Forum Replies Created
-
Thanks Tovim , good idea , i will try it .
on the futures i note the sudden change in volume , so maybe add days until the third
friday , as an input for calculating the predicted OBV .
Well i spent the weekend trying to use the OBV on actual stocks ,
generally it failed , worked for MSFT . Makes me think volume is distorted in Stocks by HFT. I will try some bond futures .
admittedly i only tried 7 stocks but the futures look a better bet .
Hi there ,
if you take marges graph from the June newsletter , as i am playing with the volume
indicators.
The OBV , on balance Volume , it is the second input ,
delete it and re insert it as the last input .
write down all the indicators , and open a new prediction and put them in any order you like.
in reply to: Too few trades in predicationsOn further study , my directional signals are not as robust as i would like.
When the market is rallying strongly , my system will loose money as the market makes new highs , it will be short and the drawdowns will be severe .ouch .
I have a gut feeling that the HFT have distorted the volume in the stock market which is why i have never had any luck with them , i never thought of using volume indicators in the futures market but i am guessing that the volume figures might be more valid . If anyone has a view on that , let me know .
Regards
Michael.
Oops , on the historical downloader ,
Marge has informed me that there is a new version that defaults to version 7 .
So that part is redundant , sorry .
in reply to: Automating Adding SymbolsHi Stu ,
Well I have not done this yet but it is something i was thinking about for a portfolio of stocks, 35-50 stocks.
Have the historical downloader installed,
In your C DRIVE add C:\NeuroShell Trader 6\DailyQuotes
” this allows you if you want to use yahoo finance , if you want to ,BUT it gives you access to the folder to update your data into. I don´t know where you are getting your data.”
You can setup a sub folder to save your data BUT and this is a big but,
In your work spaces change the symbol list to not as a 3 alphanumeric but as a number 1 to 50 or how ever many stocks you have .
You said you are already splitting the name etc . so change the 3 letter code to 1-50
keep the name part . so SYMBOL will be a number and name = name.
Put them in the sub folder .
Then each time you restart NSDT it will read the Data file and as long as NSDT does not come up with the box asking to retrain the data , it should retrain .
As the NDST is reading the symbol list and not the name , so when you update the sub folder with new data , ie over write the old data it will read it on opening.
As your work spaces now have symbol number and then the name for the stock .
EG
in current work space you have
JPM JP Morgan
IBM International Business Machine
GS Goldman sachs
Change to
1 JP Morgan
2 International Business Machine
3 Goldman sachs
Now your scanner comes up with 3 different stocks , you call them
1 Cisco
2 Boeing
3 Apple
Save the new data in your sub folder and over write existing data .
Theory as data is different NSDT should read this on opening.
Open all your work spaces
so now
it should read as above
1 Cisco
2 Boeing
3 Apple.
and retrain on opening.
I am guessing and i could be very wrong you are add / remove chart pages each week.
Print out the list of 50 stocks and the order they are in in case you want to change things.
You can always add more sub folders to historical data
First time round you will have to do it manually , that is change symbol to a number .
and then select in your charts and save all the work spaces with the new chart designations.
Hope this helps
Michael.
in reply to: Automating Adding SymbolsSomething to think about and why using the vix in the underlying inputs is a good idea. I THINK SO .
Volatility-Managed Portfolios
<p class=”note note-list”>76 Pages Posted: 12 Sep 2015 Last revised: 8 Mar 2017</p>
Alan Moreira
University of Rochester – Simon Business School
Tyler Muir
University of California, Los Angeles (UCLA) – Anderson School of Management; National Bureau of Economic Research (NBER)
Date Written: October 25, 2016
Abstract
Managed portfolios that take less risk when volatility is high produce large alphas, substantially increase factor Sharpe ratios, and produce large utility gains for mean-variance investors. We document this for the market, value, momentum, profitability, return on equity, and investment factors in equities, as well as the currency carry trade. Volatility timing increases Sharpe ratios because changes in factors’ volatilities are not fully offset by proportional changes in expected returns. Our strategy is contrary to conventional wisdom because it takes relatively less risk in recessions and crises yet still earns high average returns. This rules out typical risk-based explanations and is a challenge to structural models of time-varying expected returns.
<center></center>Suggested Citation:
Moreira, Alan and Muir, Tyler, Volatility-Managed Portfolios (October 25, 2016). Journal of Finance, Forthcoming. Available at SSRN: https://ssrn.com/abstract=2659431 or http://dx.doi.org/10.2139/ssrn.2659431in reply to: Continuous futures and High LowHi StuC , thank you for that , i just signed up for Quandl
Got the Vix continuous futures.
Because I was using the VIX index as the insert other data .
As I am trying to do a portfolio of 30-50 stocks and using the High low probability
to get slow trade signals eg 1 or 2 trades a month per stock , which adds enough trades for a month.
I saw that chart from WSG from Marge , and that was why, ages ago , i have been using VIX as an input. The index is nice but you can not trade it per se , and the futures and options have so much time decay it distorts the answers. So diabolical to use as a hedge with the DIA etf.
So I am trying to use the ANI to predict the percent change in the High and low of the vix to give the trading signals , using as 2 inputs and 5 other inputs in the prediction .
Thanks again for your input , appreciated
Michael.
in reply to: Continuous futures and High LowOn the Predictability of Stock Prices: a Case for High
and Low Prices
Massimiliano Caporin, Angelo Ranaldo and Paolo Santucci de Magistrisin reply to: Continuous futures and High LowAnother idea you can try , is locking the values of an input so that they align together.
eg. if using lag indicator , you might lock the lag period to align with another lag period that you are using.
Lag low 1-10
Lag High 1-10
By making them joined , you might get a better result , so they end up with
both lag 7 . rather than lag low 3 and lag high 6.
Good luck , you have so many choices to what you want to try. it becomes addictive.
in reply to: Too few trades in predicationsGood morning , a couple of ideas , try changing the search criteria , eg
Winners minus loser , change the time frame you are looking at .
Look for a pattern in your inputs that maybe you can add .
sometimes reduce the number of hidden neurons .
Also try putting your variables in a different order .
Have a great day
Michael.
in reply to: Too few trades in predicationsHi to Stu C , are you thinking of loading it on the cloud , eg Google cloud, AWS, or Azure ?
Michael.
in reply to: Keeping track of your NeuroshellHi Nate , i am processing the raw data in excel , usually i am using the natural log changes in prices etc , statistical inputs , then changing to a CSV format ,
Then i group the data by various search levels, eg , i want all the data that is around the same volatility as the last line of data i am using ( the current bar) , you can sort the data by whichever criteria you think will work for you. I also use the ABS value of the search , so that i get positive and negative results , eg , if you used the percentage change daily . if you do not use the ABS , you would only get a sort from say low to high , I don´t think that is a good way to get balanced data
On the percentage trades , i am looking at a barrier trading idea , I have an 80% chance that the outcome will be below a level or above the level but the problem is how to get into the trade at the right level as i was trying to avoid having too do intraday trading , only wanted to trade at the end of the day or the open .
IN the NSDT getting training 82.57 and 80.46 and out of sample 81.72 and 77.42 respectively based on change in open.
Along to go , i am afraid to say .
Michael.
in reply to: Chaos HunterHello NeuroSpark, well , this is something i have done with the Chaos Hunter,
I took approximately 2300 days of back data and used it in excel , putting extra inputs etc. I then did a sort of the data to end up with 413 days .
I then randomized the data and used 349 for optimizing and 63 days for out of sample test , I managed to get R2 of .78 , that holds up on the out of sample data set as well. Good trading but very low number of trades.
The other big problem is the Skew and Kurtosis especially as the market has been in an up trend for so long .
So i was impressed that CH could solve the problem on all randomized data with the search range extending from + 18 percent to minus 18 percent .
I think i will be able to get a better R2 reading by defining some better sort criteria .
I added some extra columns to be able to test the trading against the signal in excel. Not used in CH optimizing.Hi guys , at the moment , i love the chaos hunter ,
if you get the right inputs , it seems harder to do but the results work especially on the out of sample . I have the ANI but i prefer the chaos hunter .
I hope to be trying a very complex setup in Chaos in the next month .
So i am voting for Chaos hunter.
regards
Michael.
