Prediction Wizard

In the previous  entry in this series we stated that wee need to  have  the  Reference level, the target levels as potential profit areas and the corresponding  probabilities to reach those targets  to make  correct decisions  about scaling.

In this  entry  I try to  highlight these a bit more  trough  a specific example, that could be generalized.

The following example  will  be for playing range markets as they  constitute more than 50% of the trading periods, we experiencing.

The following schematic  drawing will demonstrate the example:

RW: Range Width, which Is    L3 – LRef  = 4 * RW / 4

LRef: this is the level, at wich we open a Long position, expecting higher price levels.

L1:  LRef + RW / 2 = LRef  + (2* RW/4)   This is a level, that we might use to  close all or part of the open position

L2: LRef +  3 * RW / 4    This is a level, that we might use to  close all or part of the open position.

L3:  LRef + RW  This is the level, at wich  we might close all or part of our open position.

STOP level:  This level  is  RW / 4  level below the LRef reference level, and we use this as our predefined Stop level.

We have our schema, now we select a time frame, select a  specific asset type, and  collect  the appropriate statistics.

Since we trade the NASADQ100 and S&P500 ultra funds , that cover the  NASDAQ100 and S&P500 indices, we use the NASDAQ100 ultra fund “QLD”  in our example. (Which moves 2% , if the underlying NASDAQ100 index moves 1% in the same direction.)

To translate the  trading schema into real trading example we demonstrate the  following  piece of “QLD” intraday chart:

This chart is a 5Min chart, and playing the  strategy we are demonstrating requires the essential  of identifying  the market real time as a Range market.

(In future  blog entries we will write about this early range market identification, which is  the focal point of using this strategy.)

Trading a range market  we need to define the Range first.

In our case it is  not a predefined   range, but defined by the market dynamically as the day proceeds.

We basically use the  Bollinger bands on the 5Min and the 15Min time frame to help us in this, and use  the SMA20, and Std 1.85  parameters, as very often  price levels does not reach  the SMA20 +/-  2*Std levels (Which is the default Bollinger parameters)  moving toward the range extremes.

So the RW, Range Width is the  2*1.85*Std, which is our  Bollinger width  at  the time of position opening, But it is dynamically changing going forward, so if the volatility increasing, than the Bollinger width increasing, and correspondingly  our Range is increasing and vice versa and we will use the current  Bollinger width / Range Width  to guide our exit strategy.

The SMA20 is at the center of  the move, and it is the level of L1 in our case.

The lower Bollinger  is the LRef, reference level, at which we open our long position.

The L2 level is  LRef +3/4 RW level, which is halfway between the SMA20 and the upper Bollinger level.

Our L3 level is the Upper Bollinger level (SMA200 + 1.85 STD)

For  these settings, we collected statistics,  as we wanted to know:

What is the probability during intraday range market, that  after hitting the Lower Bollinger band, the market   changes direction and reaches the  SMA20 level,  our L1 level?

What is the probability during range markets that the  QLD will reach the LRef + 3/4 * RW, which is 75% of the Bollinger width?

What is the probability, that the market will reach  the other end of the range, the upper Bollinger level, which is L3 in our case, and  RW  = Bollinger Band Width above our LRef reference level, and position opening level?

If our example the STOP level will be  one quarter of the Bollinger Band Width, which is RW / 4 and it will be initiated after position opening, below the LRef reference level.

The  following table demonstrate the probability results in our market, collected recently:

What might be surprising is that the market reached the midpoint level,  or the SMA20  more than 90% of the cases.

The market reached 75% of the Bollinger width / Range width  74.6% of the cases, which is very high, and reached the other  end of the range, the other Bollinger (Which is in our case 1.85Std  from the SMA200 level)  55.8% of the cases.

Running  our optimization  model we wanted to find out, what is the best  scaling strategy having the presented statistical parameters:

The result:

The best case is to close 100% of the position at L2 level or LRef + 0.75*RW, but  similarly good, if we  close 60% of the position at L2 level, and try to  let it run to the L3 level, the other range extreme.

Obviously total profit  could be better than theoretical, if we let at least part of the position try to run to the L3 level, and  not allowing it to become a loser  position part, once it is already moved through the half-point, or the L1 level.

The worst position playing strategy in this case to close 100% of the position  at midpoint level, which is L1 level.

I remember  in the early trading  period, as a newbie I did just that, scared to  lose  the profit already  in the position, once it made a sudden move  against the position.

The win ratio will not be the best,  running the position a bit further than the L1 level, but  the total profit is considerable better in the long run.

The difference between the  best and the worst position  management in this case about 20% for one  occasion, which adds up  during long periods.

Playing this strategy successfully requires knowledge and extreme patience.

With  a little bit of mistake our long–term results might be much different.

Why?

Lets calculate the Risk / Reward ratio in this scenario:

Assume that we committed to play  about optimal, and close 100% of the position at L2 level, and use a STOP –loss as RW / 4 below the LRef, position opening level:

With these settings our reward is 0.75*RW, our risk is 0.25*RW, the Reward to Risk ratio is 3, which is very good.

(If we would allow 40% of the position to run to the other range extreme, and in case it does not reach that level after passing the range midpoint we do not let it below midpoint, than the R/R ratio could be even better than 3.)

Now If we are not that alert, and open the position a bit above the LRef, say 0.25*RW  above the LRef level,  so about 25%  of the Bollinger width above the lover Bollinger level.

In this case the Risk to reward ratio, shooting for the same target:  0.5*RW / 0.5*RW = 1

So as we see, the  Risk / Reward ratio quickly decreasing, as we compromise the position opening, since we still must place the stop levels below the  lover range extreme,  or in this case below the lover Bollinger level by about 0.25*RW.

What can we do?

To  stay consistently successful even if we managed to correctly identify a range market and playing it  optimally we  need to stick to some rules here:

What we do.

The market either come to us, reach the lover  range extreme, or we do NOT open position.

It is very  tough to  accept and  to keep,  but  that is why it is a RULE and we must stick to that.

Obviously using different asset classes, different time frames,  different settings to define the rules, we get  a bit  different  solution.

Additional information to this is that range market identification is not really possible 5 – 10 minutes after market open.  The more we wait, the more  valid our decision might be, but the more opportunity we might lose also.

A compromise to wait 20 – 60 minutes before making the conclusion is acceptable. There are many other market internal  indicators, that can help us in this decision making. 

On the other side playing this strategy during very narrow range market might be a killer strategy, as  the potential reward is very low, but in this case we have increased probability for a market breakout, when the loss could be meaningful, if for example we use mental stop and delay the exit even if   just  10 – 15 seconds.

Another note that a bit safer play of the range market could be, If  you open positions a bit delayed in case of  sudden extreme moves, which is  in case of  SMA20 + 1.85 *Std  opening positions not at the time, when the prices passes the range extremes, but  only when the price return to within the range,  or  expressing differently opening position not at the time, when the price moves outside of our Bollinger settings, but when moves back and touches the BB that way, though this might be difficult to implement in the heat of the battle.

Sometimes it is really difficult  to  identify moves and correctly distinguish it  from breakouts.

To do that we also need a lot of experience.

Also the probability for range market is  higher during Midday periods (Not in the first 60 – 90 minutes and Not in the last 60 – 90 minutes, as higher volume periods might lead to trending  characteristics with a higher probability.)