QuantConnect Community

@PeterGuenther It took me much longer to find the time to collect all the information I could about the timeline of changes I made in my strategies. The first time I used DBB and UUP as “canary assets” was in Mar. 02, 2021 in the article at the following link: https://seekingalpha.com/article/4410670-momentum-strategies-djia-stocks I initiated entering a reference as follows: “ The usage of a pair (DBB, UUP) as an indicator for market risk-off periods was introduced by Peter Guenther in a blog on Quantopian.com. A continuation of that blog is now available at the following link: https://www.quantconnect.com/forum/discussion/9597/the-in-amp-out-strategy-continued-from-quantopian/” Because of the delay in editing past posts by Seeking Alpha editors, the references will post only at the beginning of next week.

Meanwhile, I intend to start a new discussion thread here at QuantConnect to complement your thread with information about Momentum strategies published by many authors. Here is the list of authors I plan to mention in that blog: Yegadeesh and Titman (2002), Mebane Faber (2010), Gary Antonacci (2013), Frank Grossmann (2013), Asness, Moskowitz and Pedersen (2015), Keller and Keuning (2016), Yulia Malitskaia (2019), Michael Gettings (2019) and Scott Juds (2020).

Additionally, I will mention two of my articles published by Seeking Alpha. The first is an article in 2014 where I used a three-month look back period for momentum evaluation when most authors were still using a period of 12 months as proposed by Jegadeesh and Titman, period confirmed by Antonacci as optimal for a period of almost 100 years. In my second article (2019), I showed that for a broad class of assets, the optimal momentum period has changed. From 1988 to 2008 the optimal period was 12 months, but for 2008 to 2019, he optimal was 3 years. This is an example as argument for using the most recent past for optimization of strategy parameters as opposed to using very long past periods.

I hope that this new thread would reach many individuals who can contribute to further development of your strategies. Here I want to mention that the reason why I have never posted anything on Quantopian was the fact they were insisting on market neutral strategies that required shorting some assets. On your thread I had the impression that there was much effort to adapt the strategy for automatic intraday trading. All my strategies are long term investments with minimal trading. My strategies have been risk-on since May 2020. When DBB-UUP indicated risk-off I did not apply it because it was overridden by other conditions I use.

Finally, I would like to post a copy of my initial QuantConnect post as a blog on Seeking Alpha. There, some Seeking Alpha members will be exposed to information about your work.      

Stephen David Bennett Wondering if you saw my post above with Link to alternative data that I have assembled for longer duration back testing. Data is from 1995. The data consists of futures and mutual funds that seem to be equivalent to the ETFs being used. https://www.dropbox.com/s/jhbv45w111szult/BackTestingData.csv?dl=0

Hi 

I tested the effect of the window lenght used for percentage change of the switching pairs. (pairs used to decide on In-OUT)

First test shows the difference between a window length of 2-800 days for the whole period from 2008 until today (one return for the whole period and variation of the window lenght)

Second test, I calculated the return with a rolling window of one year and than looked for the best performing window and its return.

Both test were done for different single pairs. IN-OUT was done with just the SPY. In the first test I switched between SPY and IEF and in the second test between SPY and Cash.

From the first test it looked that around 60 days give the best result/return for the whole period.

The second test, with a rolling window return and picking the optimal window, showed that the optimal window lenght ist ALL OVER THE PLACE….

And now? How should we decide which window length to use and to avoid fitting the window to the best historical result?

Please don't understand me wrong, I'm personally using this method for my investment and I really want to improve the method, but don't know how.

Here the result for the pair DBB/UUP. Upper graphic shows the optimal window lenght.

Following the notebook with the code:

# Make sure the plot shows up

%matplotlib inline



# Import libraries that we need

import pandas as pd 

import numpy as np

import yfinance as yf

import pyfolio as pf



import matplotlib.pyplot as plt



symbols_to_use = '^VIX ^VXV SPY QQQ MDY IEF TLT DBB UUP XLY XLP SLV GLD XLI XLU FXA FXF FXC FXY SSO QLD MVV SPXL TQQQ MIDU'



    

# indicator from in-out Quantopian

#    self.slv_sid = symbol('SLV').sid # silver

#    self.gld_sid = symbol('GLD').sid # gold

#    self.xli_sid = symbol('XLI').sid # Consumer Industrial Select Sector SPDR Fund

#    self.xlu_sid = symbol('XLU').sid # Consumer Utility Select Sector SPDR Fund

#    self.fxa_sid = symbol('FXA').sid # Swiss franc

#    self.fxf_sid = symbol('FXF').sid # Ausi dollar

    

# indicator from Toma Hentea

#    self.dbb_sid = symbol('DBB').sid  # Invesco DB Base Metals Fund

#    self.uup_sid = symbol('UUP').sid # Invesco DB US Dollar Index Fund

#    self.xly_sid = symbol('XLY').sid # Consumer Discretionary Select Sector SPDR Fund

#    self.xlp_sid = symbol('XLP').sid # Consumer Staples Select Sector SPDR Fund 



def timing_test(data,symbl_1, symbl_2,start,end):

    

    data['eqity_ret'] = data['QQQ'].pct_change(1)

    data['bonds_ret'] = data['IEF'].pct_change(1)

    

    cagr = np.zeros(end)

    for i in range(start,end):



        window_risk_on_off = i    



        # definition of momentum of tickers used to define risk on risk off 

        data['a_mom'] = data[symbl_1].pct_change(window_risk_on_off)

        data['b_mom'] = data[symbl_2].pct_change(window_risk_on_off)



        #shift 1 day to avoid hindsight bias     

        data['a_mom'] = data['a_mom'].shift()

        data['b_mom'] = data['b_mom'].shift()



        # now we make the decision if risk is on or off

        data["risk_off"] = np.where ((data['a_mom']<data['b_mom']),1,0) 



        data["dummy"] = np.where(data['risk_off'] == 1, data['bonds_ret'], data['eqity_ret'])



        strategy_ret_daily = data["dummy"][i:]



        days=len(strategy_ret_daily)

        cagr[i] = (((((strategy_ret_daily+1).prod())**(1/days))**252-1)*100).round(2)



    fig, ax = plt.subplots(figsize=(20,5))

    ax.plot(cagr)



    ax.set(xlabel='window', ylabel='cagr',

           title= symbl_1 + ' vs '+ symbl_2 )

    ax.grid()



    fig.savefig("test.png")

    plt.show()    

    

def timing_test_2(data,roll_wind,symbl_1, symbl_2,start,end):

    

    strategy_ret_daily_rolling = pd.DataFrame()

    result = pd.DataFrame(columns=['best_window','best_returns','sec_best_window','sec_best_returns'])

    

    

    # ticker of benchmark investment

    data['eqity_ret'] = data['SPY'].pct_change(1)

    data['bonds_ret'] = data['IEF'].pct_change(1)

    data['cash_ret'] = data['bonds_ret']

    data['cash_ret'] = 0.0

    

    cagr = np.zeros(end)

    #strategy_ret_daily_rolling = np.zeros((end,len_data))

    

    # running the strategie for all momentum windows

    for i in range(start,end,10):



        window_risk_on_off = i 

        #print(i)



        # definition of momentum of tickers used to define risk on risk off 

        data['a_mom'] = data[symbl_1].pct_change(window_risk_on_off)

        data['b_mom'] = data[symbl_2].pct_change(window_risk_on_off)



        #shift 1 day to avoid hindsight bias     

        data['a_mom'] = data['a_mom'].shift()

        data['b_mom'] = data['b_mom'].shift()



        # now we make the decision if risk is on or off

        data["risk_off"] = np.where ((data['a_mom']<data['b_mom']),1,0) 



        #data["dummy"] = np.where(data['risk_off'] == 1, data['bonds_ret'], data['eqity_ret'])

        data["dummy"] = np.where(data['risk_off'] == 1, data['cash_ret'], data['eqity_ret'])



        strategy_ret_daily = data["dummy"][i:]

        

        # 22 days rolling window to see the gains 

        days=roll_wind

        

        # calculating the rolling returns, on yearly basis

        strategy_ret_daily_rolling[i] = strategy_ret_daily.rolling(days).apply(lambda x: (((np.prod(1+x)**(1/days))**252-1)*100).round(2))

                

    len_x = strategy_ret_daily_rolling.shape[0]

    

    #print(strategy_ret_daily_rolling.head())

    #strategy_ret_daily_rolling.to_csv('out.csv', index=True)

    

    result = pd.DataFrame(np.zeros([len_x, 12]),columns=[

    'window_1','returns_1','window_2','returns_2','window_3','returns_3','window_4','returns_4','window_5','returns_5','window_6','returns_6'

    ])

    

    result = result.reindex(strategy_ret_daily_rolling.index)



    

    

    for j in range(0,len_x):

        if len(strategy_ret_daily_rolling.iloc[j,:].loc[strategy_ret_daily_rolling.iloc[j,:]==strategy_ret_daily_rolling.iloc[j,:].max()]) > 0:

                      

            # first max

            a=strategy_ret_daily_rolling.iloc[j,(strategy_ret_daily_rolling.iloc[j,:]==strategy_ret_daily_rolling.iloc[j,:].max()).values]

            

            # set max to zero

            strategy_ret_daily_rolling.iloc[j,(strategy_ret_daily_rolling.iloc[j,:]==strategy_ret_daily_rolling.iloc[j,:].max()).values]=0

            

            #second max

            b=strategy_ret_daily_rolling.iloc[j,(strategy_ret_daily_rolling.iloc[j,:]==strategy_ret_daily_rolling.iloc[j,:].max()).values]

            

            # set max to zero

            strategy_ret_daily_rolling.iloc[j,(strategy_ret_daily_rolling.iloc[j,:]==strategy_ret_daily_rolling.iloc[j,:].max()).values]=0

            

            # third max

            c=strategy_ret_daily_rolling.iloc[j,(strategy_ret_daily_rolling.iloc[j,:]==strategy_ret_daily_rolling.iloc[j,:].max()).values]



            # set max to zero

            strategy_ret_daily_rolling.iloc[j,(strategy_ret_daily_rolling.iloc[j,:]==strategy_ret_daily_rolling.iloc[j,:].max()).values]=0

            

            # 4 max

            d=strategy_ret_daily_rolling.iloc[j,(strategy_ret_daily_rolling.iloc[j,:]==strategy_ret_daily_rolling.iloc[j,:].max()).values]

            

            # set max to zero

            strategy_ret_daily_rolling.iloc[j,(strategy_ret_daily_rolling.iloc[j,:]==strategy_ret_daily_rolling.iloc[j,:].max()).values]=0

            

            # 5 max

            e=strategy_ret_daily_rolling.iloc[j,(strategy_ret_daily_rolling.iloc[j,:]==strategy_ret_daily_rolling.iloc[j,:].max()).values]

            

            # set max to zero

            strategy_ret_daily_rolling.iloc[j,(strategy_ret_daily_rolling.iloc[j,:]==strategy_ret_daily_rolling.iloc[j,:].max()).values]=0

            

            # 6 max

            f=strategy_ret_daily_rolling.iloc[j,(strategy_ret_daily_rolling.iloc[j,:]==strategy_ret_daily_rolling.iloc[j,:].max()).values]





           

                        

            if len(a)>1 or len(b)>1 or len(c)>1 or len(d)>1 or len(e)>1 or len(f)>1  :

                result.window_1.iloc[j]=np.nan 

                result.returns_1.iloc[j]=np.nan

                result.window_2.iloc[j]=np.nan 

                result.returns_2.iloc[j]=np.nan

                result.window_3.iloc[j]=np.nan 

                result.returns_3.iloc[j]=np.nan 

                result.window_4.iloc[j]=np.nan 

                result.returns_4.iloc[j]=np.nan

                result.window_5.iloc[j]=np.nan 

                result.returns_5.iloc[j]=np.nan

                result.window_6.iloc[j]=np.nan 

                result.returns_6.iloc[j]=np.nan 

                 

            else: 

                

                result.window_1.iloc[j]=a.index[0]

                result.returns_1.iloc[j]=a

                result.window_2.iloc[j]=b.index[0]

                result.returns_2.iloc[j]=b

                result.window_3.iloc[j]=c.index[0]

                result.returns_3.iloc[j]=c

                result.window_4.iloc[j]=d.index[0]

                result.returns_4.iloc[j]=d

                result.window_5.iloc[j]=e.index[0]

                result.returns_5.iloc[j]=e

                result.window_6.iloc[j]=f.index[0]

                result.returns_6.iloc[j]=f

                

                



    fig, ax = plt.subplots(figsize=(20,5))

    ax.plot(result.window_6,'b+', result.window_5,'b+',result.window_4,'b+', result.window_3,'b+', result.window_2,'g+',result.window_1,'r+')

    

    ax.set(xlabel='time', ylabel='best_windows= 1. red, 2. green, 3-6 blue',

           title= symbl_1 + ' vs '+ symbl_2 )

    ax.grid()



    fig.savefig("test.png")

    plt.show()

    

    fig, ax = plt.subplots(figsize=(20,5))

    ax.plot(result.returns_6,'b+', result.returns_5,'b+', result.returns_4,'b+',result.returns_3,'b+', result.returns_2,'g+', result.returns_1,'r+')

    

    ax.set(xlabel='time', ylabel='return for best_window =  1. red, 2. green, 3-6 blue',

           title= symbl_1 + ' vs '+ symbl_2 )

    ax.grid()

    

    #fig.savefig("test.png")

    plt.show()







# import data from yahoo

data = yf.download(symbols_to_use,  "2008-01-01", "2021-11-01")['Adj Close']



### testing the best window to caluculate percentage change of the two pairs



timing_test(data,'DBB','UUP',2,800)

timing_test(data,'XLY','XLP',2,800)

timing_test(data,'SLV','GLD',2,800)

timing_test(data,'XLI','XLU',2,800)

timing_test(data,'FXA','FXF',2,800)

timing_test(data,'FXA','FXY',2,800)

timing_test(data,'FXC','FXF',2,800)

timing_test(data,'FXC','FXY',2,800)





### running a rolling window for retuns and selecting the best performing windows during that time



timing_test_2(data,250,'DBB','UUP',20,400)

timing_test_2(data,250,'XLY','XLP',20,400)

timing_test_2(data,250,'SLV','GLD',20,400)

timing_test_2(data,250,'XLI','XLU',20,400)

timing_test_2(data,250,'FXA','FXF',20,400)

timing_test_2(data,250,'FXA','FXY',20,400)

timing_test_2(data,250,'FXC','FXF',20,400)

timing_test_2(data,250,'FXC','FXY',20,400)

Hi, Manoj Agarwala , apologies for not replying sooner. Yes I got that file, thanks so much for putting them together.  I dissected all of the relevant instruments and created an external data source. I'm still having a bit of an issue getting it to run. I invested about 9 to 10 hours into it last week, hopefully I'll get some more time next week to see if I can fix the issue. 

My issue is that I'm predominantly a C# developer. I've done that for 20 years. Python's probably a third language to me, so its slow going. Eventually, I did though at least get to be able to set breakpoints in local copy of lean in visual studio. Progress! Slow progress!

Carsten My guess would be (or maybe not) the window length is correlated to the overall volatility? hence the window length should be somehow adaptive to the market volatility?

Toma, thanks for sharing the edits and links. Looking forward to the ideas and articles you outlined above. The change over time concerning what is the optimal lookback period length is very interesting. Relatedly, the question you raise regarding whether to use only recent lookback periods vs very long periods is interesting and is likely to generate a robust debate. We had a bit of that debate in this thread too (e.g. one suggestion was that a strategy should ideally hold for extremely long periods, e.g. 100 years, to be considered valid etc.).

Manoj Agarwala, thanks for sharing the dataset that you have created, very nice work! 
Regarding your points: 
I never expected there to be an issue concerning dividends or splits. My understanding is that the price data is backward adjusted so that, for instance, stock splits are neutralized and do not affect the price signals that are used in the algo. So, my assumption is that, due to the adjustment, there is no effect/bias. Not sure whether you have observed something that made you believe that stock splits and dividends could have an effect (?).
The 5 years lookback, and data not being available before 2008, is a fair point. So, the algo backtest should actually be starting from 2013. I mean, the percentile approach still does what it’s supposed to do in the years before, i.e. it would be looking for the 5% extreme observations; however, of course, the sample of return observations would be smaller in the years before 2013 than in the following years. 
Regarding sensitivities, that’s correct: If you change the 5-year lookback period to 1 year and the extreme percentile from 5 to 1, the performance changes substantially. It should get you closer to the v7 performance and probably quite a bit worse, since I removed a few of the signals compared to the v7.
Regarding multi-year down markets: You also raise a fair point here incl. how the algo could be made more robust for these scenarios. Theoretically, the algo would also generate out signals in down markets, since extreme price drops occur also when the prices are continuously declining (the decline is never fully smooth). However, what is true is that there would be times when the algo would go in. I cannot tell how lucky the algo would be in terms of capturing some of the upsides in a constantly declining market; there would definitely be opportunities to capture upsides even in declining markets since these are never only in decline, i.e. there is always a bit of back and forth.

Carsten, nice work, thanks for sharing! Here is my thinking: I fully understand the beauty of rolling optimizations runs to derive and test ‘would-be-optimal’ parameter values. However, in (semi)-efficient financial markets where these is ‘no free lunch’, I doubt that we would ever find one optimal parameter value, or a very tight range of optimal values for that purpose. If this one optimal value (or range) existed, then there would be free lunch, i.e. I would use this parameter setting and make optimal returns every single year (!). In my view, this is unlikely to ever exist in financial markets. Therefore, the optimal parameter values are all over the place, since they have to be in somewhat efficient financial markets. So, I think that the only thing we can really hope for is a parameter value that is actually quite sub-optimal when evaluated for each individual year, but is something of a valid average that still leaves us with a bit of excess return. I do not have the final answer, but the three-month lookback period might just be this sub-optimal average value that leaves us with an excess return (?).

Jupyter Notebook to Assist with Manual Trading: 

I created a Jupyter Notebook that generates up-to-date signals in seconds instead of using signals produced from QuantConnect backtesting, which takes a bit of time. This will greatly assist you if you plan on live trading this strategy manually. It is located in my GitHub repo under In n Out Custom.ipynb

New Out-Signal Idea: Min-Variance Portfolio of Bonds 

We currently push all of our money into 20 Year Treasury Bond (TLT) when we are given an “out” signal. Instead of going all in on TLT, I want to buy a mixture of out-of-market ETF's like (TLT), 7-10 year (IEF), and Daily S&P Short (SPDN). To find the composition of these 3 “out-of-market” ETF's, find the Minimum-Variance portfolio fo the 3. 

I was able to reduce max drawdown by using min-variance portfolio on out-of-market ETF's 

Peter Guenther  I made minor changes to v8 and was able to improve the performance. Basically, I introduced frequent rebalancing.  Here is the original v8 performance

and here is the new v8 performance:

Hello Matthew! Thanks for sharing the Jupyter notebook, this is really useful for manual live trading.

Do you know why the recent in and out signals are different from the “official” v8 implementation that Peter posted back in October, and also your own version posted here? For example going back in on 2021-10-22 vs 2021-10-29 for your version posted here backtested in QuantConnect, and 2021-11-05 for the “official” v8 implementation…? I am not sure how much of a difference it makes at the end but thought it was worth asking the question and knowing the way you would use this notebook in your daily tradind.

Laurent Bonherbe The recent signals appear to be the same. I attached the screenshots of what I'm seeing. Could you point out what difference you're seeing on your end? 

I'd just run “Restart & Run All" on Jupyter notebook at the end of the trading day to get the latest signal and trade accordingly. If signal is in, be in QQQ. If signal is out, be in Bonds and maybe short Q's.

Matthew Wang I ran your in-out custome 2 here on QuantConnect, and those are the trade dates since 1st of Jan 2020 (putting the precise dates rather than the graph to be razor sharpe precise):

On the collab notebook we have:

OUT on 2020-02-25

IN on 2020-06-12

OUT on 2021-08-12

IN on 2021-10-22

OUT on 2021-11-24

IN on 2021-12-16

As you can see from the above it is directionally OK but we have differences between the 2, sometimes up to a week (2021-10-22 vs 2021-10-29 for the OUT… Maybe I am worrying about something I shouldn't worry about, but I would very much like to explain where those differences are coming from.

I tweaked some of the parameters (suc as smoothing period) and in/out comparison logic and was able to significantly improve. the recent years performance compared to te original v8. See year by year comparison below:

Wondering if we can combine the two and dynamically pick the best performing one?

Just what we all love, code with zero comments.

Hi @Matthew Wang (how do I tag you with a space in name?), loving the jupyter approach you've taken.

Have you tried a backtest using the Out variance with max sharpe ratio? What are the results looking like on that in comparison?

Does anyone have much experience with pandas to do a vector backtest right in the notebook?

And this is how these strategies blow up when both stocks and bonds collapse, i suggested a long long time ago adding an ultimate out of cash. Find my old replies for some detailed thoughts. 

You just need safety measures built into the portfolio management and trade execution portions of the lean framework.

That is true, Elsid Aliaj, I remember that you warned about the risk of equities and bonds moving in sync. i.e. both going down. Thanks again for this early warning. Holding cash is really a good idea these days.

If I recall correctly, Goldie Yalamanchi had posted a version of the algo where gold was mixed into the out holdings. In addition to your cash suggestion, this setup may also have worked well in the current context.
Goldie also warned about tech coming under pressure. The warning came quite a bit earlier than the actual decline, but was probably instrumental for some to remain vigilant. I think that Goldie's approach was to mix a tech-focused strategy (eg holding QQQ) and a value-based strategy (stock selection based on Quality Companies in an Uptrend).

I pulled this from one of my archived presentations circa 2017.  My thoughts on robust risk management that is would not be fitted to specific market conditions would be as follows:

If Risk-Off > Then run a 20 and 60-day correlation of MKT security compared to securities in a basket of possbile risk-off ETF securities (Long Bond, Short Bond, Gold, Energy, Yen, Swiss Franc, Commodities, etc), inversely rank the least correlated, check for positive trend over the same period and invest in the top N > If Risk-Off check correlation every N Days and rebalance.   One could add some momentum logic as well.  

Market regime changes can be very protracted and correlations are not static.   Buyer beware.  Using ETFs such as DBC, DBA or Currency ETFs can pose liquidity and volume issues if trading in size and market orders.  

Pardon me for poking my nose in this thread but distilling what I understand of the strategy it hinges on a negative coorelation between the market (represented by SPY) and long term treasury yields.   Would then not a simple correlation track between the two be enough of a “get to cash” signal to bail out?

I've cobbled together something similar for my algo by leveraging a sample posted in another thread which shows a pretty good signal in the June-July timeframe of this year.

Stole the code from here if anyone is interested.

Wouldn't dual momentum of say bonds : cash work as a better predictor than correlation? I saw some of the correlation stuff above and it doesn't seem like it can be a good enough predictor as it's bouncing all over the place.

I have a dual momentum strategy with bond mutual funds, that does well and has moved to cash pretty much all these times corona ect.
Seems like it would work as a good out signal here if cash > bond funds stop trading and move to cash.

Well I think the obvious answer is probably to do both and take a look.  My comments were based on the view that I see this whole strategy hinging on negative coorelation between bonds and stocks.  If that starts to break down the strategy may no longer hold weight.

I'm not saying you're version won't work or isn't better but it seems to provide decent signal for me.

New to this thread, I would imagine someone would have raised this question already. Why not using actual oil, gold, metal prices for a much longer backtesting period pre-2008?

Is anyone using this for live auto trading? 

sorry to hijack thread Vladimir how can i get in contact with you? Want to discuss some potential investment ideas. 

Based on current situation, what if stocks and bonds both fall? currently stocks are down and bonds are also down due to high inflation worries. in that case we will lose money if we invest in bonds right? 

Log Up I've been brining these issues up since Quantopian days so a few years nobody cares to listen, they just keep overfitting the strategy.

Also the OUT is completely arbitrary based on faulty correlations which break down. Hence a simpler normal dual momentum strategy works better. 

Because if Asset < Risk Free Rate move OUT

Out also has to have a bucket of bonds vs risk free rate of cash, no point in going to bonds when their absolute momentum is negative vs cash.

The strategy would therefore be:

If Asset > Risk Free Cash IN (in to whatever highest momentum asset or hold top 5 assets)

Because if Asset < Risk Free Rate move OUT (either bonds in a low rate env or cash in a rapidly rising rate env)

The SLV < GLD and all the sector non sense is based on faulty correlations that would break the strategy at some point in time along with OUT just being bonds another fault correlation.

Jack Pizza  I am not much experienced trader myself, but I do agree with the pitfalls of hard coded assumptions/correlations you mentioned. We have many periods in history where stocks and bonds both fell together, so felt should it accommodate to invest only if bonds are faring better than other available alternatives.

@Jack Pizza 

I had similar issues with the overfitted versions.

The question is, what is robust?

I totally agree, a dual momentum is quite robust.

Voltility should be, but there a some periods, where high volatility as well delivers good gains.

Tested as well first month / second month VIX futures. If you plug them into a Hidden Markov Model you don‘t need to specify a length component as the HMM figures it out. worked very well in the 2008 episod, performed poor in 2015/16 and performed excellent in 2020.

Than there is the good old price vs 200DMA, if one uses the volatility or ATR to decide if the price is significantly above or below the performance is not bad.

Some more thoughts On robust indicator?

Carsten it's basically a fools game based on luck, the thing that has shown semi promise is momentum if you can stomach 40-50% drawdowns which basically make it as good as all in on the stock market end of the day lol. 

But you see they love hiding that fact by showing monthly drawdowns because well everyone is using monthly drawdowns, instead of we purposely want to mislead you, as if you knew the truth you wouldn't give us your money to invest.

On here it just turns into Ego contests, of I have the most popular thread on Quantopian / qc. Dual momentum backtests are highly flawed too especially with bonds as the out. Even Cash as the out doesn't quantify risk, say you're out everything is honky dory and the dollar collapses 40% overnight…

They geniuses laugh and grin ohh that will never happen, like it hasn't 100 times over globally throughout human history. People's don't appreciate risk, and think it will never happen to them, until it does over and over. Then all the experts go and hide not to be found.

“The day after the U.S. dropped its second atomic bomb on Nagasaki, Japan closed down trading on the Japanese Exchange. Although Japan would surrender less than a month later, officially ending World War II, the Japanese Stock Exchange would remain closed for nearly four years, finally reopening on April 1, 1949. Following reorganization, the TSE emerged as the largest of five post-war exchanges in Japan.”

