All market gains since 1993 have occurred after hours

[Overnight]

Now here it is with all of the filters/rules on. Only trade Monday/Tuesday, only trade if the market has dropped by at least -0.5% in the past 5 trading days, reinvest returns.

It's not completely obvious from the graph but when looking at the trade metrics, this test has significantly better performance.

View attachment 201068

View attachment 201069

Testing testing testing. I did the same thing you did, and I never made any real money. By the time my years-long tests were completed, the market conditions completely changed, and when I went to live mode, nothing worked.

Stop with the madness. Start trading live, and STOP WITH THE SIM BACKTESTS.

They will only hurt you.

 

[krugman25]

Testing testing testing. I did the same thing you did, and I never made any real money. By the time my years-long tests were completed, the market conditions completely changed, and when I went to live mode, nothing worked.

Stop with the madness. Start trading live, and STOP WITH THE SIM BACKTESTS.

They will only hurt you.

I already have multiple strategies a currently trade. I am not going to rush to add another. I will if/when I see a strategy to trade.

You sound like you need a few more energy drinks.

 

[Overnight]

...

You sound like you need a few more energy drinks.

You sound like you need to go live. Start making the monies Lebowski!

 

[cruisecontrol]

If it were that easy, everyone would have done it. It WILL NOT WORK. Your backtests are flawed. Not your fault, per se, it is just that it doesn't work in RL

Bro, it totally does work "in RL". You shouldn't appear SO CONFIDENT when you don't actually know what you're talking about. Not a good look.

 

[krugman25]

This is research done by someone in a different trading group I am a part of. This person primarily trades IWM. This was research on overnight gap down risk. Somewhat related to what we are talking about here.

"Here's a pretty interesting study on overnight gaps down in IWM over the past 20 years that I thought I'd share. I wanted to see what the actual number of occurrences was for an overnight gap down of 2% or greater in IWM (and essentially RUT), since I sell puts on that each week. I use stop loss orders but in the event of a large gap down, the stop loss will not protect me. I downloaded intraday data from Yahoo Finance, and added a column that would calculate the % change from the opening price of the day compared to the closing price of the prior day. Summary - there is around a 1.2% risk of an overnight gap down of 2% or greater."

 

[Seaweed]

I ran it from 2013 to present. This has no gain reinvestment, commission+fees, spread (1c/share), and no additional trade filters.

I'm not sure how to put your results into context with my results.

First thing is that I wasn't able to go back to the beginning of 2013 like you since I didn't have the data, so mine started Oct 24.

Second thing is that I see your PnL is $36k, but I wonder what this is based on. Buying 100 shares? Even though we are almost a year apart in start date, my study would net 759 ES points, which is $38k, so a somewhat similar number. But when put into context, this was an average of 0.45 ES points per day. This means less than $25 per day, and then I had to subtract slippage from this and commissions, which made it negative.

So I'm not sure how you can account for a generous $10 commissions and come up with a nice positive number.

I understand what Overnight is saying with the idea of forward testing this live. But I think we can easily simulate this a little better. I looked at tick data for the ES, looking to enter just before the close and exit before the open, and I can confirm exactly what price I would get within a 1 tick (since I lose the bid/ask spread). I think what needs to be done before you can trust your data is look at a chart, see where you can get in just before the close and exit just before the open, and compare how this relates to your OHLC data you're using for the SPY.

I specifically don't want to use daily data because I want to see where I'm getting my fill in those last few seconds. I also don't know what data they use to print those daily bars. Is it the 4pm close? Is it the 4:15pm close? Is it the 5pm close? Maybe if you can show the daily log chart for the last few weeks, we can compare this to the tick data of the SPY and see how it behaves if we were actually trading it, and see how the results actually stack up if we were to trade this, not just ran stats on it with historical daily data.

It doesn't make sense that you end up with 36k in profits after you deduct commissions and slippage, and I end up with a negative number.

 

[krugman25]

I'm not sure how to put your results into context with my results.

First thing is that I wasn't able to go back to the beginning of 2013 like you since I didn't have the data, so mine started Oct 24.

Second thing is that I see your PnL is $36k, but I wonder what this is based on. Buying 100 shares? Even though we are almost a year apart in start date, my study would net 759 ES points, which is $38k, so a somewhat similar number. But when put into context, this was an average of 0.45 ES points per day. This means less than $25 per day, and then I had to subtract slippage from this and commissions, which made it negative.

So I'm not sure how you can account for a generous $10 commissions and come up with a nice positive number.

I understand what Overnight is saying with the idea of forward testing this live. But I think we can easily simulate this a little better. I looked at tick data for the ES, looking to enter just before the close and exit before the open, and I can confirm exactly what price I would get within a 1 tick (since I lose the bid/ask spread). I think what needs to be done before you can trust your data is look at a chart, see where you can get in just before the close and exit just before the open, and compare how this relates to your OHLC data you're using for the SPY.

I specifically don't want to use daily data because I want to see where I'm getting my fill in those last few seconds. I also don't know what data they use to print those daily bars. Is it the 4pm close? Is it the 4:15pm close? Is it the 5pm close? Maybe if you can show the daily log chart for the last few weeks, we can compare this to the tick data of the SPY and see how it behaves if we were actually trading it, and see how the results actually stack up if we were to trade this, not just ran stats on it with historical daily data.

It doesn't make sense that you end up with 36k in profits after you deduct commissions and slippage, and I end up with a negative number.

I'm pretty sure it's the 4PM close. Even my 15 minute resolution data from TickStory stops at 4PM. I attached the trade execution sheet in my first test yesterday. You can download that and see what prices it is getting filled at and compare it to yours.

This is 500 shares to get close to 1 /ES contract.

 

[ironchef]

That said I think you may also need to calibrate your expectations on what value can even come from such a thread. Traders in their right mind will not reveal any edge. If you really do this for a living, you should understand this is not a collaborative enterprise.

Looking back over the thread, I am amused that some people seem to be denying that that the overnight effect even exists, especially a guy whose handle is "Overnight". Perhaps they protest too much?

No, he/she trades it and doesn't want us to know it works.

 

[Seaweed]

I'm pretty sure it's the 4PM close. Even my 15 minute resolution data from TickStory stops at 4PM. I attached the trade execution sheet in my first test yesterday. You can download that and see what prices it is getting filled at and compare it to yours.

This is 500 shares to get close to 1 /ES contract.

Aye, so I started going through your data to try and initially match things up because I was getting varied results when I looked up individual days. (ie. I saw maybe a 20 ES point gain in my data, but I couldn't find it in yours when I tried to apply a simple formula to figure out the profit from your open and close data). It of course doesn't help that there are differences between dates. (ie. Is the date of the trade the day after since it closes just before the open, or is it the day its opened, which is just before the close the previous day?)

I then realized that my data is bad in the sense that I have 6 bars for every week, when I should only have 5. Must have something to do with Sunday, so its not that its bad, but there is an extra day. I then fixed that problem, but honestly, at this point, I'm too frustrated to move forward. I would never trade something like this but was curious enough to see if I could quantify it. I'm a total amateur when it comes to Excel and my trading is done visually.

Anyway, good luck with the rest of your study.

 

[krugman25]

Aye, so I started going through your data to try and initially match things up because I was getting varied results when I looked up individual days. (ie. I saw maybe a 20 ES point gain in my data, but I couldn't find it in yours when I tried to apply a simple formula to figure out the profit from your open and close data). It of course doesn't help that there are differences between dates. (ie. Is the date of the trade the day after since it closes just before the open, or is it the day its opened, which is just before the close the previous day?)

I then realized that my data is bad in the sense that I have 6 bars for every week, when I should only have 5. Must have something to do with Sunday, so its not that its bad, but there is an extra day. I then fixed that problem, but honestly, at this point, I'm too frustrated to move forward. I would never trade something like this but was curious enough to see if I could quantify it. I'm a total amateur when it comes to Excel and my trading is done visually.

I agree, just getting the data ironed out is half of the battle. By the time I got my data ironed out I was too exhausted to want to run any tests. haha. I have another support ticket out to NinjaTrader because I am now having another issue with the platform.

Anyway, good luck with the rest of your study.

I appreciate it.