C****a
发帖数: 1639 | 1 Simple strategy:
rsi.tf.2 (trend following): buy RSI(2) > 50; short RSI(2) < 50
rsi.cf.2 (mean reversion): buy RSI(2) < 50; short RS(2) > 50
bah: buy and hold
Adaptive strategy:
For individual stock, choose the best simple strategy based on a secret
measurement.
For a group of stocks, choose five best strategies based on the same
measurement.
Here is the backtesting results for AMAT/AAPL and NQ100 stocks.
Pretty amazing performance for NQ100
Without considering erosion, the total returns is 157900%, since 1999
When considering erosion (commission), the total returns is 141400 % since
1999, while CAGR is 93.3%, and
max drawdown is 36%.
The commission adjusted performance is better than the new livermore index:
http://cssanalytics.wordpress.com/2010/03/11/new-livermore-acti |
C****a
发帖数: 1639 | 2 Another view of the performance for NQ100 stocks.
【在 C****a 的大作中提到】
: Simple strategy:
: rsi.tf.2 (trend following): buy RSI(2) > 50; short RSI(2) < 50
: rsi.cf.2 (mean reversion): buy RSI(2) < 50; short RS(2) > 50
: bah: buy and hold
: Adaptive strategy:
: For individual stock, choose the best simple strategy based on a secret
: measurement.
: For a group of stocks, choose five best strategies based on the same
: measurement.
: Here is the backtesting results for AMAT/AAPL and NQ100 stocks.
|
g********4
发帖数: 4959 | 3 牛叉。可惜我一点都不懂。学电脑的就是好。文科男哭死。 |
w*******o
发帖数: 6125 | 4 Rule of thumb:
Back test makes money, doesn't mean forward trading makes money.
【在 C****a 的大作中提到】
: Simple strategy:
: rsi.tf.2 (trend following): buy RSI(2) > 50; short RSI(2) < 50
: rsi.cf.2 (mean reversion): buy RSI(2) < 50; short RS(2) > 50
: bah: buy and hold
: Adaptive strategy:
: For individual stock, choose the best simple strategy based on a secret
: measurement.
: For a group of stocks, choose five best strategies based on the same
: measurement.
: Here is the backtesting results for AMAT/AAPL and NQ100 stocks.
|
C****a
发帖数: 1639 | 5 The beauty of adaptive strategies that you don't need to have any assumption
, don't need to rely on any single
strategy, like MA crossover, support/resistance, MACD or STO.
The failure of some backtesting is because they are relying on a single
magic indicator or something. But
adaptive strategies do not have to.
【在 w*******o 的大作中提到】
: Rule of thumb:
: Back test makes money, doesn't mean forward trading makes money.
|
a*****e
发帖数: 1717 | 6 157900% ? or 1579%
sigma_i=1,100 r_i ?
【在 C****a 的大作中提到】
: Simple strategy:
: rsi.tf.2 (trend following): buy RSI(2) > 50; short RSI(2) < 50
: rsi.cf.2 (mean reversion): buy RSI(2) < 50; short RS(2) > 50
: bah: buy and hold
: Adaptive strategy:
: For individual stock, choose the best simple strategy based on a secret
: measurement.
: For a group of stocks, choose five best strategies based on the same
: measurement.
: Here is the backtesting results for AMAT/AAPL and NQ100 stocks.
|
a*****e
发帖数: 1717 | 7 我来倒乱一下。
if stock price is 1,2,1,2,1,2,1,2
RSI(2) will stand aside watching forever :D
【在 C****a 的大作中提到】
: Simple strategy:
: rsi.tf.2 (trend following): buy RSI(2) > 50; short RSI(2) < 50
: rsi.cf.2 (mean reversion): buy RSI(2) < 50; short RS(2) > 50
: bah: buy and hold
: Adaptive strategy:
: For individual stock, choose the best simple strategy based on a secret
: measurement.
: For a group of stocks, choose five best strategies based on the same
: measurement.
: Here is the backtesting results for AMAT/AAPL and NQ100 stocks.
|
C****a
发帖数: 1639 | 8 The beginning is 1, and end is 1579
Note the y axis is log.
【在 a*****e 的大作中提到】
: 157900% ? or 1579%
: sigma_i=1,100 r_i ?
|
C*G
发帖数: 7495 | 9 同学，您老也是文科男?！幸会！
hiahia
【在 g********4 的大作中提到】
: 牛叉。可惜我一点都不懂。学电脑的就是好。文科男哭死。
|
g*****u
发帖数: 14294 | 10 WOW!!! 1579x
Do you have paper trading results staring when?
【在 C****a 的大作中提到】
: The beginning is 1, and end is 1579
: Note the y axis is log.
|
|
|
N******r
发帖数: 642 | 11 it was impossible to short stocks in 2008. have you factored that in your
backtesting? |
C****a
发帖数: 1639 | 12 The beauty of adaptivity is that you can switch away from this loser, to,
like, 1,2,3,4,5,6,7
【在 a*****e 的大作中提到】
: 我来倒乱一下。
: if stock price is 1,2,1,2,1,2,1,2
: RSI(2) will stand aside watching forever :D
|
C****a
发帖数: 1639 | 13 No, not considered yet.
【在 N******r 的大作中提到】
: it was impossible to short stocks in 2008. have you factored that in your
: backtesting?
|
a*****e
发帖数: 1717 | 14 the logic relies on the secret measurement.
what is the target your want to maximize/minimize,
as you said sortino is your measure.
you evaluate three strategies by this measurement.
but I think bah is redundant,
RSI(2) is such a binary stuff, it's hard to imagine bah
better than mr and tf at the same time.
【在 C****a 的大作中提到】
: The beauty of adaptivity is that you can switch away from this loser, to,
: like, 1,2,3,4,5,6,7
|
C****a
发帖数: 1639 | 15 Even that livermore issue index has a returns like 34661%...
http://cssanalytics.wordpress.com/2010/03/11/new-livermore-acti
The backtest ends on Mar. 07, I will see what it performed since then.
【在 g*****u 的大作中提到】
: WOW!!! 1579x
: Do you have paper trading results staring when?
|
a*****e
发帖数: 1717 | 16 really? as far as I remember,
borrowing cost is higher, Banks stocks were banned for short in a short term
,
Same for technical sector?
【在 N******r 的大作中提到】
: it was impossible to short stocks in 2008. have you factored that in your
: backtesting?
|
g*****u
发帖数: 14294 | 17 I from time to time take a peek at that blog.
I am trying to get a sense of how the system perform live. Maybe it is just
me, or the answer is not totally obvious.
【在 C****a 的大作中提到】
: Even that livermore issue index has a returns like 34661%...
: http://cssanalytics.wordpress.com/2010/03/11/new-livermore-acti
: The backtest ends on Mar. 07, I will see what it performed since then.
|
a*****e
发帖数: 1717 | 18 keep rolling over the top winners?
this strategy is the most naive form of statistical arbitrage?
usually ppl long winners short losers as hedge.
【在 C****a 的大作中提到】
: Even that livermore issue index has a returns like 34661%...
: http://cssanalytics.wordpress.com/2010/03/11/new-livermore-acti
: The backtest ends on Mar. 07, I will see what it performed since then.
|
a*****e
发帖数: 1717 | 19 simply effective but DD is not tolerable
and no transaction cost considered.
【在 C****a 的大作中提到】
: Even that livermore issue index has a returns like 34661%...
: http://cssanalytics.wordpress.com/2010/03/11/new-livermore-acti
: The backtest ends on Mar. 07, I will see what it performed since then.
|
C****a
发帖数: 1639 | 20 Sortino is one of them. But the problem of using sortino/sharpe alone is
that for some stocks, e.g. AMAT,
sortino of simple strategies tend to oscillate in a certain window. You can
try it.
【在 a*****e 的大作中提到】
: the logic relies on the secret measurement.
: what is the target your want to maximize/minimize,
: as you said sortino is your measure.
: you evaluate three strategies by this measurement.
: but I think bah is redundant,
: RSI(2) is such a binary stuff, it's hard to imagine bah
: better than mr and tf at the same time.
|
|
|
C****a
发帖数: 1639 | 21 Not exactly.
They said they capture the stocks with
1. Consistent trend, and
2. Detached from the benchmark.
They did not disclose how they do that, though. The good for that index iss
that they only rebalance their
portfolio every 5 days (i.e. every Friday), which is much less stressful. I
like it.
If simply buying winner, it will be slaughtered in this decade.
【在 a*****e 的大作中提到】
: keep rolling over the top winners?
: this strategy is the most naive form of statistical arbitrage?
: usually ppl long winners short losers as hedge.
|
a*****e
发帖数: 1717 | 22 My guess : try different timeframes might reconcile.
can
【在 C****a 的大作中提到】
: Sortino is one of them. But the problem of using sortino/sharpe alone is
: that for some stocks, e.g. AMAT,
: sortino of simple strategies tend to oscillate in a certain window. You can
: try it.
|
C****a
发帖数: 1639 | 23 They describe the concept here:
http://cssanalytics.wordpress.com/2010/05/24/the-conceptual-lin
No details though. But it would not be difficult to develop something
similar, I guess.
just
【在 g*****u 的大作中提到】
: I from time to time take a peek at that blog.
: I am trying to get a sense of how the system perform live. Maybe it is just
: me, or the answer is not totally obvious.
|
a*****e
发帖数: 1717 | 24 how to understand "detached"
iss
I
【在 C****a 的大作中提到】
: Not exactly.
: They said they capture the stocks with
: 1. Consistent trend, and
: 2. Detached from the benchmark.
: They did not disclose how they do that, though. The good for that index iss
: that they only rebalance their
: portfolio every 5 days (i.e. every Friday), which is much less stressful. I
: like it.
: If simply buying winner, it will be slaughtered in this decade.
|
C****a
发帖数: 1639 | 25 I tried from 50 to 2000, haha
Yes, for longer time frame, it can partly solve, but it cannot detect the
short term opportunity. So I give up
using that one alone.
BTW, I am not the guy who use different time frame/parameters for different
stocks. :)
【在 a*****e 的大作中提到】
: My guess : try different timeframes might reconcile.
:
: can
|
C****a
发帖数: 1639 | 26 Quote from the blog:
This means that stocks that have powerful fundamentals and a strong business
model like an Apple Computer
or a Baidu.com will be able to march to the beat of their own drummer at
times, rising even when the market
may be falling. This detached behavior is especially evident on the shorter
time frames. Since high LTR stocks
must exhibit a smooth price history, the detachment often manifests itself
in a persistent/consistent manner.
This behavior helps to improve the risk-adjusted returns of holding such
stocks since they tend to have strong
momentum versus volatility.
【在 a*****e 的大作中提到】
: how to understand "detached"
:
: iss
: I
|
a*****e
发帖数: 1717 | 27 that's is kinda subjective strategy:D
business
shorter
【在 C****a 的大作中提到】
: Quote from the blog:
: This means that stocks that have powerful fundamentals and a strong business
: model like an Apple Computer
: or a Baidu.com will be able to march to the beat of their own drummer at
: times, rising even when the market
: may be falling. This detached behavior is especially evident on the shorter
: time frames. Since high LTR stocks
: must exhibit a smooth price history, the detachment often manifests itself
: in a persistent/consistent manner.
: This behavior helps to improve the risk-adjusted returns of holding such
|
t****g
发帖数: 3434 | |