Ohh it can never happen in their brains, what would you do if your'e invested and the stock market is closed for 4 years, if not totally bankrupt along with the companies or index you had invested in? Even excluding war there is so much counter party risk its insane, like dominoes they'll all fall. 

What happens with the DTC Cede and Co. which is technically the rightful owner of almost all Assets in the world since it's their name on the stock certificates and titles, says we own everything sue us in court if you want it?

See where I'm going with this? Every single strategy i've personally tested researched read about, none of them achieve anything above 1:1 performance/dd long term. You'll always have inevitable drawdowns even making the ratio 1:2.

My conclusion from all this is you can't escape drawdowns from a strategy. 

My further conclusions from a risk standpoint as far as investing goes is.

Operating Farmland, war happens your farm blows up who cares its just land, replant seed after. Weather ruins crops for a year, again replant next year. Water issues, use new technologies like desalination. Everyone will always need food….

Gold as basic insurance, if you have some sort of trading strategy, trade it through actual futures where you use physical gold as collateral on CME, but you at all times are physically owning it. Internet goes down, CME goes bankrupt, war, yada yada who cares, you have no counter party risk as you physically own it somewhere, after the dust settles it's a long time. 

Even gold just because it's worked for human history, there could be some time in the future from say asteroid mining where it becomes basically worthless, or some massive discovery. 

I would hold other non food type commodities mostly metals physically and like i said maybe run some momentum strategy if you can stomach the massive drawdowns. Assuming momentum is a psychological phenomena and not another overfitted mined hypothesis. 

Hope this confused you a lot but hopefully it helps you from losing a crap ton of money…. I wish i knew what a robust indicator was to limit drawdowns and maximize performance, but I have yet to find it unfortunately. 

So invest with a barbell type Taleb mindset, 95% in farmland, gold, 5% gamble on crypto, high flying strategies, if they last a year or two and you make 1000% great take it out put it on the barbell, and repeat. 

Always remember there is a lot more risk like 100000000% more than majority of people care to even bother thinking about. 

hope this somehow helped lol.

A comment on the current overfitting discussion
It is an important part of the In & Out discussion’s evolution that we constantly revisit the overfitting issue. The latest resurgence of the issue was mainly driven by an underperformance of the TLT ETF which is used in the latest in & out versions as the out holdings when we are not in, i.e. not invested in equities. Some comments, as a shorthand, simply seem to refer to the previous In & Out versions as the ‘overfitted versions’, with the TLT underperformance just being an additional proof that the In & Out has been overfitted as claimed in earlier comments in this thread. However, this conclusion does not seem to be valid to me.

Overfitting of what?
In the Amazing returns in & out thread, I try to explicitly pinpoint the three key components/characteristics of the in & out algos:
1. The in holdings selection
2. The out holdings selection
3. The market timer property

Recall that earlier overfitting discussions were mainly about the number of parameters that the In & Out uses and that these are too many and clearly show overfitting. Note that these parameters are neither used for the in holdings selection nor the out holdings selection, the latter of which has led to the current underperformance (the TLT decline). Instead, these parameters are used for the market timer. In the Amazing returns in & out thread, I argue that the market timer has worked remarkably well during the recent equity downturn. The market timer is arguably the soul of the In & Out. Therefore, it appears to me that the recent episode is not very suitable to demonstrate overfitting of the algo along the lines of earlier claims in this thread.

Have the out holdings been overfitted? First, it is important to note that this question has little to do with the previous overfitting discussion. It is a new question since it is about a different component. Second, in my view, overfitting is a process in which models are repeatedly estimated and optimized to fit the data better (i.e., to maximize the total return). Going through the discussion in the in & out threads, I see little evidence for such an approach on the out holdings. It seems to me that bonds were listed as the out holdings in my very first post on Quantopian. Since then this component has basically been ignored, almost neglected. There are no parameters dedicated to this component other than a hard-wired ETF that is listed. I do not recall significant efforts to (over-)fit this component. Quite the opposite, the lack of such efforts has sporadically been noted in this and the other thread. 

It seems that the issues we are seeing at the moment are a result of underfitting or lack of fitting, instead of overfitting. Specifically, the out holdings selection has been ignored and requires work.

Hey Peter Guenther actually the biggest performance issues seems like the timing model, it goes in right before the collapse. You can easily just swap in BIL (cash) instead of TLT, and see it takes a massive hit because it's IN instead of OUT. 

Trying some tests with an MA filter and stop losses.

nevermind had a wrong version seems like, I tried adding an MA filter doesn't seem to be working though… added to line 137. 

