This could have gone into the introductions but the word beginner was in there...
I've been working with Excel VBA for about 25 years, mostly doing data analysis of price history of major ETFs and stocks. Over the last year or so, I've been doing it full time.
I've been looking around for a suitable platform to discuss my work. Have a website; written a few articles. It could be turned into a software product, or maybe a book, Mostly interested in seeing the concept survive and gain recognition rather than making money from it.
The following 7 users say Thank You to semiopen for this post:
Hi semiopen - You will find a few Excel aficionados here, I am an Excel junkie myself.
So you're more than welcome to discuss your work here and I'm sure many will be interested.
A word of caution though: this site has a very strict No self-promotion policy, which means anything that can be construed as self-promotion for monetary gain is a no-no.
Other than that, I look forward to read your work!
The following 3 users say Thank You to xplorer for this post:
I'm a retired brokerage systems developer - front and back office. I decided to seriously continue development of my application, sort of a last project. Mostly I'm interested in feedback. The code/logic/methodology isn't a secret, mostly just a way of self expression similar to painting watercolors. I'm not sure how good my analysis is; seems pretty good to me - false modesty isn't one of my vices.
I think price action should be analyzed holistically - "characterized by comprehension of the parts of something as intimately interconnected and explicable only by reference to the whole." (from google)
This is a similar concept to Yin and Yang. "a concept of dualism in ancient Chinese philosophy, describing how seemingly opposite or contrary forces may actually be complementary, interconnected, and interdependent in the natural world, and how they may give rise to each other as they interrelate to one another."
Traders seem to be obsessed with situations that are profitable. I understand why profitability is important, but one has to wonder whether the single minded pursuit of an edge is the best way to achieve that.
Finite State analysis is a way to look at and hopefully understand the entire price action story.
I'll try to post concrete stuff now and again.
The following 3 users say Thank You to semiopen for this post:
I've been playing with graphics a little, meant to do a little writeup here but didn't understand what URL was needed to upload an image so I put it on my website - Three Types of Profit.
It has some analysis for SPY using the 31 day rate of change.
This is a pivot table, analyzing weekly LSMA (Least Square Moving Average) performance from April 3, 2006 through yesterday. The various columns are important to me and are described on my web site and seekingalpha (to some extent).
32 conditions are analyzed by the app - 8 different lengths of four indicators. The four indicators are: Simple MA, Exponential MA, Rate of change, and Least Square MA. I call the 8 lengths Friedmanacci numbers: 3,7,13,21,31,43,57, and 73. Possible timeframes are daily, weekly, and monthly.
There are two possible states for each condition - Above (A) and Below (B) zero (or some other number).
When the possible states are combined, as above, HProf is the buy and hold profit for the condition(s) being analyzed as well as the buy and hold profit for the instrument in general. The app invests a fixed amount of $10,000 every time the state changes and books the profit or loss for the previous state at the same time. Note that RProf (Reinvest) is higher than HProf for IWM for this time period.
This is the same information with the three profits types having running totals.
Note how IWM has lost money when above 13L but made money below.
The market turmoil during 2008 is partially to blame for the RProf higher than HProf anomaly.
There is a pile of observations that go with this analysis. For example, A is the "positive" state, "B" is the negative. A transition to a negative state makes money on the entry and exit, a positive state transition will lose on the entry and exit. Generally positive states last about twice as long as negative states. Negative states have a higher winning percentage but also have higher risk. LSMA has a remarkably even distribution of Above and Below lengths.
Entries and exits can be improved, the most obvious trading strategy is to buy the negative state at some point below the transition price and sell the positive state at some point above the transition price.
This being a futures forum, this may not apply as much, but some form of buy and hold is a great strategy for equities. I haven't heard many trading mentors dwell on that fact.
The following 2 users say Thank You to semiopen for this post:
Greetings Josef and welcome to futures.io hopefully we can have some productive conversations.
Interesting article. While a simple 43 Week Moving Average Crossover system can quickly be programmed in something like Tradestation your AR0 and AR1 analysis would be very difficult to perform.
Given your AR0 and AR1 conclusions, if you now revisited the analysis trying to improve your states (switch earlier or later) do you feel that you would have introduced a bias?
I performed a similar analysis once and came to similar conclusions. It's well known that while the SPX has an upward bias it's short term movement is often mean reverting. Hence combining a shorter moving average and a longer moving average (NOT a crossover system) - ie creating a 4 state system - greatly improved results - until 2018 or 2019 when it imploded impressively!
Also given the 10 year bull run of the stock market it would have been interesting to see how it performed a) in 2008 and 2009 and b) commodity ETFs like GLD, UNG and USO.
I will read the rest of your thread in the coming days.
The following 3 users say Thank You to SMCJB for this post:
I analyzed a 3 day 7 day rate of change strategy in late 2018, buying when both were positive, selling when at least one indicator turned negative. That worked well for 2017- mid 2019. I noticed it performed less well in 2015 (to say the least). That jump started this project as I decided to look more carefully at the performance of the two indicators as a four state system. Prior to that, I did all my strategy development on TradeStation. I had to program the strategy accounting on Excel and by the 10th iteration or so of the prototype, the algorithm got pretty good by my standards.
There is a technical argument that strategy development in its current form is unsound and I suspect that I have proven that although it is surprisingly difficult to interest people much less convince them. It is not so important if that is true or not, it mostly provides a design theology (if that makes any sense). My guess is that my stuff has some quite original views in some areas mixed with a certain amount of reinventing the wheel because I have no formal education in the areas of computer science, statistics, mathematics, or quantitative finance.
The charts above suggest a natural log relationship between the different profit measures (at least after a cursory glance - 2 weeks ago, I knew nothing about the subject), so I'm interested in doing something with that.
The accounting application can currently analyze 8 above/below conditions at a time. That can be increased simply by changing an array size.
Four states gives you something like AA occurring 50% of the time, BB about 25%, with AB and BA splitting the difference. At some point the population of the obscure states is too small to draw inferences and then there are naming convention issues.
Regarding non equity based futures, I think the analysis is still appropriate. When one analyzes the states in something non uptrending, certain states will produce losses, we don't really see that with the equity indexes - IWM is interesting that way. Just looking at some data briefly, the above state seems to become a loser before the below state.
The following 3 users say Thank You to semiopen for this post:
Here is some analysis based on state change entries. Trying to keep the Theocratic messaging to a minimum, but the typical strategy trader doesn't look at both sides as we are doing here. Whether my analysis is sound might be debatable, but my guess is that not considering stuff like this at all has to be a real disadvantage.
The table below is comprised of 16 major ETFs. DIA,IWM,QQQ,SPY,VTI,XBI,XHB,XLB,XLE,XLF,XLI,XLK,XLP,XLU,XLV, and XLY. Based on Daily data from 6/29/2012 through yesterday. The states of 32 indicators, 8 for each of the 4 types (E,L.M,R) are considered for each ETF. Accounting for virtual trades is done at all state transitions.
The groupings are by level of the indicator (x1) at a state transition. Observed results range from -10.8% to +17.4%. The averages are based on the fixed amount of $10,000 as the initial investment. The distribution of TLen (total length) CStat (count of state transitions) and CWin (count of wins) looks quasi normal to me. APPer is the average profit per period. Note how violent state changes to the upside are dubious to go long on.
The table above deals with 07M (7 day moving average). Note the distribution is still pretty normal. and the red/yellow flag over strong up moves.
The following user says Thank You to semiopen for this post:
