 PM #59525
Although this seems like a good idea what you are in reality doing is just writing a strategy that is perfectly curve fitted to the past. Just because buying at that time in those months has worked in the past does not mean that it will in the future. You are better to use seasonality to identify potential bull months and then buy at good value prices on dips or on confirmed breakouts rather than just because it is a certain time on a certain day in a certain month. Time is never a sound reason to buy or sell on its own.
Seasonality is only about trying to give yourself a few percentage points of ‘edge’. Avoiding buying in months that have historically been good months to short may give you (and the emphasis is on the ‘may’) an edge. Closing a long position just before a month that has historically been a month to short may give you an edge. Buying on the 1st of January because it is the 1st of January will not.01/15/2018 at 1:15 PM #59534
I think Vonasi is making a great point.
No one trades purely on seasonal stuff. From what i make of it, seasonal data can be used for positionsizing (maybe if its a historic good month then buy 3 contracts in stead of 2, and if its a bad month then maybe just 1 rather then 2.) and maybe even when to pause the system or not. F ex take no trades in august on dax, because august is genereally a bad month. Maybe this year august might be great, who knows, it depends on a shit ton factors. If it is a great month you missed out on profits, and thats sad but really, who cares about 1 month.. but if its not a great month then you might have saved yourself from loosing capital, and that is key. anyone can make profit at any time in any market, thats not the hard part.. it might be luck, it might be some great info only you have, but the hard part is keeping that profit and not giving it back to the market the very next day!
Use seasonal data as a tool, not as the main strategy. Thats my 2 cents01/15/2018 at 3:24 PM #59553
Hey guys, you are absolutely right! I just did this to test of there was stronger and weaker periods in the history. Obviously the code is missing a valid entry and exit signal to make it a real system. But as you guys pointed out, after adding these things it can improve an edge.01/15/2018 at 3:27 PM #59554
I may have an issue with writing today. Here is attempt nr 2.
Hey guys, you are absolutely right! I just made this test to see if there was stronger and weaker periods in the history. Obviously the code is missing a valid entry and exit signal in order to make it a real system. But as you guys pointed out, after adding these things it can improve an edge.01/15/2018 at 3:38 PM #59558
Hi again, okay so what youre talking about is seasonal data, check f ex this site: http://www.seasonalcharts.com/classics_gold.html And there are many more like it. On the forum there are also some indicators that analyze the seasonal data from the backtest data you have in prt.
Also pathfinder is a good example on a forum-algo (by Reiner) utilizing seasonal data in a good way, adding contracts depending on seasonal data. If you or anyone else have a profitable system today you should definitly look into seasonal charts/data!
1 user thanked author for this post.01/15/2018 at 5:34 PM #5957301/15/2018 at 7:38 PM #59580
There are seasonal periods that are non random and many apparent seasonal periods that are still full random that are skewed by 1 or 2 large events . I think identifying the non random periods with some sort of logic attached is the key to finding an edge in cyclical events . Just saying .. Hint supply / demand / moneyflow01/15/2018 at 8:43 PM #59581
I did write a reply on this but I guess I forgot to press ‘Submit’ so it never got sent! So here goes again.
Seasonality is a difficult beast. Last years data is more relevant than data from 25 years ago but one years data is insufficient a data sample to use on its own. So it is all a bit of a compromise unless you can put more weight to recent data – which is what I have tried to do a bit with my in sample and out of sample seasonality analysis rating on my indicator.
Plus some events can make seasonality data irrelevant almost overnight – the obvious example being when Argentina started growing soybeans. All data up until that point was sampling a totally different market to the data afterwards.
I’m not sure how much seasonality gold has (it is on my very long list of things to test!). I’m sure it does around certain dates but I think that it may be mainly driven by the fickles of mining supply and manufacturers demand along with much longer term worldwide financial cycles that cause the ‘rush to gold’ when ever there is a risk off situation happening.01/16/2018 at 3:31 AM #59610
I spent about 30 minutes playing here , this is long only and whilst still using a select period to get long i have quickly incorperated some of my mean reversion techniques that search for value entries and it does show some promise . Absolutely no stop losses and no optomising of my mean reversion strategies that are purpose built for indice . So would appear that plenty room for improvement . Like the OP straight out of the mould with no refinement whatsoever . I will run this strategy at random times of year to compare to the ” curve fitted ” dates to see if there is a noticeable difference .
You must be logged in to access attached files.01/16/2018 at 4:06 AM #59613
Ok tested exact same template starting at every month of year and the difference between results is very noticeable . But quite a few but not all month dates are profitable . Handy hint for the OP here sometimes the optimum entry is during the month before these strong seasonal months on an extreme swing low/high , this is where my mean reversion techniques excel . You need to write some ( conditions ) or ( conditions ) to nail down these entries . Hint , make the 1 month prior to optimal seasonal month entries greedier ( more extreme) than the actual seasonal month where condition can be relaxed slightly , You only want to enter during month prior when conditions arise that are unlikely to be repeated 2 months consecutively .01/16/2018 at 7:25 AM #59614
Interesting analysis BrisVegas especially the bit about entering the month before. I’ll add that onto my (now stupidly long) list to investigate.
Funnily enough yesterday I was working on an extreme swing low mean reversion strategy. I made a simple indicator based purely on price movement direction and quantity and noticed a regular pattern – so I wrote a very simple strategy that was profitable straight out of the box – always a good sign. The addition of one filter condition and my new trailing stop that I am still testing out and it was a very pretty equity curve. Add in money management and it had me very excited. Win rate a bit low for my liking but gain/loss ratio right where I would want it to be. Forward testing seems to be too good to be true which always gets me worried for some reason but otherwise I think it has the chance to be one of my best strategies so far!01/16/2018 at 9:24 AM #59623
@jebus89 did you notice that all metallas has a strong period in the beginning of the year? I’ll probably try to make proper system with entry triggers etc. for the first time of the year. Probably on the daily timeframe. I will do like Vonasi and add it to the the growing list.01/16/2018 at 10:08 AM #59624
From listening to the podcast “better system trader” there a some guys who “only” trade based on seasonality. Sometimes there are good reasons as to why that is. Car producers need special kinda metall that they usually buy in Q1 or something like that. Other times its just random as other users are pointing out. I think honestly that googling about seasonality, i think that there has been a lot of research that has been published online on this. Where people have studied these things for years, much better then we can do. But we can pick up on the research and make our own strategies based on that research! Im into reading about this stuff these days, ill post something here if i find it to be very interesting (but nearly everything is!) 🙂