Best Quantitative Trading Blogs

Here’s a selection of my favourite blogs about quantitative trading…

These blogs are great places to find new trading ideas and to learn from other trader’s research and experience. All are run or contributed to by people who are actively involved in the industry, whether at hedge funds, investment banks, or as independent quantitative traders. Hope you find something useful in the list!

https://mathtrading.wordpress.com/

http://alvarezquanttrading.com/blog/

http://www.alphaarchitect.com/blog/

http://epchan.blogspot.co.uk/

https://www.quantnet.com

http://www.quantstart.com/

http://blog.quanttrader.org/

http://blog.quantopian.com/

best-quantitative-trading-blogs

Raising Capital for Start-Up Quantitative Funds

quantitative-hedge-fund

Smaller start-ups are more common in the quantitative fund management arena than in other areas, and many successful modern day funds began with less than $100k in capital. But attracting seed capital and hanging on to it has become harder than ever. This article covers the basics that you’ll need to put in place to ensure that a small startup quant fund is able attract capital investment…

As investor interest diminished following the credit crunch, demand for many of the investment vehicles offered by quantitative hedge funds also fell. Where fund managers before had dictated the terms, often providing little transparency and with conditions not necessarily favorable to the investor, the balance of power has now shifted firmly towards the investor.

Although managers across the whole gamut of strategies were affected by this regime change, few have suffered more than those offering quantitative trading strategies such as statistical arbitrage and high frequency market making, with practitioners of these approaches having been widely held accountable for many of the recent market inefficiencies and declines.

With a dearth of investors and a relatively low level of participation in the current bull market, the environment has become more challenging than ever for those seeking to raise seed capital for new funds, or retain institutional capital in established investment partnerships.

Quantitative fund managers must now be more explicit than ever when explaining their strategies, their source of alpha, their back-office processes, risk management, and differentiation from competitors. As well as returns, the intelligent investors and high net worth individuals that quantitative approaches attract will tend to be equally concerned with volatility, market liquidity, prime brokerage, and general transparency of a fund’s operation.

To help you stay competitive when raising capital for your quantitative hedge fun, we’ve assembled the following list of essentials – if you’ve got these bases covered then prospective investors will be able to understand the unique edge that you are able to offer.

  • Explain how your performance and returns will be replicated into the future, including how they can be scaled with greater assets under management. Because investors will be reluctant to commit based on short term track records, if you don’t have several years of solid returns under your belt you should be prepare to explain precisely how you will ensure consistent performance through a full market cycle. You’ll need detailed quantitative data to show how your strategy navigates corrections and bear markets, and continues to generate alpha throughout periods of unique volatility and economic duress. Given the fraught market conditions of the last decade, this will be one of the main concerns that informed investors will have.
  • Be prepared to demonstrate the effectiveness of your strategy with a full range of industry metrics. Investors in quantitative funds are more sophisticated than ever, and you will be expected to provide all the measures they need to effectively judge your performance. You’ll need Sharpe and Sortino ratios, volatility of returns and maximum drawdowns at your fingertips. For many “fund-of-fund” arrangements, these objective measures tend to be more critical than subjective ones, and if you’re not able to provide the data when it is requested you simply won’t get due consideration.
  • Limit counterparty risk. With so many financial institutions finding themselves on the ropes in recent years, you need to ensure that you have several prime brokers and custodial banks in place. Many investors will now insist upon selecting both prime brokerage and custody providers themselves as a way to mitigate these risks, and it is important that your fund has the flexibility to accommodate these requirements.
  • Ensure that you are operationally equipped to offer separately managed account structures. This option is growing in popularity, and investors will want to see evidence that a separately managed account will perform in tandem with your core strategy. Demand for separate accounts is growing, especially amongst institutional investors.

Over the coming years, as regulation increases and more advanced and efficient algorithms exhaust alpha sources in electronic markets internationally, raising capital for a quantitative fund is only going to get harder.

Following these simple guidelines should help you to raise and retain quantitative hedge fund capital successfully.

 

Neural Networks for Trading

Several neural network researchers have investigated the problem of improving prediction by removing undesirable features or vectors. For example, by using a genetic algorithm to search the space of subsets of the universe of inputs. The fitness criterion used for evaluation is the actual prediction error made. This is an expensive operation since, to evaluate each subset of indicators, training has to be done from scratch.

A different approach to this problem is to remove some training samples (as opposed to features) from the training set. The candidate training samples that are to be removed belong to the ‘malicious’ category of data that will harm out of sample performance.

