Useful Indicator

R

Rapunzel

Hi Folks, I hope you are well.

I have just been playing around with how you can apply the concept of Sharpe Ratio and Sortino to price action to predict drawdowns on trades and also future price movement.

Quick summary if you are unfamiliar with Sortino. (https://www.investopedia.com/terms/s/sortinoratio.asp)

Sortino = Portfolio Return / St Dev Downside (volatility of negative returns) - as a measure of return per unit downside risk.

I was thinking about applying this concept to price action in the following way for example on 20 bars (H1).

(1) 'Buy Sortino' = Average Price Increase for UP bars in 20 / st dev price decrease for down bars in 20.

(2) 'Sell Sortino' = -1*Average Price Decrease for Down bars in 20 / st dev price increase for up bars in 20.

So it is supposed to measure the price increase per unit downside risk (1) and the price decrease per unit upside risk (2). And I have found from inspection of the indicator graphs created by (1) and (2) that it does have a level of predictive ability on future price action. And is a property you can model to quantify average and (especially) maximum drawdowns on a trade. If you take drawdown on a trade as the negative PnL/Risk on the trade.

(1) crossing over (2) predicts future rising prices and (2) crossing over (1) predicts future declining prices reasonably well. And I am sure it can be tweaked for even better performances.
 
Last edited:
Just back tested it on a few strategies and it works very well. The best results are when there is a recent cross in (1) and (2) and the difference between them is still quite small.

A good way to test the validity of the indicator might be something like;

for Buys: if there is a recent cross (within last 8 bars) in (1) and (2) with a small difference between their current values (0 < difference < 1), is the future price (2-8 bars) higher.

In what percentage of cases is the future price higher. And how can you modify the settings for a pair; time of last cross and difference in current values to increase the probability of a higher future price, on model data and unseen data.
 
iLikeMoney said:
Looks interesting
More...

Thanks for the post, done several tests and it works better in the H4 and above timeframe. It also works better in combination with a filter that you can use to measure the predictive reliability of an indicator for near future bars.
 
Last edited:
iLikeMoney said:
Looks interesting
More...

Hey there is edge in this trade and i thought i'd share. I've only tested the buys so far and I think that if you add your own filters you can definitely extract more edge from it.

Just using the buys for example but the sells will follow the same logic.

- FX
- Period H4
- Bars 20
- 'Buy Sortino' (BS)(20) = (average price increases for up bars in last 20) / (st dev price decreases for down bars in 20)
- 'Sell Sortino' (SS)(20) = -1*(average price decreases for down bars in last 20) / (st dev price increases for up bars in 20)

Buy Signal Logic
  • (BS)(20) Crosses Above (SS)(20) - Most Recent Cross: (BS)(20)[1] < (SS)(20)[1] and (BS)(20)[0] > (SS)(20)[0]
  • 0.75 < (BS)(20)[0] < 2
  • 0.25 <= (BS)(20)[0] - (SS)(20)[0] <= 1
  • Close[0] > SMA(20)
  • Close[0] < Upper Bollinger Band(20) @ Sigma 1
  • SL Distance = 3 x (st dev price decreases for down bars in 20) - I think anything more than 3 times multiple will diminish its value as a predictor of downside risk.
  • Minimum SL Distance = 30 pips.
  • TP Distance = 0.5 x SL Distance
The break even win rate is 66.67%. Results mostly in the range 72% ~ 84% so far, but it wont work for them all and feel like it can be improved if you add your own filters and process to get more out of it.

Let me know if you try it out and it works for you.
 
Last edited:
Rapunzel said:
Hey there is edge in this trade and i thought i'd share. I've only tested the buys so far and I think that if you add your own filters you can definitely extract more edge from it.

Just using the buys for example but the sells will follow the same logic.

- FX
- Period H4
- Bars 20
- 'Buy Sortino' (BS)(20) = (average price increases for up bars in last 20) / (st dev price decreases for down bars in 20)
- 'Sell Sortino' (SS)(20) = -1*(average price decreases for down bars in last 20) / (st dev price increases for up bars in 20)

Buy Signal Logic
  • (BS)(20) Crosses Above (SS)(20) - Most Recent Cross: (BS)(20)[1] < (SS)(20)[1] and (BS)(20)[0] > (SS)(20)[0]
  • 0.75 < (BS)(20)[0] < 2
  • 0.25 <= (BS)(20)[0] - (SS)(20)[0] <= 1
  • Close[0] > SMA(20)
  • Close[0] < Upper Bollinger Band(20) @ Sigma 1
  • SL Distance = 3 x (st dev price decreases for down bars in 20) - I think anything more than 3 times multiple will diminish its value as a predictor of downside risk.
  • Minimum SL Distance = 30 pips.
  • TP Distance = 0.5 x SL Distance
The break even win rate is 66.67%. Results mostly in the range 72% ~ 84% so far, but it wont work for them all and feel like it can be improved if you add your own filters and process to get more out of it.

Let me know if you try it out and it works for you.
More...

tested with pairs where average spread < 3 pips.
 
Hey, lots of ways to use it. Example:

entering buy trades as an example, where the SL is 2 x ATR(12).

We can look at the st deviation of the downsides in 20 for down bars (StDevDown(20)) which is the 'Buy Sortino' denominator.

and only enter trades where StDevDown(20) x M <= 2 x ATR(2). Where M is a multiple of the downside risk. In general, as the multiple (M) increases, the probability of winning the trade also increases.
 
oops! I read too fast as USELESS INDICATOR.

I have tried hundreds of indicators for many years.
Now I have zero.

Anyway, go and experiment with the various indicators.
 
You must log in or register to reply here.
Back
Top