Favorite Links

Monday, June 4, 2018

Psychology and System

Jan 14, 2018


I am writing to continue “Parameter Optimization & Portfolio Selection” in FAQ (10/25/17)

Thank you for your consideration of my professional development and your feedback to help me improve. I smile as I read your post many times and think about how I will implement your feedback in my trading and professional development.

What I take away from your feedback:

1) the steps behind security selection and my personal system (heat level, commitment etc) are more important and impactful than trying to optimize the last bit out of return through Python or C#

2) Specifically I should focus on identifying my process for selecting cryptos & how I do that again

3) You don’t understand why I include time of draw down in the denominator of bliss as it already has a time component in the return/numerator.

4) I consider your Tribe ARs seriously: learning a new language, finding the perfect system and letting go of relationships.

Security Selection (Potential New Systems)

My process in selecting cryptos as a sub-system:

1) look for instruments with large potential payoff. With Bitcoin large potential payoff simply means that the bull market continues (bitcoin has made no new 3 month low in the last 2.5 years).

I find the bull market most visible by looking at a long term log chart due to the huge return and volatility, which can obscure the upward movement.

2) Thinking about my process one thing I notice is that I accidently select the feature of disruption in Bitcoin. But this disruption of where/how services are delivered seems important because it is in common with the biggest monster trades of the last decade:

Apple phones take on computer functions, Netflix mails DVDs, Amazon delivers IT over the cloud, Bitcoin currency is created and used without gov’t permission.

Ex ante v Ex post “discovery” of disruption opportunities is problematic. But I think disruption with trend following the number of opportunities is more manageable.

Also in considering potential instruments, some “+” factors, Antifragility is a plus, Govopoly- either going with the flow or limited but effective resistance (ie bitcoin) is a plus; liquidity is a plus.

3) try trend following sub system parameters and look for negative correlation of monthly returns with both total system monthly returns and existing sub-systems monthly returns

4) optimize new sub-system parameters to get higher return and lower draw down and correlation

5) tinker to see if new sub-system fits well with existing total system and adjust the heat of all sub-systems to find the right portfolio mix.

(It's really #1 and #2 I focus on for finding sub-system adds, #4 and #5 are just what I do to optimize but if I have poor results from those optimization and tinkering steps then I either trade with the low heat the system defines or throw out the new sub system entirely if the added return or lower drawdown are not significant enough for trading costs and extra time.)

I have 2 sub systems in planning that could be good fits for “the next monster trade”: Commodity Volatility (vega trading through options) and Interest Rate Spreads.

Both of these systems are ones that could be profitable under increasing “Govopoly” as the regulations and excessive debt creation of the state will likely cause more volatility, especially in debt based instruments and across countries, so they could meet the large potential payoff requirement.

Government control of interest rates could be disrupted in a small way through a new inflationary cycle. For #3-5 I believe they both should be uncorrelated with most of the other sub systems I trade (there may be some correlation with my existing sub-system long term bond futures – I plan to mimimize this in system development step #3). I work on these potential sub-systems now in Excel and will deprioritize the mathturbation of optimizing through Python or C#.

Security Selection (Optimization in Cryptos)

The bull market in cryptocurrencies continues. Maybe the next monster trade is the same as the last monster trade? I make no changes in my Bitcoin system except to drop GBTC and trade futures as they launch in Dec, I tweak my alt coin system and add a “venture capital” ICO approach (work in process).

A) I continue with my long term trend following Bitcoin system

B) I back test and optimize my all time high stock system for use on altcoins (cyptos other than Bitcoin). My testing shows alt coins have similar price movements to the lower cap stocks I trade successfully with this system.

C) I consider ways to add uncorrelated return streams and one thing I remember reading is that traditional venture capital return (ie pre-IPO investment) is slightly negatively correlated with S&P returns over long periods. In addition I find data that shows first day IPO performance during the internet bubble expanded significantly. “In the 1980s, the average first-day return on initial public offerings (IPOs) was 7%. The average first-day return doubled to almost 15% during 1990-1998, before jumping to 65% during the internet bubble years of 1999-2000.” (“Why Has IPO Underpricing Increased Over Time?”, Loughran, Ritter 2002). Though cryptocurrency ICOs (Initial Coin Offerings) have fewer intermediaries than 90’s internet IPOs similar market conditions exist in cryptocurrencies today.

It is possible that pre-ICO or ICO returns could offer large returns and also have negative correlation with cryptocurrencies (my steps #1 and 2). If this is true I would be able to optimize in cryptocurrencies by using a small portion of capital as my own personal venture capital firm (with trend following). I seek to take small positions in ICOs that are most popular (the trend following aspect) and then upon ICO launch trade the coins within my alt coin system. If the boom in cryptocurrency securities is similar to the boom in internet securities then it is likely that there will be a couple homeruns (ie Amazon) and many strike outs (ie I work to define system (steps 1-5 above) start small and track return and correlation of returns.

Bliss (or RRO)

I poorly explain my personal version of a bliss calculation. Allow me to take another try. First maybe “bliss” is the wrong name for what I’m doing because using time in the denominator destroys the earn-back frequency output of bliss as you pointed out. Instead of bliss lets call it RRO (Return/Risk Optimizer). I’m using this RRO function to optimize portfolio selection and heat.

A) The premise for including time in the denominator of RRO is that I believe trading failure can come through either hitting the uncle point from the depth of the draw down or hitting the uncle point from length of draw down.

An extreme example for depth of draw down: 80% loss from portfolio all time high and the trader exits trading or switches systems just before things begin to turn around. An extreme example for length of draw down: no new high for 8 years such as Dunn Composite Performance from '04-'12 see Dunn Composite Performance at

In Dunn's personal case he "rode the bucking bronco" as he said and continued trading successfully, but I wonder how many customers got bucked off and exited Dunn during that 8 year bronco ride to a new high. I wonder if I trade his system if I can stay on the bronco for 8 years without a new all time high… and my stomach feels tight and a little queasy. I believe including time in the denominator acknowledges that time since an all time high plays a factor in the trader's willingness to commit to his or her trading.

B) Using time since all time high in the denominator of RRO makes a 2 dimensional simulation of the "lake" of the draw down by creating a rectangle with "width" as % draw down and "length" as time of the draw down since the last all time high. I'm not sure how best to represent this mathematically but in my calculation of RRO at a total system level I use: (CAGR%/((max DD%x2)+max months of DD)) where max DD% is a negative number and max months of DD is a positive number.

The result is a number I try to optimize (higher is better) by adjusting system variables including instrument selection, sub system heat, number of individual positions, total heat allowed, etc. This is just the formula that I use which suits me. I believe there are individual differences in sensitivity of % and time for each trader.

C) In addition to uncle point considerations if the trader is optimizing for their feelings (including trading willingness) and they feel better the shorter the draw down, then including time in the denominator may allow the trader to choose system parameters that minimize both draw down depth and length and better optimize for the trader's feelings. The amount of feeling preference for draw down length vs raw CAGR/DD% preference can be adjusted by adding a factor to either DD% or months of DD.

D) If return is y axis and time is x axis, using % draw down to budget subsystem heat allows for something I think of as “amplitude matching” of subsystems to minimizing the total system % drawdown. In the same way I believe using length of time of draw down (since all time high) to budget subsystem heat can allow for “frequency matching” of subsystems to minimize the total system length of draw down. Control over frequency is less direct than amplitude and I think more useful at a total system level (not subsystem as frequency optimization would require many parameter changes), but including it when evaluating the heat weighting of subsystems I believe, minimizes the amount the trader can stray from their individual time preference. If there is a 30% CAGR subsystem that makes new highs only every 10 years on average and a 25% CAGR subsystem that makes new highs every 6 months on average both may be traded but including draw down time when budgeting heat will move more heat budget to the latter system. Maybe all this is simply part of my “mathturbation” - or maybe this does help me as I believe. Please excuse my imprecision. I really want to show a mathematical proof; however, I don’t know the math to show this accurately and so I try to explain it in my very theoretical and imprecise tinkerer terminology.

Trading Tribe

Of the Trading Tribe ARs you give me the one that resonates is Letting Go of Relationships. I sigh. I feel I have no skills to address this and so no chance at success through any kind of DIM. I don’t quite understand the other 2: Learning a Language – I am definitely fearful about this but also willing and able to pull the trigger. I feel a tightness in my shoulders when I think about it, which is very similar to trades that don’t feel good. I pull the trigger anyway. And I will eventually pull the trigger to learn a language but your feedback on “the next monster trade” is compelling and urgent. Finding the Perfect System – this seems to me like the ultimate, never-ending, always-fun puzzle. I’m looking for “my perfect” – that highest return for the total heat and draw down (and length of draw down) that I’m willing to take and still sleep good at night level. I have relatively few issues trading my systems (small mistakes), I’m having a good time, and it feels like trading could be right livelihood for me.

But even if I don’t agree or don’t understand your suggestions about the language/system Tribe entry points, I know you are right about letting go of relationships. At this point there are no tribes within 3 hrs drive of me (and the one 3hrs away doesn’t appear to be active). I know I have the option to start a Trading Tribe and yet I am hesitant to commit to this. I am skilled at managing people but my experience is very different than tribe; primarily I simply create the right incentives and help people solve specific problems through creative thinking and logic. I have little practice in accepting and understanding extremes in people’s emotions – if anything this is a weakness for me. I feel the chief would need to be a better role model than I could be and if I do go forward with starting a tribe I need to recruit someone better than me in this skill to be the leader.

2017 Results and New Year Commitments

For 2017 I achieve 120% return,17% max draw down. Roughly the return is 80% driven by cryptocurrency (despite this being <20% of the portfolio), 10% futures (mostly seasonal; reg. trend following was ~flat), 10% stocks (this is 2x S&P return on my stock allocation). The year is life changing and I eliminate any reasonable doubt of not having financial freedom. I have realistic expectations for future return and I don’t expect a >100% return again unless there is a melt-up/down in all commodities or cryptos continue their upward trend at the same rate. The 17% draw down does not reveal my daily volatility - many days I am up or down >2%, though the weekly and monthly results are much smoother. Trading aggressive instruments with high heat/instrument but medium total portfolio heat is working well at the portfolio level for both return and for my sleeping at night as I am now used to the daily volatility. I look forward to 2018 and commit to trading with greater accuracy (I make mistakes on a handful of trades in 2017), completing the new sub-systems under consideration that I mention above (ICO, commodity vol, interest rate spreads), and expanding my trading network in the Seattle area to ultimately test willingness of potential Seattle area trading tribe leaders.

Please continue to coach me and give me your feedback.

Thank you Ed!

I am grateful for your help!

Thank you for sharing your process and for raising several issues.

Roughly speaking, we might say:

R <-- P * S

Where R (0-100) stands for your trading Returns, P (0-10) stands for your Psychology and S (0-10) stands for your system.

Now if you already have a ninety-percent (9 out of 10) System and if you have a thirty-percent (3 out of 10) Psychology, you have a 27% Return.

You might notice that if you work really hard on your System you might get it from 9 to 9.5 and get your Return up to 28.5%.

If you work on your Psychology and get it from 30 to 75 you might get your Return to 67.5%.

You might consider taking your feelings about <needing a leader> and <running away from your feelings and into complexity> and <this is no ordinary bubble> to Tribe.

No comments:

Post a Comment