Philip, I can now do monthly, weekly, whatever…I figured out the mechanism to do so technically. You had mentioned that the RSI Monthly filter will not yield enough trades on 4H bars, so that’s why we decided on RSI Weekly filter. Recall I ran original RSI Monthly filter simulation on Dow-30 Daily data; hence the decision to run RSI Weekly filter sim on the 4 hour data.
Yeah I think there was a miss communication on my part. When I meant the monthly will be more picky, it was a compliment lol. But you are right though. There won’t be that many trades. I’ll do them manually and if the results are good I’ll ask you to fix the code so we can also get the avg. win, avg. loss calculations automatically. I can also do it on all the forex pairs since I have the data.
Ok EURUSD yielded only 2 trades. Both in 2015. The first yielded 22 pips and the second 606 pips. I will keep updating this posts with other results.
GBPUSD: RSI gave a signal starting October 2008 and into 2009. The 2009 trades would have been for a loss, although those losses would have been dwarfed by the 2008 wins. I couldn’t trade them because I only have data since 2012. There were no trades from 2012-2015.
USDCAD: +1161 pips, +86.8 pips. Missed the 2009 trades due to lack of data.
Philip, I just re-ran with monthly RSI filter, and I got the exact same results as those with RSI Weekly. Unless I’ve made a mistake, it looks like RSI Weekly or Monthly filter yields the same results on the 5-year EUR/USD 4-hour data I have. Again, I’m using go Long on RSI > 66.67 and going short on RSI < 33.33, assuming other algorithm criteria are met.
No I just manually traded EURUSD on monthly, it yielded two trades only. Let me share my charts.
I think I just found a bug in my RSI calculation on the interval required. This is going to take some analysis on my end. Sorry for all this; I need to look at it in earnest. Unfortunately, part of the process, and thanks for challenging this.
No thank you for double checking. Just so we’re certain (and since I’ve done them) here are pictures of the charts.
So here we see on the monthly chart that RSI 14 closed below 33.3 only at the end of December 2014. So we start looking for trades since the beginning of 2015.
Here is the first trade, 22 pips of profit.
Here is the second. 606 pips of profit.
So two trades, 100% success rate. a total of 628 pips. I think that would equate to $6280 if you were to do it on wealth lab but I’m not sure. I don’t know how it calculates the value of a pip per dollar there.
Anyone of the thread participants such as Bucketman, Rookie or Jalapenofan can confirm that these were the only two trades that occurred since 2015.
Just so you guys know, there may be a limitation imposed on my Wealthlab version, which is no longer supported by the vendor. I spent a lot of time yesterday revisiting my RSI implementation. Natively, Wealth lab allows you to easily obtain Weekly or Monthly information from Daily information - that is why I was able to apply Monthly RSI filtering easily from my Dow-30 Daily data.
They do have an Intraday -> Higher timeframe function, but upon closer inspection it appears that the 4-hour -> Weekly and 4-hour -> Monthly extrapolations are NOT working as advertised. I think the intended use for this functionality was developed for Day Traders so that 1 minute bars could be extrapolated to 5 minute bars, 5 minute bars could be extrapolated to 15 minutes bars, etc. It doesn’t seem to be scaling right for 4 Hours -> Weekly and 4 hours -> Monthly. I thought it was, but I looked at the RSI charts more diligently last night, and it’s not calculating these “macro” RSI’s correctly.
This certainly is a backtesting issue when trying to apply Weekly and Monthly RSI filtering on 4 Hour bars. I’m going to look at this again more tonight, but early returns suggest that I may be dead in the water in trying to automate this.
So, to date, my Dow-30 Daily (w/RSI Monthly filtering) back-testing CAN be trusted to be accurate, and my EURUSD 4-hour (w/no RSI Filtering) CAN be trusted, but my EURUSD 4-hour (w/RSI filtering) CANNOT be trusted to be accurate.
More on this as it develops.
That’s ok. Thanks for giving it a shot though.
I can definitely do a Daily RSI(14) from the 4-hour bars, but doing a Weekly RSI or Monthly RSI doesn’t look feasible at this point, unless I get a brain rush on how to solve. Here’s how the RSI Daily Filter Test looked:
Profit Graph:
Performance Data:
Trades Log:
I’ll be happy to test other data if you guys can provide me with 4-hour backtesting data for other CCY pairs.
After reviewing this thread, and trying to trade it by writing trades down from the past and seeing if they would profit or not, I gotta say this is great strategy. I am still trying to understand what is trying to be accomplished with RSI and such, but then again, I’m not sure what is going on with it. Can anyone explain what the RSI and all the tests are for? The method works great. I don’t see a reason to change it if that is whats happening.
The intent of the RSI filter was to use an additional method of filtering out the chop - that is, to help avoid taking trades when the instrument is not trending, since this is a trending system. The value of the back-testing is to try to simulate trading Philip’s algorithm in all kinds of markets, in all kinds of time frames, and with different instruments, to see how the system fares for longer durations.
Hello PhilipPirrip,
first thank you a lot for the described system. It looks good but I still have some problems with it. Please have a look at the following example. I have a buy entry on the dailys and then the trend changes. After many pips downwards a sell signal comes which would be the exit situation. I have seen even much stronger moves with bigger losses. My question is now if this is just part of the game and if you also have that huge losses (in relation to your timeframe) or if you have a way to prevent these?
Thank you.
Dough I’ve been thinking. What if I give you the dates where RSI 14 monthly was in an extreme, can you run the test starting and ending with those periods? So rather than do the test from 2009-2015. Can you do it for example from September 1, 2010-November 17 2010 and so on?
If possible the may be we can automate it by creating an RSI 14 dashboard. Whenever the trader gets a notification he starts to run your code.
Its part of the system unfortunately. Trading on the daily really requires smaller lot sizes as well. That’s why I prefer the 4HR.
The problem is it’s a lot of manual checking, with different Price Series involved to match-up (Monthly vs. 4-Hour), and although I can certainly do it, the goal was to be able to check this in an automated fashion for many years of data across a whole bunch of CCY pairs. I think we really need to accomplish this to verify all historically, and to be able to test different hypotheses as well. I’m confident that your primary entry/exit algorithm along w/Fib extension is working well in my back-testing model. It’s this last piece, and this shortcoming in Wealth-Lab that is a royal friggin’ pain in the ass.
I’m still looking at it to see if there’s a way to automate this, but currently, I’m not optimistic. If somehow I can get this piece to work, then we can verify more data, and I know that I can then build a piece of software that will publish (maybe on a website; maybe in this forum; maybe on a Google Share, etc.) all future potential trades to look for in real-time.
But, as the back-testing has proven, the algorithm by itself (w/o chop filtering) will produce some losing years.
Hi Philip,
If you take a quick look at GBPUSD, depending on your H4 chart, the EMA 50 is above the EMA 20, and the stochastic is or is about to cross its own overbought average. My concern with this setup is that both EMAs are now pointing upward and may cross in the opposite direction. In stronger, successful directional moves, EMAs look to hold their positions well, and since they are now bending in the “wrong” direction, would you avoid trading this setup (nor is this first overbought setup with this pair).
I will likely pass on this setup. Am I being too conservative? I have to believe that this sort of filtering would be of greater value than the RSI approach. :28:
Well it’s actually the second signal as well which I don’t trade at all…now that’s an even better filter 
Yessir…that’s why I mentioned it was not the first setup.
Can you tell me when your next H4 bar appears? Mine will be at 7PM CST or about 2 hours from now.
Thanks again!
In 20 minutes.