The Quantitative Discretionary Trader Discret Trader - Adam Grimes.pdfThe 800 lb Gorilla in the Room Extensive academic research questions whether it is possible to make excess profits

The Quantitative Discretionary Trader: Art and Science

Adam Grimes

Outline The Nature of the Market

Toward a Critical Approach to Market Problems

The Two Forces

– Mean Reversion

– Momentum

Application

– Applying the patterns

– Tradeoffs in designing trading strategies

– The process of becoming a trader

© 2013 by Adam Grimes, LLC. All rights reserved.

Some Questions How do markets “really move”?

Are there non-random, repeatable patterns in markets?

Is it possible to make money trading?

– We see successful traders. Are their results explainable by chance?


The Market Tools for Analysis The Two Forces Application

The 800 lb Gorilla in the Room

Extensive academic research questions whether it is possible to make excess profits (above index benchmark) from trading.

Opinions are divided, but many researchers believe that markets are highly efficient and that random walks describe price movements well.

– Efficient Markets Hypothesis

– Random Walk Hypothesis.

This is bad news for traders, if true.



The Efficient Markets Hypothesis

A capital market is said to be efficient if it fully and correctly reflects all relevant information in determining security prices. Formally, the market is said to be efficient with respect to some information set, φ, if security prices would be unaffected by revealing that information to all participants. Moreover, efficiency with respect to an information set, φ, implies that it is impossible to make economic profits by trading on the basis of φ.

– Burton Malkiel



Three Forms of EMH Weak form efficiency

– φ: includes only past prices

Semi-strong form efficiency

– φ: includes all publicly-available information

Strong form efficiency

– φ: includes all information, even secret, inside information

(Most researchers find strong form to be fairly strongly contradicted by empirical evidence, and weak form to be strongly supported.)



Poor Challenges to the Theory

Non-normal distributions of returns and prices

– They are non-normal.

Baseline drift

Isolated examples of outperformance

– We can create “talented” coin flippers under the right conditions.

Inefficiencies due to microstructure or transaction costs.

– In analysis, consider impact of stale quotes and illiquid markets.



Credible Challenges to EMH The assumptions are flawed.

– All participants receive information at the same time

– No frictions or financing costs

– Everyone always acts rationally

– People never make mistakes (e.g., ticker confusion)

There are overreactions and underreactions to news.

Inside information

Cost of information

Long-term mispricings: booms, bubbles, and busts



South Sea Company Share price 1719-1721



Credible Challenges to EMH Autocorrelation in returns

– Opinions differ, depending on the study and timeframe.

– Autocorrelated returns could = trending markets

Volatility clustering

Patterns in prices

– (More on this later)



EMH: One Trader’s Perspective Markets are almost always, but not always, efficient.

– Inefficiencies = potential opportunities for profit. – Inefficiencies come in “spurts”.

Price movements are usually random. Behavioral factors strongly influence price movements at

times. Market efficiency may be an evolving process.

The bottom line:

– Price movements are mostly random. – It is really hard to make money in the market, but… – There are patterns in prices. – Behavioral and psychological factors are important.



Randomness We have poor intuition and perception of randomness.

Even well-educated traders suffer from this because it is a fundamental element of human perception.

There are more runs in random data than people expect.

– A trend in prices does not mean it’s non-random.

– The traditional “patterns” of technical analysis appear in random price data.

The feeling of “that can’t be random” often comes from a poor understanding of randomness.

Why do we care? Because if it’s random, you can’t make money with it.



Coin Flips: Which Group is Random?

Group 1

H, H, H, H, T, T, T, H, T, T

T, H, T, T, H, T, T, H, H, H

H, H, T, H, T, H, H, H, H, H

Group 2

H, T, T, H, T, H, T, H, H, T

T, H, T, H, T, T, T, H, T, H

H, T, H, H, T, T, H, T, H, H



Runs Tests for Random Data

Group 1

H, H, H, H, T, T, T, H, T, T

T, H, T, T, H, T, T, H, H, H

H, H, T, H, T, H, H, H, H, H

4/2/3 [5 heads]

2/2/3 [5 heads]

2/5 [8 heads]

Group 2

H, T, T, H, T, H, T, H, H, T

T, H, T, H, T, T, T, H, T, H

H, T, H, H, T, T, H, T, H, H

2/2 [5 heads]

3 [4 heads]

2/2/2 [6 heads]



Runs Tests for Random Data

Group 1

H, H, H, H, T, T, T, H, T, T

T, H, T, T, H, T, T, H, H, H

H, H, T, H, T, H, H, H, H, H

Group 2

H, T, T, H, T, H, T, H, H, T

T, H, T, H, T, T, T, H, T, H

H, T, H, H, T, T, H, T, H, H



Truly random (RNG using noise from nuclear decay.)

