1/ If we take a frequentist approach to TF we can understand why our diversified models don't perform well in predictable regimes and why we excel in unpredictable regimes. 2/ We need to understand that we want the opposite of predictive certainty. We are harvesting divergence from predictable assumptions.

Food for thought

1/ Many are concerned with a TF’s use of Realised Balance to assign their bets. They would suggest that equity no matter where it resides is important and that trades should be allocated by risk at all points in time. 2/ Now if dealing with spatial correlation across different time series (markets), then this would be the case. However absolute momentum methods we are harvesting serial correlation across time.

Nonlinearity can be expressed spatially across space and serially along time.

Food for Thought

1/ As traders we like to think that a successful outcome can be attributed to a Managers skill, but what if I was to tell you that an Outlier Hunter’s success is defined by a process that harvests ‘luck’? 2/ You see, an investor never likes to think that their success is defined by ‘luck’. Such derogatory connotations should never be uttered to a Manager with an ego. But to a Trend Follower, we left our egos behind a long time ago.

Food for thought

1/ An Outlier hunter is not attracted to the visual form that a price series takes, as there are many different reasons for how a trending price series can manifest. 2/ There can be random trends that offer no enduring potential, trending moves that arise as part of a larger mean reverting cycle, and non random directional trends of enduring persistence.

An Outlier hunter is seeking the latter form of trend.

1/ A Tail hunter is concerned with only one form of trend. The trend with directional extension.

There are many forms of trend hidden in this series such as random trend components, trends within mean reverting cycles and a primary trend of directional extension. 2/ Unlike many other forms of momentum seeker, to capture these directional trends, we need loose pants to allow for the many other forms of trend that interfere with the trend we are seeking.....the Outlier.

Food for thought

1/ As a tail hunter we are less concerned with market correlations than many other forms of Trend Follower.
This is because the portfolio impact of these non linear anomalies dwarf the impacts of market correlation in our trade distributions. 2/ So for example we could trade two highly correlated markets such as sugar or London Sugar and get two very different trade distributions.

Below is how an ensemble of TF systems used by an Outlier hunter tackles Sugar. We will compare and contracts this with London Sugar.

@rjparkerjr09 @TwitterSpaces @moritzseibert @systematictrend @OMikhailov @AbsoluteGnosis @AlpDogu_ @brunogcampos @mikeharrisNY 1/ @mikeharrisNY has provided a great idea for a discussion next week. As he states, all 'real world' systems tend to be non-ergodic in nature. They inevitable tend towards chaos. @rjparkerjr09 @TwitterSpaces @moritzseibert @systematictrend @OMikhailov @AbsoluteGnosis @AlpDogu_ @brunogcampos @mikeharrisNY 2/ This is because there are no 'things' in this universe. Only processes. When we delve into the nature of 'anything' that we have statistically defined as a 'thing' it quickly vaporises into simply a 'process' that is ephemeral in nature.

@rjparkerjr09 1/ Here is an example to explain how most processes in the real world (including trading) are non-ergodic in nature. What this means is that the time average differs from the ensemble average. Still confused?

I will try to break it down here, @rjparkerjr09 2/ Let's assume we have a trading strategy which has a win rate of 50% and an average win of $50 and an average loss of -$40.

According to our expectancy equation we generate a positive expectancy of (0.5 x $50)- (0.5 x $40) = $5 per trade.

All good. Now let's compound

Let's look at diversification.

1/ It is tempting to conclude that with maximal diversification, a trading process will emulate the price properties of the market (or an index), but while this is pertinent to a 'Buy and Hold' process, it is not pertinent to a Tail Hunter. 2/ A Tail Hunter uses a process of 'exclusion' to only participate in material price moves and avoids trading when price is displaying 'non-trending behaviour'.

What does a Classic Trend Follower mean when they refer to ‘warehousing risk’?

1/ There is a general market principle when it comes to risk. “Risk can never be removed, but it can be transferred”. 2/ In fact, portfolio managers are experts in ‘risk transference’ as we know that a portfolio can hold onto far more risk than what a single return stream can stomach.

1/ Why I don’t trust a backtest – Another Boring Narrative…But with a Touch of Science 2/ Many argue that a backtest is the ultimate verdict for how a model works, but I would like to contend that you can’t trust a backtest. Why?

Underfitting and Overfitting

1/ When we discuss ‘overfitting and underfitting’ with trading models there is not a ‘one size fits all answer’.

The riddle relates to what is referred to as 'Regime Dependent' Vs 'Regime Independent' models. 2/ A Classic TF model responds to exotic Outliers (anomalies) in the market data as opposed to the edge residing in a repeating condition of a historical data series.

We apply a regime independent model, focussed towards market transitions between stable regimes.

@rjparkerjr09 @TopTradersLive 1/ The thing is that we do see the benefits of diversification by trading more markets . We see it in the TTU TF Index.

60xTF Programs (albeit many of them being partial sinners) produce a CAGR/Max DD that continuously converges towards 'the optimal portfolio'. @rjparkerjr09 @TopTradersLive 2/ We all know that correlations of small wins and losses 'offset' in a diversified portfolio to smooth the total portfolio result. This impact quickly dissipates as diversification increases....as there is no further offset available.

More Food for thought on a Lazy Sunday

1/ Most investors view diversification as approximating the following performance growth function.

The model assumes that there comes a point with diversification that marginal benefits decreases to a plateau. 2/ But let's see what happens with a real example of a TF portfolio that diversifies extensively across more markets.

Below are the performance results we are working with in this example.

I have highlighted the columns of interest.

More Food for Thought on System Diversification

1/ Consider the Constituents of the TTU TF Index. In this case we have 60 TF Funds with a >15 Yr Track Record.

Here is how each Program plots w.r.t. CAGR and Max Draw%.

We could consider each Program as a separate system. 2/ Now notice where the Index plots in the CAFR/Max Draw% Distribution Plot. A CAGR of 8.77 and a Max Draw % of 18.04%.

This plot is not the result of the average or median performance of the constituents in the Index. Here are those results.

Food for thought.

1/ An equity curve (return stream) provides an account of your system trading performance when responding to market data.

We can break this return stream into 2 components. That which is a valid signal and that which can be attributed to noise. 2/ By far the majority of the trade results for a trend follower are the results of applying a TF system to a noisy market.

Only 5-10% of our trade results can be attributed to performance relating to Outliers = Signal.

Positive Expectancy is Not Enough in this Game of Investing for the Long Term

1/ Ole Peters in his landmark paper in the Nature magazine “The Ergodicity in Economics" demonstrates a very simple game that is quite counterintuitive in outcome.

nature.com/articles/s4156… 2/ Here we have a practical example of that game using Ole Peter’s example with slight positive expectancy.

The Difference between Continuous signals and Multiple TF Systems.

I have highlighted an Outlier of unknown extent in the graph.

A continuous signal invests between 0-100% in a single possible trajectory.

Multiple unique systems invest across many possible trajectories. Multiple systems squeeze the juice out of the many different possible forms of Outlier.

Continuous systems invest in the precise form of a single trajectory.

1/ Thoughts for a Lazy Sunday

All traders are trend followers (you heard it right)…like it or lump it…but what sets each of us apart is the type of trend we seek to target and whether or not that type of trend holds an edge.

Are you Type A, Type B or Type C? 2/ Type A Trend – A Random Trend
Here is a randomly generated price series where trends clearly exist. The data was drawn from the bulk of the market distribution. This type of trend has no ‘future endurance’.
This is the gamblers trend. In Hurst Terms H=0.5

Analogy for the price action sought with a 'convergent portfolio' . Potential for diversification is limited.

acs.psu.edu/drussell/Demos…

Versus an analogy for the price action sought with a 'divergent' portfolio. Potential for diversification is vast.

upload.wikimedia.org/wikipedia/comm… Analogy 1 is for Convergent methods. Analogy 2 is for divergent methods.

The human brain uses Bayesian reasoning to perfect its predictions. The more confirming data we receive, the more precise our expected outcomes.....but this is the opposite of how we should approach the nature of tail events. We want "loose pants" as opposed to "spandex shorts". While the Kelly Criterion is a perfect outcome for geometric returns under 'certainty'....it is a nightmare waiting to happen under 'uncertainty'. That's why an 'uncertain trader' uses fixed fractional position sizing.