Wednesday, 30 May 2012

white soldiers

oasis analysis of candlestick White Soldiers yields some interesting results, at least we can compare.

A White Soldier is described as the de-facto 'in your face' indication of a price increase.

The results are interesting because they appear, at least on first glance, to be a little random.

To the right/above is a screen shot of a random selection of companies and the dates that White Soldiers were triggered, along with close price, next day open price and then percentage increases over 5, 10, 15 and 30 days. 

The parameters for White Soldiers stipulates a downward trend needs to have been established, but does not say how and so I have used the psar on close to denote a change of trend. I will be experimenting with other trend setters.


Below is an interesting extract from a report from the current companies that have been traced.

The columns are, company, average percentage increase, max percentage increase and number of times this has occurred over the last 1.5 years:

AMER     11.7     12.2     4
AMER     12.5     13.3     4
PFD         14        17        4
DOM       15.6     17.4     4
PFD         17.3     21.3     4
TCM        17.5     21.8    4
PFD         17.7     21.3    4
DOM       18.1     21.4    4
OPTS       18.7    20.9     4
PFD         18.8     21.3    4
OPTS      19.9     20.9    4
PFD        18.4     21.3    5
OPTS     18.5      20.9    5
OPTS     18.8     20.9     5
OPTS     17.8     20.9     6

As you can see the last line for OPTS, the formulae have been triggered 6 times with a consistent 85% average percentage increase of the maximum percentage increase and consistently over 11.5%.
The element combinations for this one line are:

dfnotv     OPTS          17.8                   20.9                   6
dfnrtv      OPTS          17.8                   20.9                   6
dfnstv     OPTS          17.8                   20.9                   6
dfntv      OPTS           17.8                   20.9                   6
dfortv     OPTS          17.8                   20.9                   6
dfrstv      OPTS         17.8                   20.9                   6
dfrtv        OPTS         17.8                   20.9                   6

All of these formulae trigger at the same points, so essentially we have an extra dimension to the specific betas for each company. Not only can we analyse when individual betas trigger but also we can analyse relationships between more than one beta triggering as well.

The upshot of this is that for OPTS we have 7 formulae that when triggered give a good probability of an increase of at least 11.5% over the next 5 days.

Lets have a look at the dates for one of the triggers - dfntv
(for information, d is related to the stochastic on the macd, f is related to the psar on the adx, n is related to the psar on the vi, t is related to the psar of the psar of the close, v is related to the stochastic of the rsi of the opening price):

OPTS 22-MAR-12        215.6         208.5          17.1           34.1          34.1           34.1
OPTS 23-MAR-12        220             216            21.6            29.4           29.4            29.4
OPTS 26-MAR-12         221           216              22.8           29.4            29.4             29.4
OPTS 27-MAR-12        221.5          221             20               26.5         26.5              26.5
OPTS 28-MAR-12          222.777       221           23.6              26.5          26.5          26.5
OPTS 29-MAR-12            243.25        221.25       26.3             26.3         26.3           26.3

The columns above are company, trigger date, close price, next days opening price, and then percentage increase over the next 5, 10, 15 and 30 days.

And what we see here is again backed up by the previous post, triggering of a formula within certain periods indicates quite an increase probability of a price increase. Here the period is 1 day.


The way oasis is set up makes it very easy to search for candlestick patterns, so while the company traces are running (2 hours per company, but running in parallel and so my throughput is currently 10 companies per hour, still going to take days to complete though), I am coding the candlestick pattern recognition engine and entering the world of Doji, Spinning tops, Marubozu, White Soldiers, Black Crows and other assorted candlestick patterns and their meanings.

I like the fact that candlestick pattern analysis is another way to layer formulae on top of, and included in, new beta formulae, as well as a method to substantiate current betas

Saturday, 26 May 2012

ACHL analysis

Some initial results for the people from ACHL

Below is the listing of the trace elements, company, average percentage increase and max percentage increase for ACHL for the last 1.5 years, a period that is relevant for these calculations but not too distant as to mar the results.

I have designated my trace elements as alphabetic characters, so for the first line, for the first column, each of the letters a, f, m and o designate different elements to a formula - the individual elements being concatenated to produce the full trigger.

Just as a note and to back up the previous post, the formula for 'a' is 'trigger on psar on open and psar on close together' - below you will see the figures for just element 'a' in the second section.
(and just for info, 'f' is related to the adx, 'm' is related to the macd, and 'o' is related to the vi)

The last 2 columns are the average percentage increase using this trigger over the subsequent 5 days and the  maximum percentage increase over the same period. The percentages are calculated on the next day's opening price, the approximate price at which we would buy.

afmo               ACHL                                               113.3                  113.3  
afn           ACHL                                               113.3                  113.3 
afno         ACHL                                               113.3                  113.3
afo         ACHL                                               113.3                  113.3
ag       ACHL                                               113.3                  113.3
agl         ACHL                                               113.3                  113.3
aglm           ACHL                                               113.3                  113.3
aglmn        ACHL                                               113.3                  113.3
aglmo           ACHL                                               113.3                  113.3
agln       ACHL                                               113.3                  113.3
aglno          ACHL                                               113.3                  113.3
aglo          ACHL                                               113.3                  113.3
agm          ACHL                                               113.3                  113.3
(this is only part of the report)

We have quite a few formulae that indicate not only that the trigger to each of these formula gives one answer and that answer is that the stock will rise but also that these are quite separate and different formula, even though all have the element 'a' in them.

If I scroll further down the report to get away from the 'a's I get:

a                ACHL                                               65.3                   113.3
bgo             ACHL                                               45.3                   128.2
bdgo             ACHL                                               45.3                   128.2
bcgo            ACHL                                               41.1                   122.8
bcdgo          ACHL                                               41.1                   122.8
bcego            ACHL                                               36.3                   108.6
bfglo             ACHL                                               36.3                   108.6
bfgo                ACHL                                               36.3                   108.6
bdego            ACHL                                               36.3                   108.6
beglo                ACHL                                               36.3                   108.6
bdfgo                ACHL                                               36.3                   108.6
(this is another part of the same report)

What we can see here is that taking 'a' out of the equation reduces our average percentage increase, but increases our max percentage, but only temporarily... it may appear that the above report is ordered by the first column (which is a random select of elements designated a lower case character) but in fact the list is ordered by the third column, average percentage increase; the interesting aspect is that the same elements appear repeatedly in the high probability, high percentage formulae.

Lets focus on the top listing as this lists formulae that if triggered yield the highest probability of a rise. I was not expecting to see so many different formulae giving the same result, this is quite interesting, I was expecting only 1 or 2 triggers, and I will be investigating further. I need to think a little more about this - but what I can do is post here when the formulae in the top section get triggered in the near future for ACHL.

Another aspect that is interesting and also beneficial to the processing time is that the highest yielding formula have between 3 and 6 elements only. This is also a surprise. Percentages fall off rapidly after 7 elements, and there is little use in calculating these - most 7 and 8 element triggers return no results and so there is no point in adding even more elements, and 1 and 2 element triggers mostly yield inadequate percentages. What this means is that for 26 elements, instead of calculating 67,108,862 iterations, the trace will only need to process 313,560 combinations.

Friday, 25 May 2012


Been doing some deep thinking about exactly how this all works

Decided that I needed psars on the positive and negative adx's and on the open and close bollinger bandwidth percentages, adding an extra 24 indicators (bringing the total to 167) and 2 graphs

The MXP updated charts with these new graphs and in high definition detail are now on flickr here:

What is more interesting is that I have programmed a trace system, taking individual elements (either single or complex) and allocating success coefficients, and then going through all non repeating combinations of the elements per company. As you can imagine this takes a lot of processing as well as time - for 20 elements there are 1,048,575 separate combinations. The trace elements consist of the individual elements that have been used in the beta's such as simple psar on close, or more complex such as stochastic on the psar on the rsi = 0 over the last 9 days.

What I have noticed is that the traces are very date and actual close price dependent, and so merely taking an average will not work. We have to be a lot smarter to get to the essence of each individual company.

For a given formula oasis is able to calculate on average how much the stock has risen over 5 days as a percentage. As an example lets take the simple element of the psar on close. Below is a random selection of companies that indicate the average percentage increase over subsequent 5 day periods if we followed just  the psar on close indicator and only taking into account the last 6 months:

ACHL 64.3
ACL        0.133
ACM 16.1
ACMG 7.65
ACS        2.55
ACTG -0.9
ACU 15.25
ACWD 0.05

What we see is that we would have done pretty well to follow ACHL, less so the others and not at all with ACTG.

Lets have a look at ACHL in detail for a second - following are all the dates that the psar on close indicated a buy with the corresponding percentage increases over the subsequent 5 days in the last 6 months.

ACHL 09-JAN-12 110.5
ACHL 07-MAR-12 18.1

And so ACHL would have been an excellent company to trade on just the psar on close over the last 6 months.

And what about ACHL from about Jan 2010, ACHL's average 5 day increase over this 2.5 year period for just psar on close is 21.1%, with the detail as:

ACHL                 07-MAR-12                 18.1                
ACHL                 09-JAN-12                 110.5                
ACHL                 12-OCT-11                 75.3                
ACHL                 16-SEP-11                 1                    
ACHL                 11-AUG-11                 -2.2                
ACHL                 29-MAR-11                 -0.4                
ACHL                 29-DEC-10                 4.4                  
ACHL                 15-SEP-10                 6.9                  
ACHL                 02-AUG-10                 4.2                  
ACHL                 27-MAY-10                 11.1                
ACHL                 20-JAN-10                 4.7

On average still not bad, but as you can see we do have a couple of negatives here, not large ones but still negative.

If we add in another element, for instance the simple psar on open, then we have 3 combinations: psc, pso and psc and pso together. Lets see what this looks like for ACHL:

psc - psar on close is above

pso - psar on open is:

ACHL                 05-MAR-12                 19.1                
ACHL                 05-JAN-12                 116.2                
ACHL                 27-OCT-11                 1.5                  
ACHL                 19-SEP-11                 -0.5                
ACHL                 12-AUG-11                 0.5                  
ACHL                 01-JUN-11                 2.5                  
ACHL                 30-DEC-10                 6.9                  
ACHL                 03-AUG-10                 6                    
ACHL                 24-MAR-10                 4.2

and psc and pso together is:

ACHL                 07-MAR-12                 18.1                
ACHL                 09-JAN-12                 110.5                
ACHL                 15-SEP-10                 6.9

And so we have eliminated pretty much all of the unwanted triggers, and we are left with an average 5 day increase, post trigger on both pso and psc of 45.1% that has worked over the last 2.5 years.

What this means is that if we were purely watching the open and close psars on ACHL and buy when these are both triggered then within 5 days we would sell with an average price increase of about 45%.

What we can also see is that the triggers also seem to be date related. Clusters of triggers are noticeable around months, which indicate a much increased probability of price rise. That is, if you see one trigger in one month then that is good, if you see a second trigger soon after then its very good, and if you see a third trigger, go all in - I will post here on the blog if ACHL triggers on psc and pso in the next few months, as you can see from the charts below ACHL is currently on a down trend.

I have uploaded the full high def oasis charts for ACHL onto flickr here:

(For the high def chart, choose the chart, then choose it again, then choose 'view all sizes' (top right) and then choose original)

And so oasis keeps adding element combinations until it has gone through all possible iterations for each company. What I am really looking for are the formulae where the individual percentage increases are above 11.5%. 

What the trace does is highlight company's specific success coefficients to produce better, more tangible beta formulae as demonstrated above. From analysis the current beta formula work well, but lose in the timing strategy, both too early entry indicated and too early exit. It is these 'too early' indicators that this process has been designed to address - currently there is nearly always a slight dip in the share price before the 11.5% increase realisation which is accompanied by a variable length of time. I could just wait and watch, and indeed this is what I have been doing and testing different strategies on the gaming sites, but the answer has to be better beta formulae.

The crux of this work is that the traces allocate success coefficients to specific companies for specific beta formulae - that is, no longer is the beta an all encompassing, all things to all companies type of formula, but a more specific, dedicated company beta - specific company, specific date, specific 5 day period.

I have just started running the traces.

The traces link in very nicely to the RNS analysis; this is proving a lot faster (we only need to analyse specific lists of companies rather than all of them).

If this leaves you really confused but intrigued to know what this is all about then you may want to have a look here:

Thursday, 10 May 2012

# hashtag twitter

Added hashtags onto the twitter feeds for ease of search and sorting.

These will appear in subsequent auto runs, starting... now!/spacecadet99999

Friday, 4 May 2012

lets see some charts in detail

I have created an HD flickr photo library so that I can upload detailed screen shots of the charts I use and you can have a look. I have uploaded CWC, TCG and MXP  - charts people have requested.

The charts can be found here:

Please choose the chart you are interested in, and then choose it again, then pick 'view all sizes' and choose 'original' - the chart will be about 5mb.

If we look at the MXP chart, straight away from the first graph we see that the psars quite closely follow the rises and falls of the share price close (in black). psc is psar on close, pso is psar on the open price.

As a help, mc is macd, st means stochastic, bb is bollinger, rsi is rsi, etc.
o, l, c, h at the end means on open, low, close, high.
So strsic means stochastic on rsi on the close price,
And stpsvi is the stochastic on the psar on the vi - its just vi, there is no concept of open or close here.

For the alpha and then the beta formulae I go through each graph and try to figure out what is being indicated and when with respect to risers.

Please note that these graphs are very date dependent, I have noted the date on the upload.

If there is a company that you would like to see, please comment and I will load.

we have lots of news

The RNS (regulatory news service) analysis has proven more useful than originally thought. oasis can now analyse companies that are within a month of publishing, have just published and post publish. The interesting analysis here is that the 14% of risers above 11.5%, half of these (7% of total) are greater than a 20% rise.  It is worth following the RNS's as part of an LSE trading method.

I have a couple of RNS related alphas that are proving useful and will continue to formulate to produce a beta.

At the same time I am running extensive real life testing on new betas within the LSE (London) share site here:

And searching for share gaming sites to run tests for the US betas.

My real life dealing of the tweeted betas, I am happy to say, are proving themselves.