Another approach is to extract features by making linear combinations of the features in the full feature set. We propose a decision boundary method for doing this. Such a method favors features that discriminate between the classes of interest rather than the fidelity of representation as principal component analysis does. This technique works well if none of the inputs in our universe confuse the predictor.

Several related techniques exist in the field of pattern recognition, all dealing with feature subset selection. All of the existing methods handle the case when the features in the full feature set are all useful to the classification task. The motivation there is to select features until any further increase in prediction is not justified by the added complexity of having an extra feature. However, in our application the problem is quite different. We are typically given a large set of features constituting the full feature set where not all of the features help and some features may surely confuse the predictor.

A further related technique is the well known analysis of variance (ANOVA). This technique computes the sum of squares between classes SS(between) and the sum of squares within each class SS(within). This technique is described for situations where each class is characterized by only a single quantity or feature. The ratio of SS(between) to SS(within) can be used as an alternative criterion for both of the proposed feature selection algorithms. The advantage of using the number of mismatches as the criterion is the ability to handle cases where each class is characterized by more than one feature, as is typical in many prediction tasks.

Principal Characteristics of the Forex Market

To begin, it is useful to examine the general characteristics of the forex market. This will facilitate understanding of the related time series. Most of our discussions will center around the spot forex market as this is the most important segment of the forex market.

The spot market has the following characteristics:

  • Among all the financial markets, the spot market may be regarded as the most technically efficient. This attribute makes it difficult to apply conventional modelling techniques to forecast the dynamic behavior of the forex time series with a view to establishing profitable trading positions.
  • Spot transactions in the forex markets are carried out 24 hours per day using an electronic network such as the dedicated dealing system offered by Reuters. The absence of any formal exchange makes it very easy to trade on the spot market. Transactions on the futures market, by contrast, are made via the exchanges.
  • Ignoring credit related issues, another important reason for the very high trading volume on this market is the possibility of high leverage. This feature attracts even small time speculators and investors to enter and trade currencies. The low barrier to entry to trade forex has been facilitated by the financial companies, brokers, and investment banks all over the world. Broker bonus offers are commonplace, and enable trading with minimal capital.
  • The Forex market is perhaps the only market which is open virtually 24 hours a day, five days a week. The market opens early in New Zealand and Australia, followed by Japan. A little later several trading centers in the Far East such as Singapore and Hong Kong enter the global market and provide further liquidity. As the market closes in Japan, trading begins in the Middle East and Europe, closely followed by the United Kingdom. While the UK is still actively trading, North America begins to trade, and when it is time for the US close the 24 hour cycle starts over in New Zealand. On Saturday and Sunday very limited trading is reported from the Middle East markets.
  • Trading volumes are substantial, particularly in the major currency pairs. Therefore the market is considered very deep and liquid for the purpose of initiating and closing positions at any time of day or night. In general, forex market trades are executed in a fraction of a second, as a counter-party is quickly found by any reputable broker under normal market conditions.
  • Funding costs for overnight forex positions are based on the interest differentials of the two currencies involved in the trading position. These are normally low unless a high interest rate currency is sold against a low interest rate currency. Therefore, in general, it does not cost much in terms of position funding to hold trading positions in the forex market.

You can find a full list of brokerage services for forex trading at this site: Directory World

Brokers for Quant Trading

The variance in trading performance was considerable. A range of
871,000 fSEK was obtained, with a maximum debt of –6,500 fSEK
as the worst result, to a maximum account of 864,500 fSEK as the
best. The mean outcome for the 52 participants was 79,616 fSEK
(SD = 170,029 fSEK) and the median = 14,025 fSEK. Ten
participants (19.2 %) went bankrupt, that is, lost all their fictitious
capital, another 6 participants (11.5 %) lost money but not all of it,
and the remaining 36 participants (69.2 %) were able to gain
money by trading.
Comparing the treatment groups
For the two treatment groups (receiving a lecture), the number of
bankrupt/surviving and losing/winning traders were compared by
Chi-square test, resulting in non-significant differences
(Chi2 = 2.42, p = .12; Yates corr. = 0.68, p = .41, and
Chi2 = 1.01, p = .31; Yates corr. = 0.39, p = .53, respectively).
The group means (Mgroup1 = 97,658 fSEK, Mgroup2 = 24,677 fSEK) did
not either differ significantly (t = 1.37, p = .18). Thus, there was
no significant effect related to the two lecturers and the two
groups receiving a lecture. Thus, they were considered to be
equivalent and therefore collapsed into one group when compared
to the control group in the following computations.
Position size of bankrupt, losing, and winning traders
The participants were not asked about what kind of position-size
strategies they used. Even without knowing if they used any
strategies at all, some conclusions can be drawn by looking at their
series of position sizes. There are several different ways the
traders have taken their positions. Some traders used a constant
fSEK-value position size, regardless of current total capital, while
some others used a constant percentage of current total capital.
Further, there seem to be traders varying the percentage of their
position size, for example, some of them were increasing their
position size after a losing trade, while others were doing the
opposite, that is, increasing the size of their position after a series
of winning trades.
How much of available capital that was put at risk in one separate
trade ranged from minimum allowed position size of 0.5% of
available capital to the maximum possible size of 100%. At level 1,
representing a trading system with expected value of 0.45 (i.e., a
19
gain of 0.45 fSEK per 1 fSEK put at risk), position sizes of 20% or
more were sometimes taken. This was done, even though such a
big position could be the last position taken, if a -5 to 1 losing trade
came up.
The bankrupt traders were apparently taking higher risks. They
were risking 22.9% on an average trade at level 1, while the
surviving traders were risking 6.6%. This difference was significant
(t = 19.3, p < .0001) indicating that the larger the position size,
the greater the risk of going bankrupt. When calculating differences
between the losing and winning traders, it was found that the
former took positions of 15.0% on the average compared to 6.0%
for the latter (t = 16.7, p < .0001).
The tendency described above was similar at level 2 (expected
value of 0.91). Bankrupt traders were risking 23.7% while surviving
traders were risking 3.7% (t = 15.6, p < .0001). Losing and
winning traders were risking 6.5% and 3.8% respectively (t = 4.1,
p < .0001). Accordingly, Hypothesis (1) and Hypothesis (2) were
confirmed.
Position size in the treatment and control groups
Receiving a lecture had an effect on the treatment group to take
smaller position sizes than the control group at both levels,
supporting Hypothesis (3). Thus, position sizes for the treatment
group were 5.5% at level 1 and 3.6% at level 2. For the control
group, the position sizes at level 1 and level 2 were 12.0% and
5.3%, respectively. The differences between the two groups were
significant at both level 1 (t = 14.8, p < .0001) and level 2 (t = 3.0,
p < .001).
All in all, the traders who received a lecture took smaller positions,
than those in the control group, and the traders that took the
smaller positions did not only survive in the simulated market, but
were also able to gain money over the long run. The average
position sizes of the different groups of traders are presented in
Table 5.
20
Table 5. Average position size over groups of traders
Level 1 Level 2
Group of Traders M SD M SD
Winning 6.0%. 7.8% 3.8% 5.2%
Losing 15.0% 14.0% 6.5% 15.1%
Surviving 6.6% 8.0% 3.7% 5.1%
Bankrupt 22.9% 18.5% 23.7% 29.9%
Treatment 5.5% 7.5% 3.6% 6.9%
Control 12.0% 12.3% 5.3% 9.8%
The only variable available for manipulated by the participants
were the size of their position. None of the traders did risk exactly
the same amount, trade by trade, at exactly the same time as
someone else. As a consequence, except for some of the traders
going bankrupt, there were not two traders getting the same
amount of capital. Accordingly, how big or small capital a trader
would get in the end was primarily determined by the size of the
trader’s position.
Profits and losses in the treatment and control groups
The participants in the treatment group lost all their money to a
lesser extent (2 out of 32 = 6.3%) than those in the control group
(8 out of 20 = 40.0%) and thereby confirming the fourth
hypothesis (Chi2 = 9.03, p < .01, Yates corr. = 6.98, p .05; Yates corr. = 1.85, p > .05).
Neither did Hypothesis (5) receive statistical support (t = -1.11, p
> .05) when comparing mean amount of capital gained by trading.
The treatment group (M = 58,888 fSEK, SD = 152,480) did not
gain more as a group than the control group (M = 112,782 fSEK,
SD = 194,382).
21
Effects of gender and prior knowledge of trading
The treatment and control groups did not differ significantly in the
distribution of gender (Chi2 = 3.40, p > .05; Yates corr. = 2.38,
p > .05) and prior experience H(2, N = 52) = .43, p > .05.
There was a main effect of Gender, F(1, 46) = 7.17, p < .05, Prior
knowledge of trading/investing in the stock markets,
F(2, 46) = 8.26, p < .001, and there was an interaction effect of
Gender and Prior knowledge, F(2, 46) = 6.47, p < .005.
Tukey HSD post hoc-test confirmed Hypothesis (7) (p < .0001)
that female traders gained more capital (M = 249,938 fSEK) than
men did (M = 77,372 fSEK). Having prior knowledge of
trading/investing was of an advantage, if one was active trader
(p .05;
Yates corr. = .51, p > .05). There was neither any difference
between the three conditions of prior knowledge of
trading/investing H(2, N = 52) = 1.22, p >.05.
Discussion
One purpose of this study was to find evidence for the importance