Created by a human (professional statistician) in a best effort to appear random

Random Walk The classical “Drunkard’s Walk”

– Intoxicated man takes an (equal sized) step right or left

– Equal probability of either right or left.

– No memory of past steps

– The path he walks is a type of a random walk

– (Disciplined drunkard? Talk about a model not fitting reality!)

Can also generate from a coin flip

Pt = Pt-1 + x, x = -1 or +1 with 50% probability



Binomial Tree (Lattice)



Random Price Series 1



Random Price Series 2



Traditional Technical Analysis:

Support and Resistance Levels

A price area which may offer sufficient supply (resistance) or demand (support) to stop prices.

– Always think potential support or resistance…

Usually zones, not exact levels.

– If drawing, think crayons, not an architect’s pen.

Many ways to find these levels:

– Visible chart points

– Calculated levels (Gann, Fibonacci, Murrey Math, etc.)

– Opening range

– Volume analysis



Support and Resistance in Action

These levels were calculated for Bidu, Inc. (Nasdaq: BIDU) at the end of 2010 based on the daily chart.

Consider intraday engagements of these levels with the lessons of traditional technical analysis in mind:

– Broad zones, not exact prices.

– Support → Resistance and vice versa

– Gap through levels can be significant

– The more times a level is tested, the more valid it is.

With calculated levels:

– If a level fails to hold, price usually moves to the next level.



The Levels on the Daily Chart



Intraday Levels



Intraday Retests



Moving Through Levels



…But Stopped at the Next Level



Equally Useful on All Timeframes



The “Math” On December 28, 2010 I generated these levels publicly

– http://youtu.be/5mPp7_fajvo

I drew random horizontal lines on the chart!

– These levels are completely random.

– (There is no “file drawer” problem. This was only done for BIDU.)

If you are using levels, are they better than this?

– Are you sure?

The lesson:

– Any line(s) on a chart will appear to be significant.

– You cannot trust your perception.



Tools for Analysis Financial markets generate a tremendous amount of

data, most of it numerical.

It is tempting to apply analytical tools from other disciplines to market data, but be careful.

– Physical analogues (Newton’s laws, etc.)

– Tools from digital signal processing (radar, audio, etc.)

– “Quasi-mystical” approaches (ratios, astrology, etc.)

Behavioral and psychological factors are important

– The social sciences have shown us that crowd behavior is often amenable to statistical analysis.

The market is not a math problem, but it looks like one.



Analyzing Market Data The goal is to be able to distinguish non-random

price action from the noise. – “Hey, that’s different.”

– Significance testing may be flawed.

Also need to assess economic significance.

Consider risk and volatility. – Sometimes it might be preferable to make less money with

more certainty.

Markets are very noisy and have a high degree of randomness – Simple and robust mathematical tools are often better.



Simple and Robust Tools Basic statistics

– Ways to summarize a data set

Measures of central tendency

Measures of dispersion

Simple statistical tests – Linear regression

– Hypothesis tests (can be misleading)

– Checks for stability (e.g., stationarity)

Test procedures – Exploratory data analysis

– “Bin” analysis

– Event study methodologies



General Advice Do some work by hand. There is value in having close

contact with the data. Conversely, make sure you also look at large amounts of

data with appropriate tools. Avoid data snooping bias. Don’t “shotgun test” many

patterns. Have an idea. Understand why something should work. (Have an idea.)

– Every good market idea has a reason why it should work. – It might not be obvious, but it is there.

If something is too good to be true, you made a mistake. Consider economic significance in addition to statistical

significance. Above all…



Be careful of the easy answer.

Make sure you understand your tools deeply.



What is the Correlation?


A. Closer to 1 B. Closer to 0 C. Closer to -1 D. Can’t tell


What is the Correlation?


A. Closer to 1 B. Closer to 0 C. Closer to -1 D. Can’t tell


Correlation = -1.0

How?

Correlation of Daily Returns The process was:

– Day 1: Series A: +1.0%, Series B -0.5%

– Day 2: Series A: -0.5%, Series B + 1.0%

Over any period of multiple days, both series advance, but they do so in perfect inverse daily correlation.



Correlation = +1.0



Tools of Traditional Technical Analysis:

Fibonacci Ratios Ratios are derived from the “Fibonacci sequence”

Fn = Fn-1 + Fn-2

Whether or not a number is in “the sequence” is meaningless.

– Can generate the ratios using any seed numbers

– There is no value to using “Fibonacci numbers” e.g., as inputs for indicators or bar lengths.

Superstition and mathematical illiteracy?

The ratios between numbers in the sequence are important in nature, art, music, etc.

~61.8%, ~38.2%, ~23.6%

– Fibonacci traders also use 50%. Why? (It is not a Fib ratio.)



Testing Fibonacci Ratios Ralph N. Elliott reasoned that if these ratios are so

important in nature (they appear to be), they should also be important in financial markets.

Are swings more likely to terminate at or near Fibonacci ratios?

How to test?

1. Create a system for defining swings in markets.

2. Collect a large number of swings from different markets, asset classes and timeframes.

3. See if they tend to terminate near Fibonacci ratios.



Defining Swings Many ways to do this, but these tests used a tool I

developed called AlgoSwings®

– Uses both price structure and volatility to define swings

Example, to define an upswing, watch for the reversal to a downswing.

– Highest point of the upswing is the “pivot high”.

– Basic swing trigger is a reversal that violates the N day low.

– Also require a swing to reverse at least X ATR’s (average true ranges) from the pivot high.

– All of these inputs can be modified.



Algo Swings with Fibs



Algo Swings with Measured Swings



A

B

C

D

Retracement Summary Stats


Added a simple trend filter

– Removed all retracements which were > 100% of the “setup” leg.

Other definitions of trend were possible (and the work was also done with no filter), but this is the simplest.


Testing the Ratios If Fibonacci Ratios accurately describe retracements

in markets, we should see a distribution of retracement measurements that cluster near the ratios.



Retracement Ratios

with Trend Filter



Retracement Ratios

with Trend Filter



Going Further:

Only the Best Trades Perhaps ratios are more significant in “good” trades.

– Length of extension > 25% of setup leg

Assures adequate profit potential

– For a long trade, the low of the retracement must be higher than the low of the setup leg.

– The extension must make a new high relative to the setup leg.



Retracement Ratios

for Qualified Extensions



Retracement Ratios

for Qualified Extensions



A Related Question: Do shallower retracements lead to stronger

extensions?

Categorize extensions by the size of the preceding retracement.

If this is true, then shallower retracements should have larger extensions.



Fib Summary:

Fibonacci Fibs?


We see no evidence that Fibonacci Ratios describe important structural points in the market, over a large and diverse sample.

The received wisdom that smaller retracements lead to larger extensions is also strongly contradicted by the data.

Filtering for best trades or volatility did not do much to affect results. (Not all tests are presented here.)


Fibonacci Lessons It is likely a serious mistake to use Fibonacci ratios as

entry points, stops, or profit targets.

This challenges many beliefs and much of the literature.

– Much of the literature is based on superstition and quasi-mystical appeals to authority. (E.g., the claim that the height of the Giza pyramid, measured in inches, is a Fibonacci number.)

These tests may have missed something. This is not conclusive disproof, (but the burden of proof should lie with those selling Fibonacci-based systems/products.)

Look for retracements to terminate around 50% of the extension, with a large margin of error.



The Tools of Traditional Technical

Analysis: Moving Averages Many different types of moving averages.

Traders use moving averages in many ways

– As support or resistance

– As a trend indicator

Questions

– Is there non-random price action around moving averages?

– Are there “special” moving averages?

50 day, 200 day, etc?

– Do moving average based trend indicators “work”?



Moving Average as Support



Moving Average Value Changes

As the Bar is Formed



Moving Moving Averages The value of the average changes as the bar is

formed.

More significant for shorter period moving averages and for front-weighted averages.

May or may not be a problem, but some of the best tests of moving averages were not tradable in real time if buying on the average.

Will not affect failed tests, so this could skew results.

A possible solution is to use the previous bar’s MA.



Event Study Methodology Define test universe.

