Zigzag Bollinger Band

[snailbeard]

Following on from the above:
SB cannot take the trade because of a conflict between the long-term direction

Code: Select all

// Multi-day swing
st_iMain4D1Direction: 1,
st_H24Direction: 1, 
iResultDirection: 1
st_b3LongMusketeers: 1

and the short term swing:

Code: Select all

// Intraday swing
st_bH4ZzH4Bb_CurrSwgDir -1
st_iH1ZZ_CurrDir: -1

In addition to the main direction we also have

Code: Select all

st_b3LongMusketeers: 1

and SB never goes against the 3 Musketeers, we must go back to the earlier logic and see if we can get the 3 Musketeers to stand down:

All for one, And one for All

They should only be active during a strong swing.

Otherwise we have a raft of flags all indicating a downward bias:

Code: Select all

st_bH4CloseBelowFiboLevel: 1 
bFiboPullBack: 1

st_bH4CloseBelowWP: 1
st_bPriceCloseToWP: 1, 
bMoveUpUnderWP: 1, 
bMoveUpUnderDP: 1

Hopefully I can resolve this conflict without getting a flood of bad trades.

[snailbeard]

After improving the logic of the different entry methods and fixing a few more bugs:
SBNG finds two subprime intraday swings.
It might be possible to add stacking and overlap additional entries at D and E.

[snailbeard]

I have been trying to moderate trades at the end of a swing, but I was either removing lots of winning trades or getting a lot of stoploss hits resulting in PL-equilibrium.

There is already some adjustment for low volatility which reduces the target but nothing for strong swings.
So I thought I would try extending the target to the next weekly pivot after and similarly for the stop-loss.

Firstly, June 2012 has well defined swings on H4:

AUDUSD-H4-Jun2012.png

These were the results using the normal target and stop-loss:

AUDUSD-Jun12-ST-M1-noWP-SLTP-run20Mar14.png

After adding the WP adjustment there was a significant improvement, however, the test period is limited to AUDUSD June 2012, so it too soon to draw any conclusions but the results are quite promising:

AUDUSD-Jun2012-ST-M1-with-WP-adjusted-SLTP-20Mar14.png

[snailbeard]

Some entries can be difficult to spot especially when the market is choppy. This often happens if price has returned to the daily average price. Price will often swing backwards and forwards across this level especially if the D1 slope is almost flat.

I initially asked why did SB miss this sell setup, however the M15 zigzags are somewhat erratic and produce mixed messages. SB did initially have a sell setup but then dropped it due to very low volatility. However, there are many occassions when the best setups emerge in the consolidation phase. Something to return to later.

[snailbeard]

Its getting towards the end March, I have just finished a long round of testing and adjusting for May, which is mostly a good strong downhill ride.
The first ZzBb EA used to miss out on a lot low hanging fruit on strong swings due a limitation of the entry method. The difficulty with any swing has always been those sharp reversals counter to the general swing.
It is almost impossible to jump over these without missing out some more fruit after the reversal rejoins the main swing. However, I am combining a number flags to 'delay' entries back into the main swing.
This might require further refinement but so far on May's data it has illiminated 3 bad trades and just one good trade. The lost good trade can be regained later by adding more entry patterns.

To summarise the aims of ZzBb-2:

a) Increase the number of entries per week and sometimes per day:
I am down to 8 winning trades for one month but theres still scope to add more entry patterns

b) improve the ratio of wins to losses
currently:
8 wins, 0 losses:
due to either notorious curve fitting or brilliant data modelling (usually it ends up being the former
However, there is a lot more data-dependent adjustment of threshold parameters than before.

c) better determination of likely price movement direction by adding checks by weekly pivot, Fibonacci levels and H4 candle patterns.

Some of these changes should be more generic than over-specific curve fitting.

[snailbeard]

The Buy trades were disabled because there are still bugs to be squashed, nevertheless, running over both May and June (Buy only) it is mostly doing the right thing but not often enough:

[bazze]

snailbeard,is this a EA you have shared here in the forum?

[snailbeard]

snailbeard,is this a EA you have shared here in the forum?

Managing just one master copy is hard enough, it is very temperamental, a small change here or there has surprising consequences sometimes. I would prefer to make the signals available than try to support lots of users struggling to install it, run it, and trying to understand why it does what it does. It is also still in pre- MetaCrap build 600 source and ex4: so MetaQuotes can pull the plug anytime and it is just the kind of thing they do to their end users!

I'm sure that some of the other EA's in this forum must be much better and more straight forward to use. Many people in this forum have been writing EAs for years and have a lot more experience at it than I do.

I still think there must be a more concise mathematical representation of the core concepts which could be implemented in a few thousand lines instead of 20,000 (or more).

[snailbeard]

Determining the end of a swing requires a lot of analysis and some guess work.
If the days move is a reversal after a long hard slide down do we really want to risk going up?
If we get a stong reversal and move of more than a 100 pips when would it be appropriate to say yes we are now swinging up but where is the next ceiling.

The best guess requires the current price relative to Fibonacci levels, signs of more forward momentum after a significant pullback. so what do we have in this chart:
Walk away or take the trade?

[snailbeard]

It turns out in this particular case that there is enough strength in the reversal to take us to new highs, even though the general longer range outlook is down: