EDS Problem (ARK)

[traderx]

Since 30th November (6 trading days) Arc International has risen from £2.07 to £3.29 a phenomenal 59% gain. If we are ever going to get seriously rich this is the type of gain we need to hit. PROBLEM: I have done some testing of my best EDS scans and they all failed to select this share on either 30/11 or 1/12 bar one scan, which selected so many shares that it would not have been possible to identify this one as a big winner. What this tells me is that the share price is rising secretly in terms of key TA indicators. You will notice no significant increase in volume. This is partly linked to the observations of Cookie re MMs, whereby shares are volatile without justification. There are hundreds of EDS routines and I have not seen one that overcomes this problem. CONCLUSION: There is a fundamental flaw in EDS capability. Alternatively, the EDS system is so complex that it would require considerable expertise and knowledge to overcome this problem. I have personally tested many, many scans, mainly from American contributors to AIQ-EDS Groups but none of them particularly impress me. We need to write a routine that recognises the point when a share is oversold after ITMA cross down over STMA at the critical point where STMA is 2 or 3 days from turn around when often the price starts to rise. Problem is that moving averages often lag share price action and any successful routine must overcome that problem. Any ideas Reality and other EDS users?

Warning: I am not recommending anybody buy ARK, it will surely turn down on profit taking. By now astute traders will have taken or about to take profits so beware.

No financial advice intended

Tx

 

[Uncle]

How much data do you have on your chart for ARK. I have data only from the 21/9 and therefore would not place ANY credibility on TA indcators with such a short time span

John

 

[traderx]

Thank you for getting back to me so quickly John, good point you make, perhaps I could have chosen a better example. This problem is a general one, there are many shares that have risen rapidly with the turn around EDS scans rarely pick the big winners, (including HG) I am convinced that it is possible to write a routine to overcome the difficulty. I will keep working on it. Find it very frustrating, so if any body has any ideas then I would be happy to discuss. My E-mail is [email protected].

Tx

 

[Cigar]

Spotting the turn wasn't difficult, its the degree of rebound that is surprising. I don't have EDS but spend a lot of time checking for divergences. ARK gave classic signals and confirmed the turn on RSI-5day at 236p (not that I bought or spotted it)

I wonder if the degree of its previous downtrend gives an indication to the rebound. ARK fell ~50% in the previous 2 months, and looks like rebounding upwards to the 50% level of the decline.
(Oxford Glyco is doing the same now, with the same signals, be interesting to watch)
Shire has fallen from ~1500 to 1075 (~30%)but is rebounding to lower levels. Shire is also a bigger share.

Just a thought.

 

[spock]

Had A quick look at Arc on Omnitrader-it gave a buy signal on 27 Nov at 246 but price then fell to 208 a couple of days later!! To be fair it then gave some secondary buy signals so anyone still in them would have caught the rise.
Interesting that you should mention OGS because that is a pretty strong buy on OT at the moment and could well be my share of next week!

 

[traderx]

Interesting debate boys, just what we need to prepare for return of better market conditions. I realise that any one of us looking at MACD crossup would recognise the signal but the question I am posing is, how can we identify a share about to go north big-style from scanning techniques? Otherwise these things tend to happen before you become allerted. It must be possible, I will eventually devise a formula but it will not be easy and might take a long time. ARK has risen quietly, no doubt intended by the MM at minimum cost to them. If we discover the right formula and I know there is no Holy Grail, then we greatly improve the RR ratio, improving probability of success. I really believe this is the key to huge profits, although I bear in mind Dr Iraj's firm belief, that profit is best made from trending stocks. We can never discount advice from the Master.

Tx

 

[traderx]

Spock I was impressed by Omni-trader when Doc Iraj was posting 'Share of the Day'. Unfortunately, I invested in AIQ and EDS which is disappointing by any standards but a good TA system on AIQ. As indicated, AIQ is renouned as complex and difficult to master but I believe it will produce the goods in the end. I will eventually buy Omni but cannot justify it right now, having spent $800 on AIQ in August and little return due to market conditions. AIQ expert ratings does not work well in bad market, that is more or less acknowledged by AIQ people. I would like, if possible for us all to discuss scanning systems to help each other. I think Omni-trader is equally relevant, if not more so than AIQ. Only reason AIQ gets attention because many members bought it. But please do share some of Omni's share picks and hopefully, the members with AIQ will do the same to achieve a concensus. I know we always worry about security of information but I am beginning to think we should have an alternative system for talking 'off air', perhaps by E-mail between members when appropriate. I hope this gives food for thought.