– Represent all important asset classes

600 stocks, by market cap

16 liquid domestic futures markets

6 major currencies against the USD

Random walk simulations

– Consider historical period (1/1/2001 – 12/31/2010)

– Address several volatility regimes.

– Adjust for baseline drift.

Compute summary stats for each asset class.



Event Study (cont.) Define a precise condition that will result in a trading

signal.

– Symmetrical for buy and sell.

Work through each bar of the universe looking for the condition.

Record returns for each bar following the condition.

Combine all N+1, N+2, … returns to get a composite for each bar following the signal.

Create excess return measure (signal – baseline) for each bar.



Event Study Summary Output



Moving Average Touch and Hold Tests



Moving Average Fade Break Test



MA Fade Break, 50 SMA



MA Fade Break, 200 SMA



MA Fade Break Results Something interesting is going on here. Evidence that asset classes do not “trade the same”. In Equities:

– Statistically significant sell signal – Buy signal that decays into a sell signal – Effect sizes are not small – Evidence that contradicts random walk!

Results (not shown here) look basically the same for different periods or types of averages.

Results in Equities are also strong using random period moving average. – Wait. What?



Thinking About MA Breaks There seem to be no “special” moving averages.

If all moving averages seem to work the same way, why use them as support or resistance?

There is an interesting pattern in the break test, but what if we run the test without the moving average?

– For a buy, price has to close below the MA, and the previous bar’s close had to be above the MA.

– This almost always means the entry bar’s close is below the previous low.



Fade Close Outside Previous Bar’s Range



Moving Averages, Explained? Particularly in Equities, the effect of any moving

average is wholly explained by normal mean reversion.

Will see same effects with any or no moving average.

Be very careful about how you use moving averages

– Support and resistance? No.

– As actual levels to enter and exit trades? Maybe.

– But what about the slope of a moving average?



Slope of 50 SMA as Trend Indicator



Uses of Trend Indicators Traditionally, we are told to buy and to avoid shorts when

a trend indicator is up.

Many trend indicators allow a neutral condition as well.

Often, we are told to use higher timeframe trend condition as filter.

– This is smart marketing. (It’s a little harder to test.)

Test procedure:

– Categorize returns by the previous bar’s trend condition.

This is critical. Otherwise, you assume information you would not have had at the time.

Many mistakes are made here!



Slope of 50 MA, Categorical Returns



Triple Moving Average Trend Indicator

Apply multiple moving averages to a chart.

Only trade when they are in the “right” order. I.e., only buy when the faster period moving averages are higher than the slower period.

Can also use slopes of multiple moving averages.

A common variant uses 10, 20, and 50 SMAs.



Triple MA Trend Indicator, Categorical Returns



Moving Average Conclusions There are no special moving averages.

Moving averages do not provide support and resistance.

The slope of a moving average is not a reliable indicator of trend.

Trend indicators derived from the position of multiple moving averages are not reliable indicators of trend.



A Useful Construct: The Two Forces Price movements are shaped by

– Mean reversion Large movements are reversed

– Momentum (range expansion) Large movements lead to more movement in same direction

Over a large sample of market action, these two forces balance each other out. – Random walks describe price movements well when the forces

balance.

Is it possible to identify conditions that will show when one force is likely to predominate? – This is a quantitative way to address the trading problem. – (The answer is yes.)



Mean Reversion Fading large single bars

– Volatility-adjusted measures are best

Fading N-day runs

Fading breakouts of N-day highs or lows – Donchian channels

Fading large excursions from moving averages – Keltner channels or Bollinger bands

– Perhaps we should be suspicious of precise values?

Different asset classes and timeframes have different tendencies for mean reversion!



Fading +/- 3σ Closes Close in top 75% of range, and Open in Bottom 25% (for sells).



Crude Oil with 20 Day Channels



Trading 100 Day Channels (at least 5 days between entries)



Stocks at 52 Week Highs Many fundamental systems include some token

technical criteria.

A common one is to only buy stocks that are at or near 52 week highs.

The logic is that a stock going up is supported by buying pressure, and that the market has already voted.

This is very logical, but, unfortunately, also wrong.

Buying stocks at 52 week highs puts you on the wrong side of one of the strongest statistical tendencies in the market: mean reversion.



Trading 260 Day (~52 week) Channels (at least 5 days between entries)



Indicators: Standard RSI-14 (70/30)



RSI-14 Test



Momentum Volatility compression

Pullbacks

These tendencies can be captured in some traditional chart patterns.



10 and 50 Day Historical Vols



Volatility Compression Test Most breakouts fail (mean reversion).

Can we qualify breakouts with a volatility condition?

Ratio of 5 to 40 day ATR is < 0.5.

Current day’s range is >= 5 day ATR.

Current day closes in top 50% of range.

These are crude criteria, but good for general test.

(Volatility evolves differently in forex, so sample size too small.)



Volatility Compression Test



Pullback Test Buys are allowed after the market closes above the

upper channel.

Entry trigger is a touch of the previous day’s 20 period XMA.

Only one entry allowed per touch of the channel. Condition is reset after entry.

Rules are symmetrical to the sell side.



Keltner Pullback Entry



Keltner Pullback Entry Test



The Two Forces: Summary Mean reversion and momentum would balance each

other out in random walk markets.

Certain trigger conditions can tilt the scales in favor of one condition over the other.

These are simple but robust tests.

– Avoid “full system” tests for tendencies. (These are common but potentially misleading.)

Understand the balance of these forces in your chosen market and as they apply to your trading patterns.



Now what? There is more to successful trading than knowing

what to do.

Successful trading comes from the successful application of correct principles of price behavior.

If we know the right thing to do, why is it so hard?

– “It is not so much that the market is against us; it is that the market sets us against ourselves.” (TAAS, 347)

Most people are not prepared for the extreme variability of trading returns.



Human Performance Varies Babe Ruth’s Hit Rate, 1951-1968



Example: Hypothetical Trading System

Wins 50% of the time. Wins are always 1.2R

Loses 50% of the time. Losses are always 1.0R

What is the expected value of this system?

(50% * 1.2) – (50% * 1.0) = 0.1

(What does this mean?)

Assume we trade 250 trades from this system, risking $2,000 per trade with a starting equity of $100,000.

“Should” end up with $150,000

(250 * $2,000 * 0.10) = $50,000 profit

© 2013 by Adam Grimes, LLC. All rights reserved. The Market Tools for Analysis The Two Forces Application

Results Are Not Always What You Expect

© 2013 by Adam Grimes, LLC. All rights reserved. The Market Tools for Analysis The Two Forces Application

Learning to Trade: Two Mistakes

The Quest for the Holy Grail

Always looking for a better indicator, system, or tool.

Reliance on gurus.

Paralysis by analysis.

Easy to blame losses on imperfect tools.

“Tell me about your Mother…”

The idea that all your answers are to be found in psychology.

“Anything works” you “just have to find what works for you.”

Places the blame for losses on imperfect psychology.



Building a Solid Foundation Both answers are right, but both are wrong without the other.

– Must have a system aligned with market reality. – Must be able to apply it.

Solid principles can be applied anywhere in the spectrum of discretionary → systematic trading. – The total gut-feel trader.

Probably not many of these who work successfully outside of a few specific environments.

– The nearly completely systematic trader If you are not trading 100%, purely systematically, you are partially

discretionary.

Discipline is key – The market has evolved because of traders’ mistakes. – The difficult thing is often the right thing.



Two “Styles” of Trading With-Trend

– Seeks to enter a position in alignment with the preexisting trend, or at the beginning of a new trend.

– Common structures are pullbacks and breakouts. – Ideal entries are often around “centers”.

Counter Trend

– Looks to take positions against the current dominant trend on the trading timeframe.

– Is there any true countertrend trading? – Hope is to be able to capitalize on a shorter trend against the

main trend. – Trades often come around extremes.



Comparison of the Two Styles

With-trend

Trades can often be held a long time. – “Ride your winners”.

Winners tend to be larger than losers.

May reward a more passive mindset.

Counter-trend

Trades tend to have limited expectation. – Important to take profits

proactively.

Losers are often bigger than winners.

Often require aggressive (decisive) psychology. – Immediate gratification

More trades



If the Only Tool You Have Is a Hammer…


Trading Mean Reversion The concept is to trade large moves.

Markets “probe” areas looking for stops (attempting to generate volume.)

Some traders simply fade large moves.

– May quantify many different ways.

– Requires scaling in.

– Invites emotional reactions.

– May work for years then blow up and lose everything.

The Failure Test is a good alternative.



The Failure Test Price probes beyond a level, finds no conviction, and

quickly reverses.

– Can occur in a trending market or a trading range

– Consider relationship to higher timeframe trend.

Objective entry pattern with a clearly defined risk point

Other names:

– Spring or Upthrust (Wyckoff)

– 2B (Victor Sperandeo)

– Bull or Bear Traps



The Failure Test:

Entry Trigger



The Failure Test: Trade Location



A Third Possibility



Failure Test:

Stop Location



EURUSD



S&P 500



Gold Futures, 2011



Soybeans



Be Careful of this Environment



A Sample Intraday Failure Test Entry

Adapt Failure Test Pattern, but use HOD and LOD as the reference levels. – Could do this with awareness of important levels on daily chart.

– Could also consider the Taylor Rhythm.

Do not trade in first half hour or last hour of the day.

5 min S&P Futures bars – Keltner Channels (2.25 ATRs +/- 20 XMA)

Previous HOD/LOD must have been outside Keltner Channels.

Must be at least 3 bars between entry bar and previous HOD/LOD. (Avoid “sliding” trends.)

Must have > .75 point failure back inside HOD/LOD.

Wait for the close of the bar.



4/11/13



4/12/13



4/15/13



4/16/13



4/17/13



Trading Momentum:

The Pullback Trends tend to move in alternating patterns of with-

trend strength and retracement.

The pullback captures this fundamental element of price behavior.

Uses market structure to limit and define risk.

A pullback is a shorter-term countertrend move in a larger trend.



Many Names for One Idea



Two Important Variations



Pullbacks



Pullbacks: How to Enter Trades


Broadly speaking, two plans (for long trades):

A. Buy failure tests near the bottom of pullbacks

B. Buy breakouts after a pullback

These two plans offer a tradeoff between trade location and confirmation.


Stop Placement: Tradeoffs



Consider Profit Targets



Pullback: Sample Trading Plan Using Keltner Channels and Modified MACD: A long setup occurs after sharp momentum:

– Price goes through the upper Keltner Channel – MACD makes a significant new high

A long entry occurs: – After price pulls back near the EMA – On the bar that “hooks” the fast line of the MACD in the direction of the trend.

Do not trade after a climax. Initial stop placement:

– At the point at which the pullback is violated – Never moved in the direction of risk

Profit targets – Take 1/3 of the position off at 1X and 2X risk. – Watch for failure around previous trend extreme.

(Examples show 10 minute S&P 500 E-mini futures, 24 hour data.)



2/20/13



2/21/13



2/22/13



2/25/13



2/26/13



2/27/13



This Is Only a Beginning Other patterns:

– Breakouts

– The Anti

– Other Pullback entries

– Pattern Failures

Using multiple timeframes to refine entries and manage risk.

Position sizing

Trade management

Understanding trading psychology

Monitoring performance

Ongoing research

Developing the skills of successful trading






Author’s blog: http://adamhgrimes.com


Waverly Advisors, LLC:

Research Products Tactical Playbook

– Written for the active trader on the daily/weekly timeframes – Exact trade recommendations

Hybrid systematic-discretionary methodology – In-depth technical “drill down” into a set of markets. – Bigger-picture overview of all liquid asset classes.

Tactical Portfolio Outlook – Written for the longer-term manager

Addresses both the allocator and the longer-term active trader. – Emphasis on executing with ETFs in a long-only and long-short environment – Focus on Equities, Equity Sectors, and other asset classes – Macro perspective on risk factors and major economic events.

Options Market Outlook – Proprietary, quantitative analysis of options market – Incorporates both volatility and directional analysis – Macro risk factors and cross-asset perspective – Actionable trade ideas



Contact Information Adam Grimes

email: [email protected]

blog: http://adamhgrimes.com

Waverly Advisors

website: http://waverlyadvisors.com

phone: (607) 684-5300

Chris Noye, Managing Director, Head of Sales

[email protected]


Links Research samples

– Tactical Playbook http://bit.ly/1bsUwlI

– Tactical Portfolio Outlook http://bit.ly/1cmENFV

– Options Market Outlook http://bit.ly/GYkQef


http://bit.ly/1bsUwlI

http://bit.ly/1cmENFV

http://bit.ly/GYkQef


Documents

The Quantitative Discretionary Trader Discret Trader - Adam Grimes.pdfThe 800 lb Gorilla in the Room Extensive academic research questions whether it is possible to make excess profits