Академический Документы
Профессиональный Документы
Культура Документы
FROM A DRAWDOWN?
Marcos Lpez de Prado
Lawrence Berkeley National Laboratory
Computational Research Division
Key Points
Investment management firms routinely hire and fire
employees based on the performance of their portfolios.
Such performance is evaluated through popular metrics that
assume IID Normal returns, like Sharpe ratio, Sortino ratio,
Treynor ratio, Information ratio, etc.
Investment returns are far from IID Normal.
If we accept first-order serial correlation:
Maximum Drawdown is generally greater than in IID Normal case.
Time Under Water is generally longer than in IID Normal case.
However, Penance is typically shorter than 3x (IID Normal case).
SECTION I
The Need for Performance Evaluation
Investors
funds
Allocation
N
4
SECTION II
Stop-Outs under the IID Normal Assumption
=1
, : = +
where is the critical value of the Standard Normal
distribution associated with a probability of performing
worse than , , i.e. : = , . Then, quantile
loss is defined as
, : = max 0, ,
9
Maximum Drawdown
PROPOSITION 1: Assuming IID outcomes ~ , 2 ,
and > 0, the maximum quantile-loss associated with a
1
significance level < is
2
=
4
=
2
10
= 2 2 +
4000000
3000000
2000000
1000000
0
-0.1
0.1
0.3
0.5
0.7
0.9
1.1
1.3
1.5
-1000000
-2000000
-3000000
-4000000
-5000000
3
Time Under the Water
13
Example 1
PM1 has an annual mean
and standard deviation of
US$10m (SR=1), and PM2
has an annual mean of
US$15m and an annual
standard deviation of
US$10m (SR=1.5).
20000000
0.05 [PM1]
15000000
0.05 [PM2]
5000000
-5000000
-10000000
0.5
0.05 [1]
0.05 [2]
10000000
1.5
PM2
2.5
Example 2
2000000
1000000
0.08 1 = 0.02 2
0
0
0.5
1.5
-1000000
Quantile (in US$)
0.08 [1]
-2000000
-4000000
-5000000
-6000000
-8000000
2.5
-3000000
-7000000
0.02 [2]
3
Time Under the Water
PM1
PM2
SECTION III
The IID Normal Assumption
0.012
0.01
0.008
0.006
0.004
0.002
0
-0.1
-0.08
-0.06
-0.04
pdf1
-0.02
pdf2
pdf Mixture
0.02
0.04
0.06
0.08
pdf Normal
0.1
SECTION IV
Stop-Outs under first-order auto-correlated outcomes
First-order auto-correlation
It is well established that hedge fund strategies
exhibit significant first-order auto-correlation. E.g.,
see Brooks and Kat [2002].
There are various reasons why strategies returns
exhibit first-order serial-correlation:
Unmonitored risk concentration (quite different from VaR).
Inconsistent profit taking and stop loss rules.
Serially correlated and cointegrated investments.