Tx

 

[ChartMan]

try the slope of price (last 5 days) <0 and TCI slope(last 5 days) rising.....
I got one that does that-I'll try it....

Martin

 

[Cigar]

I would be all for a discussion of share selection methods and individual selections.
I post selections and I actually hope someone will come back and say 'why on earth...., can't you see...'.
I want to learn and become more proficient. I'm more than willing to offer my views, what we don't want is people just to take and not give.

 

[traderx]

Reality, based on your suggestions I put this short scan together and guess what mate, you were right, brilliant! This routine picked did not pick ARK on 30/11 or 1/12 but did so on 4/12, on the third trading day after the bounce with 90 pence of profit still available if you got in on the 5th at or about 2.40p. It has today picked 120 stocks, too many I feel. If you run First Rule independently it picked ACC at 46 pence after collapse from £1.12. It looks interesting one for a bounce on monday. I think another member has already identified this share. Moving on, this scan looks disappointing when back tested with a RR Ratio of 1.13 but I think it could be developed into something really useful. What do you think chaps?


!Negative Phase value with 5 day slope of phase < 0
BadPhaseDown if [phase] < 0 and slope([phase],5) < 0.
!Trend changed up
!If the 10 day slope 5 days ago was negative and the 5 day current slope is positive
TrendChangeUp if slope([close],10,5) < 0 and slope([close],5) > 0.

GetRich if BadPhaseDown and TrendChangeUp.

Tx
No financial advice intended

[This message has been edited by traderx (edited 09-12-2000).]

 

[traderx]

The above scan also picked CHB Chubb which looks interesting and PTY which is subject of a separate thread form Waldorf and some of the boys. Its a shame that the good Doc is not around to give a steer. Never mind, we still have Reality and Uncle John to keep us right.

For discussion only - no financial advice intended

ps. under orders from my good lady to go out with to David LLoyds - be back on BB later............. have already been xmas shopping this aft!

Tx

 

[traderx]

By the way, just before I go out, I share this with you. I was down at local Mercedes dealer this morning and saw the new 'S' Class 280. Must be one of the most beautiful cars in the world. Worth every penny at 46k.
THAT IN A NUTSHELL IS MY MOTIVATION

end of message

Tx

 

[spock]

Traderx,
Interesting your comments re AIQ which I tried and did not like- I much prefer OT though of course its not perfect and as always you have to "filter " the selections.It has, however, easily paid for itself in the first month despite me being too eager to buy its selections when I first got it and I'm happy to give its input when required or when I think it shows a real winner.

 

[Dr Iraj]

Hi Trader XXX.

Your observation is admirable. I feel I owe you this explanation, as your point is one of the most important and relevant questions in advanced technical analysis. I try to explain exactly what the problem is.

Here we go.
When you write an EDS routine you are introducing a combination of indicators with a predefined settings.

Example

Buy if RSI < 30.
The indicator is RSI and the setting is 30.

Now you have made two assumptions in here
1) 30 is the correct settings for oversold stocks
2) RSI is the correct indicator to trigger a buy signal for ALL STOCKS.
Now if you had an EDS routine to satisfy the above condition then you end up with say 20 oversold stocks.
But are these selections really oversold? No of course not. All we have short-listed is 20 stocks with RSI < 30. Our assumption is that RSI <30 means an oversold situation. This is generally true but market does not give you that MERCEDES if you go by generalization.
Oversold if RSI <30 is naïve and assumptions like that triggers loads of false buys.

What is the solution?

Short list the stocks and then go back to stocks one by one and see how many times the stock has bounced off your predefined RSI. (In this case 30) If for example stock A has been bouncing off 20 times (in the LAST PAST 12 MONTH) of the figure 15 then the oversold situation is 15 and not 30. So your original assumption of RSI< 30 was not an oversold situation at all. What you need to find is that pattern that a certain stock has been reacting towards a certain indicator in the past as a bases for the future. If stock A behaved badly and kept ignoring RSI and bounced off different values all the time then RSI is THE WRONG INDICATOR TO BE USED FOR THAT STOCK.

****** YOU NOW, UNDERSTAND WHY ONE SHOULD BE CAREFUL WITH EDS AND ITS CAPABILITIES? ******
.

This is so so so so important and I insist you to understand it.
The modern software automatically finds that relevant indicator and its appropriate settings for a given stock in the past and tries to use the settings to trigger a buy signal. (The technology uses Genetic algorithms and its job is to search for the correct indicator and its settings to forecast the future price movement.)

FIXED SETTINGS AGAINST ADAPTIVE SETTINGS

AIQ does not offer adaptive algorithm to find the best settings and the best indicator. It is you who choose the indicators and not the software. New software goes as far as selecting the settings as well as the most appropriate indicators.

WHAT HAS IT GOT TO DO WITH ARK?

A lot actually. Firstly ARK has no history. (Hence you cannot confirm the stock reaction to various indicators) Secondly AIQ cannot consistently pick stocks, which creep up based on FIXED SETTINGS AND THE TRADER DICTATED CHOICE OF INDICATORS. If you wish to find those creepers then you need software, which can study stocks individually, and find the most influential and correct indicator and optimized settings. These software study individual stocks and signal if a 3 days successive rise is the start of a long term trend or a just a short term price adjustment to positive news.

WHAT NEXT?

If you know any one who owns Neuroshell then you are one step ahead of every body else. Give me a ring if you need help re Neuroshell. Try www.neuroshell.com

My own interest in research is in approximate reasoning, fuzzylogic, neuro network and their applications in financial price modeling. Neuroshell uses all above.

I hope I have answered your question but if I have not please sent me an email where I can give you my Tel to discuss this matter further.

PS:- How correct of you to clock the limitation in AIQ? (Still is the best screening software around. This is why I am using it so often).

As soon as my loft is converted to a trading center I will invite Uncle, Cigar and those serious traders for a get together.

Mean while keep your interest in TA. It works and it works well as long as market stops its stupid random behavior. The sooner the better.
(Any body knows any software which can give a future lottery numbers based on past numbers? This is my email [email protected])

 

[Uncle]

Three Cheers Doc.............Nice to know your still taking an interest in this temp BB, and am sure your explantion above will be of benefit to many.

Am sure I speak for the majority in saying we hope you will continue to make a contribution as often as possible until such time as your are able to get the new BB up and running again.

Best wishes

John

 

[spock]

Hi Doc!
Could not agree more with your point-the market is continually changing its character and any predictive system must change in line.What was a good indicator in a bull market may be bad in a bear market or may not be suited to a trading range market so the parameters set for "trigger" points should also change as the market changes.
Various neural network software packages attempt to do this in a sophisticated manner and systems like OT continually optimise the parameters for any indicator by backtesting over a specified period which moves forward with the market.These more dynamic systems in my opinion offer greater chances of success.

 

[traderx]

Hello Doc

I am overwhelmed that you should once again help us out with this basic problem. As always you know immediately the solution. I underline the comments of John. It is great to see you posting again. I rather hoped you might be keeping an eye on things but I never thought you would post at this time. I am sure you have made many members happy this Sunday morning to know you are still around. I hope you are well. Turning to the issue, I now understand the rationale behind this thinking and amazingly, I have not previously applied the methodology of 'right indicator' and 'right level' for the indicator to my trading. Brilliant! By the way the fuzzy EDS model is proving to be very effective, given that market conditions have been adverse. If anybody had traded those shares with the highest fuzzy factor, then they have made a profit. I am keeping an eye on progress of fuzzy model by keeping records and when favourable conditions return I will trade for real on the basis of the model. Look after yourself and hope to hear from you soon.

Kindest regards

Mick (Tx)

[This message has been edited by traderx (edited 10-12-2000).]

 

[traderx]

I have been looking at Boots (BOOT) over the past 12 months. The method is 100% correct when applied to this stock. THIS IS POWERFULL STUFF. The indicator that works with Boots is Stochastic. Every single time the reading was between 3 and 5 the stock took off, putting on huge short term gains. The significant thing about Boots being that it has not, in itself, been a good performer this year until last couple of months when it saw an upturn. Nevertheless, the stochastic seems to be totally reliable. I used to trade Boots almost exclusively and make money from swings in price over 8 days (T+10) the good old days when I used java charts on UK Invest, mainly looked at stochastic when red line went steep north over green from low position. The amazing thing is, I never did know why I was so successful until today. I stopped trading it when retails stocks fell from grace. It was no accident that I managed to extract capital from Boots company. Only now am I beginning to realise that very little happens by chance in this market, hence the dictum, the chart knows the news is coming. Incidently, Boots is presently retracing from an upward trend, stochastic 39 suggests it will come off a lot more towards lower BB before the bounce when the stochastic reaches a figure of <5. I could be wrong but lets watch it and see.

Tx

Tx

 

[shelman]

Glad to see you're alive and kicking, Doc.
Steve

 

[Joe]

Hi Doc,
Beware of neurostock which is similar to neuroshell. They don't respond. I am still awaiting a permanent licence depite a number of requests and my version no longer works.

I'm not sure whether they are genuine but incompetant or what.

Be very careful.

Regards
Joe

[This message has been edited by Joe (edited 10-12-2000).]