How good is RSI for predicting extreme events analysis.

• 03/09/2020 at 11:40 PM #121620
  Vonasi

  The recent massive volatility and massive drops in index prices due to coronavirus and oil prices got me thinking about how good the RSI indicator is for predicting that a market is oversold in such extreme conditions. Often it will enter oversold territory but then stay there for quite a while and so if you thought the first bar that it showed oversold was a good time to buy then you would be a bit disappointed as the market continued to plummet. So I then I got to thinking about analysing how long it usually spends on average in oversold levels might be useful so that we can see the really extreme events – that is an RSI that is at a low level for an extraordinarily long time. So I coded this simple indicator which counts how many bars the RSI is below a certain level and calculates the average number of bars that it usually stays below this level as well as storing the maximum run length ever achieved below oversold. Anything above the average might be considered an extreme event.

  I think the results are quite interesting. If we look at the FTSE100 daily with the common RSI14 and oversold level of 30 then we can see that the current run of bars under 30 is now 3 compared to the historical average of a run of 2.5 days so that could be considered an extreme event but not as extreme as the run of 10 days below 30 that happened in October 1987.

  Some further analysis of the results and also looking at the reverse of the indicator which analyses overbought as well as turning the idea into a strategy has concluded for me that actually the RSI is not a very good indicator for telling us what is overbought and what is oversold if we take just the really extreme events for our analysis.

  //p = 14 //RSI period //lev = 30 //oversold level if count <> 0 and rsi[p] <= lev then count = count + 1 else count = 0 endif if rsi[p] <= lev and rsi[p](close[1]) > lev then count = 1 endif if count = 0 and count[1] > 0 then total = total + count[1] qty = qty + 1 avg = total/qty maxcount = max(maxcount,count[1]) endif return count as "count", avg as "average", maxcount as "maximum count"

  1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21

  //p = 14 //RSI period //lev = 30 //oversold level   if count <> 0 and rsi[p] <= lev then count = count + 1 else count = 0 endif   if rsi[p] <= lev and rsi[p](close[1]) > lev then count = 1 endif   if count = 0 and count[1] > 0 then total = total + count[1] qty = qty + 1 avg = total/qty maxcount = max(maxcount,count[1]) endif   return count as "count", avg as "average", maxcount as "maximum count"

   

  3 users thanked author for this post.

  dannesson, Edmond, GraHal

  

  Vonasi

  Moderator

  Topics: 192

  Replies: 5088

  Been thanked: 1811 times
  03/10/2020 at 10:16 AM #121655
  Nicolas

  Vonasi wrote:

  concluded for me that actually the RSI is not a very good indicator for telling us what is overbought and what is oversold

  Seriously? You broke my dreams! 😥

  

  Nicolas

  Keymaster

  Topics: 81

  Replies: 20830

  Been thanked: 4743 times
• Author
  Posts