of position sizing. The results showed that in order to survive
trading in a simulated stock market, using a trading system with
expected value of < 1.0, one should take positions in sizes of
approximately 3.7% – 6.6% as the surviving traders, rather than
22.9% – 23.7% as the bankrupt traders. Further, to be able to
increase one’s account over the long run and actually make money
by trading the simulated market, one should not risk much more
than 6% as the winning traders did on an average. Accordingly,
deciding how big one’s position of shares should be was of crucial
importance. If the participating traders would lose all their money,
get into debt and not be able to trade anymore, or if they would
gain profits of up to 871,000 fSEK, as the best performing trader
did (an increase of 8,500%), was primarily determined by their
position-sizing strategies, since position size was the only variable
they could affect.
22
Of course, the results were also influenced by chance, since the
outcome of a trade, to win or lose, was determined by randomly
pulling a marble out of a bag. However, all the participants were
trading the same positive-expectancy trading systems. They should
all be able to gain money over the long run, but not everyone did.
The traders participating in the same sessions, did all get the same
trades, winners as well as losers, and no traders, other than some
of those going bankrupt, did get the same results.
Even though this study focused on allocation made by individuals
trading one type of commodities, these findings are in line with
those of Brinson, Singer, and Beebower (1991) where the main
determinant of the differential return of the pension funds was
asset allocation.
Further, was it possible to teach traders to implement less risky
and more profitable position-sizing strategies, so they could survive
in the markets and gain money? Yes, as the results reveal, the
fourth hypothesis was confirmed. The participants that received a
lecture in position-sizing, risk management, and psychological
biases did not lose all of their capital to the same extent as the
control group. All in all, it gave a trader in the treatment group a
tenfold bigger chance of surviving in the stock markets. If the
traders can continue to trade over the long run, there is a greater
chance of getting opportunities of great returns, than if they were
standing by the sidelines. There was a tendency of more traders
being able to trade profitable in the treatment group. Although this
difference was not statistically significant, it is encouraging for
further explorations.
However, the treatment group was not able to produce larger
profits than the control group, when comparing the two groups’
mean results. This outcome may be explained by the fact that the
lecture mainly focused on how to cut losses short and to prioritize
short-term survival first, in order to get long-term gains. Maybe,
the first part of Larry Hite’s basic rules about winning in trading
was not emphasized enough, leading the treatment group to take
too small positions? ”(1) If You don’t bet, you can’t win. (2) If you
lose all your chips, you can’t bet.” (Schwager, 1993, p. 189). Again,
an explanation why the control group gained more as a group is
probably position size. If you bet big you will lose big when you
lose. Evidently, if you bet big you will win big when the draw goes
your way.
23
For future studies and/or education, more emphasis should be
directed toward maximizing gains, “letting the profits run”, for
example, by hands-on training in position-sizing strategies.
Gender was a contributing factor in the results obtained by the
participants. Women did gain more money than men, but they did
not survive significantly better than men when trading a simulated
market. If the findings of Powell and Ansic (1997) that women are
risk averse when deciding in financial matters, was the reason for
this, remains to be investigated.
According to Myagkov and Plott (1997), risk seeking seems to
diminish with experience. This view can be supported by the main
effect of prior knowledge of trading/investing attained in this study.
The active traders performed better than the traders with little or
no experience. The study was carried out in a laboratory setting,
with most participants having little or no prior experience of
trading stocks. This can make generalization difficult and further
research is needed in more realistic settings.
Further, the willingness to take risks is highly dependent of what is
at stake. The only real money the participants could lose was the
remuneration. A more realistic risk-taking behavior would probably
be expressed if the participants were risking their own money
while trading. However, this would, most probably, rise some
ethical as well as practical difficulties.
In order to minimize the risk of anyone tampering with the data
forms, future studies are encouraged to gather the data
electronically, by using computer-generated versions of data forms.
Finally, being able to decrease the risk for a trader of getting ruined
to a tenth, even if demonstrated only in a laboratory setting, is
highly inspiring. With such small means as a three-hour lecture,
only by verbal information on certain, well-known relationships,
there can be more people being able to gain money by trading in
the stock markets, as long as the behavior shown can be
generalized “in vivo”. Further exploration of the importance of
position-sizing is essential. Trading is not an easy game and most of
us need all the support we can get to beat “our enemies”, in order
to make better and more profitable decisions.
The speculator’s chief enemies are always boring from
within. It is inseparable from human nature to hope and
to fear. In speculation when the market goes against
you hope that every day will be the last day – and you
loose more than you should had you not listened to
24
hope – to the same ally that is so potent a successbringer
to empire builders and pioneers, big and little.
And when the market goes your way you become
fearful that the next day will take away your profit, and
you get out – too soon. Fear keeps you from making as
much money as you ought to. . . . Instead of hoping he
(The successful trader) must fear; instead of fearing he
must hope. He must fear that his loss may develop into
a much bigger loss, and hope that his profit may
become a big profit. It is absolutely wrong to gamble in
stocks the way the average man does.” (Lefèvre,
1923/1993, p. 130)