2 +1 1
+1 1
2
++1
2
1
1
23
+1
=
0 +
1
+
2 +1
+1
2
++1
2
1
1
SECTION V
Mean
0.0055
0.0089
0.0099
0.0095
0.0052
0.0096
0.0052
0.0048
0.0095
0.0085
0.0072
0.0069
0.0080
0.0071
0.0071
0.0104
0.0080
-0.0017
0.0111
0.0068
0.0101
0.0094
0.0056
0.0089
0.0084
0.0111
Phi
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
Sigma
0.0170
0.0202
0.0264
0.0215
0.0174
0.0188
0.0094
0.0116
0.0192
0.0216
0.0177
0.0129
0.0130
0.0104
0.0200
0.0410
0.0382
0.0535
0.0508
0.0248
0.0367
0.0360
0.0159
0.0390
0.0363
0.0464
MaxQL
3.53%
3.10%
4.80%
3.28%
3.96%
2.48%
1.16%
1.90%
2.63%
3.69%
2.95%
1.64%
1.42%
1.03%
3.79%
10.98%
12.38%
-15.79%
6.09%
9.02%
9.33%
3.05%
11.50%
10.58%
13.17%
t*
6.3996
3.4905
4.8667
3.4435
7.6477
2.5827
2.2389
3.9492
2.7554
4.3218
4.1164
2.3883
1.7701
1.4444
5.3200
10.6100
15.4963
-14.2615
8.9046
8.9357
9.9416
5.4422
12.8580
12.5579
11.8933
MaxTuW
25.5985
13.9621
19.4669
13.7740
30.5909
10.3309
8.9554
15.7968
11.0216
17.2870
16.4656
9.5530
7.0803
5.7777
21.2800
42.4399
61.9851
-57.0458
35.6185
35.7430
39.7662
21.7686
51.4319
50.2317
47.5731
Penance
3.0000
3.0000
3.0000
3.0000
3.0000
3.0000
3.0000
3.0000
3.0000
3.0000
3.0000
3.0000
3.0000
3.0000
3.0000
3.0000
3.0000
-3.0000
3.0000
3.0000
3.0000
3.0000
3.0000
3.0000
3.0000
Mean
0.0055
0.0089
0.0099
0.0095
0.0052
0.0096
0.0052
0.0048
0.0095
0.0085
0.0072
0.0069
0.0080
0.0071
0.0071
0.0104
0.0080
-0.0017
0.0111
0.0068
0.0101
0.0094
0.0056
0.0089
0.0084
0.0111
StDev
0.0170
0.0202
0.0264
0.0215
0.0174
0.0188
0.0094
0.0116
0.0192
0.0216
0.0177
0.0129
0.0130
0.0104
0.0200
0.0410
0.0382
0.0535
0.0508
0.0248
0.0367
0.0360
0.0159
0.0390
0.0363
0.0464
Phi
0.3594
0.3048
0.2651
0.1844
0.3535
0.5458
0.1644
0.4557
0.3916
-0.0188
0.4838
0.5059
0.4528
0.2982
0.5780
0.3593
0.3112
0.0907
0.1969
0.3231
0.2011
0.2314
0.0422
0.0505
0.0954
0.1608
Sigma
0.0158
0.0192
0.0255
0.0211
0.0163
0.0158
0.0093
0.0103
0.0177
0.0216
0.0155
0.0111
0.0116
0.0100
0.0163
0.0383
0.0363
0.0533
0.0499
0.0235
0.0359
0.0350
0.0159
0.0390
0.0361
0.0458
t-Stat(Phi)
6.2461
5.1907
4.4601
3.0419
6.1295
10.5612
2.7035
8.3023
6.9021
-0.3051
8.9720
9.5874
8.2430
5.0670
11.4865
6.2431
5.3109
1.4776
3.2575
5.5360
3.3299
3.8573
0.6842
0.8200
1.5542
2.6428
MaxQL
6.65%
4.74%
7.27%
4.15%
7.52%
5.40%
1.33%
4.00%
4.34%
-6.69%
3.12%
2.00%
1.08%
11.60%
21.71%
22.57%
-22.77%
11.00%
12.84%
14.15%
3.25%
12.59%
12.55%
17.61%
t*
14.5551
7.3222
9.0236
5.4157
16.9638
10.7065
3.4722
11.9696
7.3855
-13.3986
8.9080
5.9134
3.2508
22.1308
23.4821
30.2969
-21.7061
18.2415
13.8963
16.4723
6.0074
14.3295
15.4084
16.8089
TuW
Penance
52.1831
2.5852
24.4918
2.3449
32.1120
2.5587
19.1093
2.5285
61.9700
2.6531
30.4208
1.8413
11.6921
2.3674
39.0229
2.2602
22.6758
2.0703
--43.7383
2.2644
25.0456
1.8116
15.3920
1.6029
8.9163
1.7428
74.4170
2.3626
87.9134
2.7439
116.2881 2.8383
--84.0775
2.8735
67.7961
2.7166
52.7651
2.7971
62.5481
2.7972
23.5097
2.9135
56.6921
2.9563
60.4123
2.9207
65.0637
2.8708
The t-Stat of is
inconsistent with
the IID Normal
assumption in 21
out of 26
strategies, with a
95% confidence
level.
4
2
1
1
4
as
1 +1
0 + 1
1
+1
1
1 +1 1
2
+
+1
1
1
2 1
29
MaxQL
0.0353
0.0310
0.0480
0.0328
0.0396
0.0248
0.0116
0.0190
0.0263
0.0369
0.0295
0.0164
0.0142
0.0103
0.0379
0.1098
0.1238
-0.1579
0.0609
0.0902
0.0933
0.0305
0.1150
0.1058
0.1317
t*
6.3996
3.4905
4.8667
3.4435
7.6477
2.5827
2.2389
3.9492
2.7554
4.3218
4.1164
2.3883
1.7701
1.4444
5.3200
10.6100
15.4963
-14.2615
8.9046
8.9357
9.9416
5.4422
12.8580
12.5579
11.8933
Alpha1
0.0500
0.0500
0.0500
0.0500
0.0500
0.0500
0.0500
0.0500
0.0500
0.0500
0.0500
0.0500
0.0500
0.0500
0.0500
0.0500
0.0500
0.0500
0.0500
0.0500
0.0500
0.0500
0.0500
0.0500
0.0500
0.0500
Mean2
0.0055
0.0089
0.0099
0.0095
0.0052
0.0096
0.0052
0.0048
0.0095
0.0085
0.0072
0.0069
0.0080
0.0071
0.0071
0.0104
0.0080
-0.0017
0.0111
0.0068
0.0101
0.0094
0.0056
0.0089
0.0084
0.0111
Phi2
0.3594
0.3048
0.2651
0.1844
0.3535
0.5458
0.1644
0.4557
0.3916
-0.0188
0.4838
0.5059
0.4528
0.2982
0.5780
0.3593
0.3112
0.0907
0.1969
0.3231
0.2011
0.2314
0.0422
0.0505
0.0954
0.1608
Sigma2
0.0158
0.0192
0.0255
0.0211
0.0163
0.0158
0.0093
0.0103
0.0177
0.0216
0.0155
0.0111
0.0116
0.0100
0.0163
0.0383
0.0363
0.0533
0.0499
0.0235
0.0359
0.0350
0.0159
0.0390
0.0361
0.0458
Alpha2
0.1205
0.1014
0.0975
0.0796
0.1207
0.1312
0.0728
0.1331
0.1114
-0.1400
0.1224
0.1029
0.0814
0.1688
0.1243
0.1139
-0.0873
0.1145
0.0872
0.0940
0.0567
0.0586
0.0667
0.0795
In some cases, firms may be firing more than three times the number
of skillful PMs, compared to the number they were willing to accept
under the (wrong) assumption of returns independence.
30
HFRIFWIC Index
HFRIFWIJ Index
2.9
HFRIEMLA Index
HFRIFOFM Index
HFRIEMA Index
HFRISTI Index
HFRIFWIG Index
HFRIENHI Index
HFRIEM Index
HFRIFOFS Index
2.7
HFRIFOFD Index
HFRIMI Index
2.5
HFRIEHI Index
HFRICAI Index
HFRIEMNI Index
Penance
HFRIFOF Index
HFRIFWI Index
2.3
HFRIFIHY Index
HFRIFOFC Index
2.1
HFRIEDI Index
1.9
HFRIDSI Index
HFRIFI Index
HFRIMAI Index
1.7
HFRIRVA Index
1.5
0.0
0.1
0.2
0.3
Phi
0.4
0.5
0.6
Although positive
serial correlation leads
to greater drawdowns,
longer and longer
periods under water,
Penance may be
substantially smaller. In
particular, Penance is
smaller the higher
(Phi) and the higher
SECTION VI
Conclusions
Conclusions (1/2)
1. Far from being a theoretical argument, wrongly assuming
that returns are IID Normal has measurable costs to firms
and investors.
2. Assuming IID Normal returns leads to the Triple Penance
rule: Regardless of the Sharpe ratio of a strategy, it takes 3
times longer to recover from a maximum drawdown than to
produce it, with the same confidence level.
3. However, taking serial dependence into account leads to
Penance lower than 3x.
4. In particular, under first-order auto-correlation, Penance is
lower the greater the Sharpe ratio and also the greater the
serial dependence.
33
Conclusions (2/2)
5. In some hedge fund strategies, if the accepted probability of
false positives was 5%, the actual rate at which skillful PMs
are fired is up to three times greater.
6. This is extremely costly: If two out of three PMs are wrongly
fired
35
SECTION VII
The stuff nobody reads
Bibliography (1/4)
Bibliography (2/4)
Bibliography (3/4)
Lpez de Prado, M. and A. Peijan (2004): Measuring the Loss Potential of Hedge
Fund Strategies. Journal of Alternative Investments, Vol. 7(1), pp. 7-31. Available
at http://ssrn.com/abstract=641702.
Lpez de Prado, M. and M. Foreman (2012): Markowitz meets Darwin: Portfolio
Oversight and Evolutionary Divergence. Working paper, RCC at Harvard University.
Available at http://ssrn.com/abstract=1931734.
Magdon-Ismail, M. and A. Atiya (2004): Maximum drawdown. Risk Magazine,
October.
Magdon-Ismail, M., A. Atiya, A. Pratap and Y. Abu-Mostafa (2004): On the
maximum drawdown of a Brownian motion. Journal of Applied Probability, Vol.
41(1).
Markowitz, H.M. (1952): Portfolio Selection. Journal of Finance, Vol. 7(1), pp. 77
91.
Markowitz, H.M. (1956): The Optimization of a Quadratic Function Subject to
Linear Constraints. Naval Research Logistics Quarterly, Vol. 3, 111133.
Markowitz, H.M. (1959): Portfolio Selection: Efficient Diversification of
Investments. John Wiley and Sons.
39
Bibliography (4/4)
Bio
Marcos Lpez de Prado is Senior Managing Director at Guggenheim Partners. He is also a Research
Affiliate at Lawrence Berkeley National Laboratory's Computational Research Division (U.S. Department
of Energys Office of Science).
Before that, Marcos was Head of Quantitative Trading & Research at Hess Energy Trading Company (the
trading arm of Hess Corporation, a Fortune 100 company) and Head of Global Quantitative Research at
Tudor Investment Corporation. In addition to his 15+ years of trading and investment management
experience at some of the largest corporations, he has received several academic appointments,
including Postdoctoral Research Fellow of RCC at Harvard University and Visiting Scholar at Cornell
University. Marcos earned a Ph.D. in Financial Economics (2003), a second Ph.D. in Mathematical
Finance (2011) from Complutense University, is a recipient of the National Award for Excellence in
Academic Performance by the Government of Spain (National Valedictorian, 1998) among other awards,
and was admitted into American Mensa with a perfect test score.
Marcos is the co-inventor of four international patent applications on High Frequency Trading. He has
collaborated with ~30 leading academics, resulting in some of the most read papers in Finance (SSRN),
three textbooks, publications in the top Mathematical Finance journals, etc. Marcos has an Erds #3 and
an Einstein #4 according to the American Mathematical Society.
41
Disclaimer
The views expressed in this document are the authors
and do not necessarily reflect those of the
organizations he is affiliated with.
No investment decision or particular course of action is
recommended by this presentation.
All Rights Reserved.
42
Notice:
The research contained in this presentation is the result of
a continuing collaboration with
Prof. David H. Bailey, LBNL
The full paper is available at:
http://ssrn.com/abstract=2201302
For additional details, please visit:
http://ssrn.com/author=434076
www.QuantResearch.info