presentations/abstract.txt at master · algoquant/presentations · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
### EARL Boston 2015 ideas

rationale


{\scriptsize{"The premier market anomaly is momentum. Stocks with low returns over the past year tend to have low returns for the next few months, and stocks with high past returns tend to have high future returns."}}\\

the question then is:
	can momentum also work on shorter time scales?
	do higher moments (skew, kurtosis) also have predictive power?
	but higher moments are hard to estimate from low frequency (daily) returns.
	higher moments decay with time (term)
	many packages aren't optimized for speed
	how to use high frequency data to estimate higher moments

estimating moments
	Rogers-Satchell estimator eliminates ON volatility spikes
	show daily seasonality of volatility and volume - borrow from FRE7241_homeworks_tests.R
	correlation between volatility and volume, lead and lag
	plots for 2010 flash crash
	volatility estimation bias variance tradeoff

perform running strategy
	describe skew strategy - borrow from FRE7241_homeworks_tests.R
	describe vwap strategies - borrow from FRE7241_homeworks_tests.R
	inputs: returns, variance, pos_skew, neg_skew, volume, vwaps
	positive convexity with respect to large skew
	individual scatterplots
	convert all variables to categorical
	vwap has low t-val, but sign(vwap) has large t-val
	optimize vwap in period up to 2011, and test it out-of-sample
	perform lm() with zero intercept, and summarize regressions
	perform regressions over periods of high and low volatility and volume
	strong signal from skew in period of high vol?
	predict and calculate table confusion matrix
	regularize different vwaps (shrink) in-sample
	test vwaps out-of-sample
	apply gbm package: Generalized Boosted Regression Models?
	volatility is proxy for volume with lag?
	percentage of time pos_skew or neg_skew aren't zero?
	perform regressions over different calendar periods
		both vwap and skew more show stronger dependence during financial crisis
		vwap has steady persistence, but not skew?
	feature selection is most important
	frequency of trading: contrarian skew strategy versus vwap strategy
	add transaction costs by strategy
	plot daily seasonality of trading rates
	plot trading rates versus realized volatility
	evaluate strategies individually and create shaded plots
	evaluate strategy performance over different periods and link it to high volatility regimes
	contrarian skew strategy worked mostly during financial crisis
	simulate random prices and run through strategy
	future directions, open questions
	apply volume threshold to indicators - doesn't work because of daily seasonality?

investigate if skew is just proxy for past returns, or if it adds extra information
	start with daily aggregations
	trade next day at OPEN
	perform regressions in high and low volatility periods
	skew is negative on average, but changes over time
	cor.test()

expand slide "Daily Strategy Using Skew Oscillator"
	investigate if normalized skew is a contrarian indicator (oscillator) for future returns
	show that relationship is stronger in high volatility periods

investigate relationship between skew and returns in high vol periods
	show that relationship is stronger in high volatility periods

illustrate estimator efficiency as more data is used
MC simu

CFM anomaly
Donier Volatility Volume Market Liquidity Crash.pdf
volatility divided by square root of trading volume is a measure of market illiquidity
show that you can measure liquidity-adjusted imbalance over time

Why Do Markets Crash?  Bitcoin Data Offers Unprecedented Insights
http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2583743


order-book liquidity and price impact
the price impact is proportional to volatility times square root of ratio
of volume imbalance divided by trading volume

LeBaron Stock Momentum Volume
	Persistence is directly related to the current rate of change of volume

# sources:


### quote references:


### slides

# Are High Frequency Traders Prudent and Temperate?
	explain the concepts

# prudence and temperance

# packages xts, quantmod and TTR

# OHLC aggregations are good

# moment estimators
	volume weighted
	with O/N jumps or not?

# bias/variance tradeoff
      \vskip1ex
      Using high frequency data for estimation bias/variance tradeoff


### EARL Boston 2015 official

title:
Quantitative Portfolio Management with High Frequency Data

abstract:
Studies have shown that estimating portfolio risk parameters using high frequency data significantly improves their quality, and results in better portfolio performance.
We demonstrate elements of a computational framework in the R language that uses high frequency data for estimating risk and return parameters.
The parameters are used as input into models based on machine learning techniques such as cross validation (backtesting) and regularization.
The computational framework utilizes popular R packages for data scrubbing, aggregation, estimation, and machine learning.
We apply the framework to portfolio rebalancing, and we study alternative active management strategies.


### R Finance Chicago 2015

relationship between realized daily skewness and

But skewness and kurtosis are hard to estimate from low frequency (daily) returns.

There is much evidence that the third moment of returns is important for asset pricing.

Investor
for higher moments has also been studied, with
Their preference for higher moments has also been studied, with
the utility function

We use intraday data to compute weekly realized variance, skewness, and kurtosis
for equity
returns and study the realized moments time-series and cross-sectional properties.
We investigate if this weeks realized moments are informative for the cross-section of next weeks stock
returns. We and a very strong negative relationship between realized skewness and next weeks
stock returns. A trading strategy that buys stocks in the lowest realized skewness decile and
sells stocks in the highest realized skewness decile generates an average weekly return of 24
basis points with a t-statistic of 3:65. Our results on realized skewness are robust across a wide
variety of implementations, sample periods, portfolio weightings, and rm characteristics, and
are not captured by the Fama-French and Carhart factors. We nd some evidence that the re-
lationship between realized kurtosis and next weeks stock returns is positive, but the evidence
is not always robust and statistically signicant. We do not nd a strong relationship between
realized volatility and next weeks stock returns.