Design Issue on Lowest and Highest functions impacts on Donchian Ichimoku
- This topic has 18 replies, 6 voices, and was last updated 5 years ago by
.
-
Hi GraHal,
not sure how to call them in PRT, but basically when there is no price from the market within the 1min, the candle is not drawn on the chart. In PRT code, you will find out that close, high, open, low are undefined.
Of course, this occurs more often on lower TF, this is a normal and correct behavior regarding what is happening on the market.
The issue I’m pointing at, is not about undefined candles(prices) but the way they are used within standard functions.
for lowest[26](Low), it looks as if PRT finds the lowest on the latest 26 defined Lows. I think this is incorrect, PRT should look at the 26 latest units whether they are defined or not, and just ignore then if they are undefined.
For example, in your code, you want to exit your position when Close crosses the Kijun. If a strong up trend occurs, followed by a long Range possibly containing undefined candles(prices), then the Kijun will be calculated with prices taken before or within the up Trend yielding a Kijun which too low, yielding to an incorrect exit signal.
I spotted the same behaviour on the Donchian channel, yielding possibly to late signals.
1 user thanked author for this post.
Bonjour Nicolas,
I’ve tried doing the loop with min/max functions, and it works fine, but it is very time-consuming during calculation, specially when you some backtesting. I didn’t understand the second workaround you’re proposing., but for sure it will be much slower than a native function. Still, would you mind explaining your idea?
Moreover, I have multi TF strategies which implement Ichimoku on different TF. Hence, if you consider the SpanB on the 1H TF implemented in PRT on 1min TF, it means you’re looking at 52*60 candles backwards. The effect of this design issue gets worse in such cases.
Worse now, I have some doubts on how undefined prices are dealt with on other functions. I haven’t tested Moving Averages, I thought I could rely on native functions. I think all of them need to be tested regarding this issue.
Yes, there could be a lot of impacts in the indicators.
1 user thanked author for this post.
You’re not going to look back in the previous 5 min candlestick for some data. So really, why shouldn’t it be the same for the Highest and Lowest functions?
I agree. If a bar is missing / undefined then the calculation should miss out the data (lack of) from that bar.
If in MA(50) bars, 5 are missing / undefined then the average would be more accurate (for all the reasons Etienne shows in posts above) if calculated using the data from the 45 bars that are present / available … in my opinion.
1 user thanked author for this post.
Find exclusive trading pro-tools on 