Quant Data Analysis for Brokers

All computations were performed on the data from the forty-fifth
(the fifth last) trade. The two treatment subgroups, each receiving
a three-hour lecture, were compared to each other in order to
17
determine if they differed significantly in means of number of
bankrupt traders, or by means of accumulated trading capital.
Average position-size differences between bankrupt/surviving and
losing/winning traders were computed by use of one-tailed t-tests.
This method of analysis was also used when comparing the position
size of the treatment and control group.
The difference between treatment and control group, regarding
the number of participants going bankrupt and the number of
participants being able to gain money by trading were computed by
chi-square analysis.
Trading in the stock markets is associated with both losses as well
as gains. There are not many traders who have been able to trade
over a longer period of time, without taking any losses. On the
other hand, every now and then, many of them have had
opportunities to make a substantial profit.
An essential requirement to receive a profit from an opportunity is
to be ready to take the chance when it occurs. If one does not
have the money to take it, the opportunity is gone. Therefore, it is
of outmost importance to survive in the short term, so one is able
to stay around for the next opportunity for a good profit.
Consequently, it was of greater theoretical importance to this
study whether a trader lost his/her entire stake or was able to
survive in the market by trading stocks, than the absolute amount
gained. This was the reason for primarily using data at nominal level
and chi-square analyses. Accordingly, a 10,001 fSEK difference
between a bankrupt trader and a winning trader is more important
than a difference between one trader gaining 560,000 fSEK and
another trader gaining 570,001 fSEK.
In the real world, a trader losing all his/her money can not be
compensated by huge gains of another trader. On the contrary, in
the light of one’s own failure, the knowledge of other traders’
success will most probably make the grapes taste even sourer.
Nevertheless, the difference between the two means of total
capital accumulated by the treatment and control group,
respectively, was computed by use of one-tailed t-tests.
Analysis of variance, chi-square analysis and t-test explored the
effect of prior knowledge of trading/investing in the stock markets
and gender.

Statistical Arbitrage Trading

The primary outcome measure was the amount of fictitious
Swedish kronor (fSEK) as scored on the distributed form. This
score was used to determine “survivability” and the number of
traders able to earn money trading over the long run.
A trader who, at some point during simulation, reached a total
capital of zero fSEK or less, that is, lost all of his/her money, was
defined as bankrupt and did not “survive” trading the markets.
Consistently, traders increasing total capital to an amount greater
than the initial 10,000 fSEK were defined as winning traders, able
to trade profitably over the long run. Remaining participants, who
lost money (decreasing total capital to less than 10,000 fSEK) but
not all of it, were defined as losing traders.
There was a lot to gain (a remuneration of 200 SEK) for the traders
meeting the profit objective on level 2, but there was not much to
loose even if they lost all but 1.00 fSEK of their trading capital. As
long as they had just a little fictitious capital left, they were entitled
to keep their 100 SEK remuneration. This could cause traders to
totally abandon their position-sizing strategy, when time was
running out, in order to meet the objective. To have a profit
objective that must be met within a tight timeframe and not really
need to take the consequences if wrong is not a realistic scenario.
In order to reflect a more realistic image of the trader’s positionsizing
strategies, focus was shifted from the outcome of the very
last trade to a trade earlier in the sequence. The trade in focus was
set to the forty-fifth trade at each trader’s final level. A number
close enough to the final trade, yet far enough away from risking to
be strongly influenced by a “make or brake” bet. All the
participants who lost their total capital did so considerably prior to
the forty-fifth trade. They have therefore, most probably, not
unjustly been defined as bankrupt.