What infrastructure do you need?
- back testing
- paper trading
- real time evaluation
- promotion scaling
What are the best websites?
- blue owl
- see ch. 2 section
How much math do you need?
- undergrad stats, financial, algorithmic
Who should you talk to?
- Peter Gardner
What is your strategy? (the simplest)
- day trading scalping
- stat arb
- mean revert
- stat infer
- etf correllation/surrogate
- seasonal sectors
Is 1%/day realistic – how do you find out?
- go on day trade forums
- ask around
What chapters can you omit (2&6 look key)
What is the Sharpe ratio?
- a measrure of the profiablity of a strategy compared to a benchmark
Does your DMA broker provide:
- real time data
- clean historical data
- fundamental (company financial) data
What are limit orders? Need to re-read pp 23
What area of math deals with time series data?
What exactly is statistical arbitrage?
- exploiting a price discrepancy, eg one stock with different prices on different
= Action points
Find your day trading strategy.
Your backtesting & execution platform should be identical. Use excel / matlab /
google docs for basic analysis, initial hypothesis testing.
Check out TradeStation brokerage (they offer backtesting & order submission)
Re-read ch. 4 when you can actaully understand what the financial terms mean.
Thoroughly investigate stop loss
= Ch. 1
statistical arbitrage trading deals with the simplest finan-
cial instruments: stocks, futures, and sometimes currencies.
Author has a physics Phd., worked at IBM TJ Watson research lab & many top tier
financial instutions. When he tried complex strategies they all failed!
Use the simpliest strategy possible
It is important not to have a need for immediate profits to sustain your
daily living, as strategies have intrinsic rates of returns that cannot
You should hope to have steadily increasing profits, but most likely it won’t be
200 percent a year
Some fund managers spend 1 day a month monitoring reports & news, setting up the
Author spends 2 hours in the morning reading reports & preparind days trading.
> I’m looking for a 100% automated system
you will need to spend time doing research and backtesting on new strategies.
The Nonnecessity of Marketing
- the business of quantitative trading
- allows you to focus exclusively on your product (the strategy and
- the software), and not on anything that has to do with influencing
- other people’s perception of yourself.
- > This is HUGE!
How do you find the right
strategy to trade? How do you recognize a good versus a bad strat-
egy even before devoting any time to backtesting them? How do you
rigorously backtest them? If the backtest performance is good, what
steps do you need to take to implement the strategy, in terms of both
the business structure and the technological infrastructure?
= ch. 2 Fishing for Ideas
Many academic strategies are too complex, out of date or expensive to test.
many traders’ forums or blogs may suggest simpler strategies that are equally profitable
strategies from traders’ forums may have worked only for a little while, or they
work for only a certain class of stocks, or they work only if you
don’t factor in transaction costs.
the difficulty is not the lack of ideas. The difficulty is to de-
velop a taste for which strategy is suitable for your personal circum-
stances and goals, and which ones look viable even before you de-
vote the time to diligently backtest them
Low capital a/c’s (YOU)
With a low-capital account, we need to find strategies that can
utilize the maximum leverage available. (Of course, getting a higher
leverage is beneficial only if you have a consistently profitable
> Leverage = risk
Capital availability also imposes a number of indirect con-
straints. It affects how much you can spend on various infrastructure, data, &
Author uses Interactive Brokers
Backtesting data with surviourship bias may be acceptable to test Intra day
> Intra day scalping seems to be the strategy for me
more regularly you want to realize profits and generate income, the
shorter your holding period should be.
Sharpe ration is:
average exces Returns / std dev excess
where excess = portfolio – benchmark
or how consistently does the portfolio perform
Bollinger bands: every time the price exceeds plus or minus
2 moving standard deviations of its moving average, short or buy, re-
What are your transaction costs (see IB api)
A historical database of stock prices that does not include stocks
that have disappeared due to bankruptcies, delistings, mergers, or
acquisitions suffer from the so-called survivorship bias, because
only “survivors” of those often unpleasant events remain in the
This is especially true if the strategy has a “value” bent; that is,
it tends to buy stocks that are cheap.
it is very likely that you can optimize parameters in such a way that
the historical performance will look fantastic. It is also very likely
that the future performance of this strategy will look nothing like
its historical performance and will turn out to be very poor.
Every time a carefully constructed model that seems to work marvels
in backtest came up, they inevitably performed miserably going forward. The
main reason for this seems to be that the amount of statistically independent
financial data is far more limited compared to the billions of independent con-
sumer and credit transactions available
The AI strategies that work for me are usually characterized by these properties:
- based on a sound econometric or rational basis, and not on random discovery of patterns.
- few parameters that need to be fitted to past data.
- linear regression only, and not fitting to some esoteric non-linear functions.
- conceptually simple.
- All optimizations must occur in a lookback moving window, involving no future
- unseen data. And the effect of this optimization must be continuously demonstrated using this future, unseen data.
look for those strategies that fly under the radar of most institutional investors
- have very low capacities because they trade too often,
- trade very few stocks every day,
- have very infrequent positions (such as some seasonal trades in commodity
- futures described in Chapter 7).
- Those niches are the ones that are likely still to be profitable because
- they have not yet been completely arbitraged away by the gigantic
- hedge funds.
Parametersless trading: all parameters are dynamically optimized in a moving
Benchmark model on a test & training set.
- eg IS: 1985 – 2005, OOS: 2006-2011
Backtesting is about conducting a realistic historical simulation of
the performance of a strategy. The hope is that the future perfor-
mance of the strategy will resemble its past performance
Data: Split/dividend adjustments, noise in daily high/low, and
Performance measurement: Annualized Sharpe ratio and maxi-
Look-ahead bias: Using unobtainable future information for past
Data-snooping bias: Using too many parameters to fit historical
data, and avoiding it using large enough sample, out-of-sample
testing, and sensitivity analysis.
Transaction cost: Impact of transaction costs on performance.
Strategy refinement: Common ways to make small variations on
the strategy to optimize performance
= ch. 4 Setting up your business
Minimising transactions costs
To cut down on commissions, you can refrain from trading low-
price stocks. Typically, institutional traders do not trade any stocks
with prices lower than $5. Not only do low-price stocks increase
your total commissions costs (since you need to buy or sell more
shares for a fixed amount of capital), percentage-wise they also have
a wider bid-ask spread and therefore increase your total liquidity
In order to minimize market impact cost, you should limit the
size (number of shares) of your orders based on the liquidity of the
stock. One common measure of liquidity is the average daily volume
(it is your choice what lookback period you want to average over).
As a rule of thumb, each order should not exceed 1 percent of the
average daily volume.
Often, the moment you start paper trading you will realize that
there is a glaring look-ahead bias in your strategy—there may just be
no way you could have obtained some crucial piece of data before
you enter an order! If this happens, it is “back to the drawing board.”
You should be able run your ATS, execute paper trades, and then
compare the paper trades and profit and loss (P&L) with the theoret-
ical ones generated by your backtest program using the latest data.
If the difference dais not due to transaction costs (including an ex-
pected delay in execution for the paper trades), then your software
likely has bugs.
If you are able to run a paper trading system for a month or
longer, you may even be able to discover data-snooping bias, since
paper trading is a true out-of-sample test.
traders usually pay less and less attention to the performance of a paper trading
system as time goes on, since there are always more pressing issues
(such as the real trading programs that are being run). This inatten-
tion causes the paper trading system to perform poorly because of
neglect and errors in operation. So data-snooping bias can usually
be discovered only when you have actually started trading the sys-
tem with a small amount of capital.
Regime shifts refer to the situation when the financial market
structure or the macroeconomic environment undergoes a drastic
change so much so that trading strategies that were profitable be-
fore may not be profitable now.
eg decimalization of stock prices. Prior to early
2001, stock prices in the United States were quoted in multiples of
one-sixteenth and one-eighteenth of a penny. Since April 9, 2001, all
U.S. stocks have been quoted in decimals.
= ch. 6 Money & Risk management
Every optimization problem begins with an objective.
Kelly formula (optimal leverage for a trading strategy)
true (geometric) random walk, by which
I mean there is a 50–50 chance that the stock is going up 1 percent or
down 1 percent every minute.
you will lose money, at the rate of 0.005 percent (or
0.5 basis point) every minute! This is because for a geometric random walk,
the average compounded rate of return is not the short-term (or one-period)
return m (0 here), but is g = m − s 2 /2.
it is extremely helpful to
have a collaborator or consultant to duplicate your backtest results
independently to ensure their validity. This need to duplicate results
is routinely done in scientific research and is no less essential in
“representativeness bias”—people tend to put too much weight on
recent experience and underweight long-term average
We must remember that we are operating in a probabilistic
regime: No system can avoid all the market vagaries that can result
start with a small portfolio and gradually gain psychological preparedness,
discipline, and confidence in your models.
In order to proceed slowly and cautiously, it
is helpful to have other sources of income or other businesses to
help sustain yourself either financially or emotionally (to avoid the
boredom associated with slow progress). It is indeed possible that
finding a diversion, whether income producing or not, may actually
help improve the long-term growth of your wealth.
Position entry signals: decide when to buy
Exit strategy: when to sell
- A fixed holding period
- A target price or profit cap
- The latest entry signals
- A stop price
- One of the most intuitive commodity seasonal trades is the
gasoline future trade: Simply buy the gasoline future contract that
expires in May near the middle of April, and sell it by the end
of April. This trade has been profitable for the last 11 years, as
of this writing in April 2008. (See the sidebar for details.) It ap-
pears that one can always depend on approaching summer driv-
ing seasons in North America to drive up gasoline futures prices in
> This is an example of the kind of stategy you should employ:
- based on economic reality
- easy to understand
HIGH-FREQUENCY TRADING STRATEGIES
The reason why these strategies have Sharpe ratio is simple:
Based on the “law of large numbers,” the more bets you can place,
the smaller the percent deviation from the mean return you will
aim to exploit tiny inefficiencies in the market
inefficiencies and need for liquidity persist
day to day, allowing consistent daily profits to be made. Further-
more, high-frequency strategies typically trade securities in mod-
est sizes. Without large positions to unwind, risk management for
high-frequency portfolios is fairly easy: “Deleveraging” can be done
very quickly in the face of losses, and certainly one can stop trad-
ing and be completely in cash when the going gets truly rough. The
worst that can happen as these strategies become more popular is a
slow death as a result of gradually diminishing returns. Sudden dras-
tic losses are not likely, nor are contagious losses across multiple
Though successful high-frequency strategies have such numer-
ous merits, it is not easy to backtest such strategies when the
average holding period decreases to minutes or even seconds.
Transaction costs are of paramount importance in testing such
strategies. Without incorporating transactions, the simplest strate-
gies may seem to work at high frequencies. As a consequence, just
having high-frequency data with last prices is not sufficient—data
with bid, ask, and last quotes is needed to find out the profitability
of executing on the bid versus the ask. Sometimes, we may even
need historical order book information for backtesting. Quite often,
the only true test for such strategies is to run it in real-time unless
one has an extremely sophisticated simulator.
High-speed execution may account for a large part of the actual profits or losses.
(To recap: Capacity is the amount of equity a strat-
egy can generate good returns on.) It is far, far easier to generate
a high Sharpe ratio trading a $100,000 account than a $100 million
account. There are many simple and profitable strategies that can
work at the low capacity end that would be totally unsuitable to
hedge funds. This is the niche for independent traders