Also add momentum approach mentioned in the stock selection forum. Would add it to both the IN and OUT portion so have maybe 4 ETF's in IN and 3 in OUT and select highest momentum one

I would add it from the other momentum algorithms, but the ordering is a bit weird and i can't debug easily on here, would probably take the symbol and concat / append it to _IN and _OUT after the momentum check is ran. I'll try working on it later, but maybe if someone has time can do it much quicker. 

Momentum would be 1,3,6,12 month with weights of 12,4,2,1

A simple stop loss protects losses as opposed to cash. straight cash you are at around 16%/16% Returns/Drawdown, with iEF + stop loss 1%, you're at 20%/!8%

Manoj Agarwala  , i have been trying to add momentum indicators on your version without success….  how would you add moving average ? 
 

Zimman  

 self.spy = self.AddEquity("QQQ", Resolution.Daily).Symbol  
       

self.sma = self.SMA(self.spy, 50, Resolution.Daily)

Zimman  

this attached backtest has a lot of momentum stuff, just have to port it over

you use a simple price above below like this: 

if self.Securities[self.spy].Close > self.sma.Current.Value:

thank you, Jack

Some recent learnings i had analyzing and extending the IntOut:

• it might be prudent to not trade the bonds while inflation is high (see chart)
• 20 vs. 30 yield inversion (see my example earlier) improves results slightly in 2022 and does not cost performance and previous regimes
• I changed QQQ to SPY for self.MRKT - this leads to more stable performance accross 2002 - 2022

self.MRKT = self.AddEquity('SPY', res).Symbol 

i think that change from QQQ to SPY makes sense. Its true that QQQ has done really well in the recent past, but then again its fairly concentrated. SPY is much more diversified since it contains banks and energy companies. So even if the switching from QQQ to SPY makes the best test worse I wouldnt trade QQQ and instead trade SPY, since I view the backtest with some level of uncertainty about the future. 

Secondly about the bonds, I think its not a good idea to come up with some ideas to change the strategy in response to the recent losses. Its not clear to me that high inflation will always lead to loss in bonds. It depends on market expectations and whats priced. making changes in response to losses often backfires. My suggestion is to keep the code as it is and accept the loss and hope for better times ahead.

Narendra Kulkarni the code was based on faulty assumptions of bonds and equities always having a negative correlation…. why would you want to trade something based on obvious fallacies…..

Santa tried to fix it with bonds inverting, and easier way like the few algorithms above is you add cash and check momentum against bonds, if cash is performing > bonds move to cash.

I agree with Jack Pizza . Jack Pizza , can you post your favorite example that adds the bond momentum vs cash logic? 

@Narendra, the choice between trading QQQ or SPY should always favor QQQ. This is a decision that could have been made even 10 years ago. In fact, it could have been made from inception.

SPY holds the 500 most valued stocks, and QQQ holds the highest 100 from the same group, technically ordered by value. The 400 stocks of SPY which are not in QQQ by definition have to be valued less than the lowest-valued stock in QQQ. Those 400 stocks will therefore be a drag on SPY's average long-term performance.

A simple chart can show this.
 

Narendra Kulkarni i didnt trade the SPY instead of QQQ, i changed the self.MRKT signal from QQQ to SPY that is used to calculate the outlier percentile return vectors for the In/Out signal

@Santa24, using SPY as signal generator should not be a better alternative to QQQ, again by construction.

SPY is averaging over 500 stocks while QQQ is doing the same on the highest 100.

The result: SPY's price movement is less volatile as can be seen in the provided chart above. This will have for average effect to delay trading signals. The other 400 SPY stocks lower SPY's average performance compared to QQQ. It also dampens average price movements also as shown in the above chart.

Trading implies exploiting price variations. The lower the average price variations, the lower the average profit potential. 

Quantopian archives can be found here. 

In/out with relative momentum attached.  This combines the in/out trigger with Bold Asset Allocation relative momentum (Bold Asset Allocation (BAA) - Keller by Tristan F - QuantConnect.com).  The relative momentum is likely suboptimal.  For example, it uses end of month values to determine the SMA, which doesn't make sense for a strategy that can trigger on a daily basis.

The offensive (in) and defensive (out) asset universes both have cash (as BIL), and the offensive includes consumer staples (XLP), which is a good diversifier to QQQ.  Results are not as stellar as just QQQ and TLT up to 2021; but the relative momentum and diversification of cash and consumer staples avoids the 30% drawdown in 2022.

If you chose to build on this, please share any improvements.

Tristan F this is a better way than some of the other modifications of adding momentum, question is how can we add an SMA filter to trading in? It seems the trading logic is lumped for both in / out with the weights logic.

Not sure if it would effect performance or not, but from my prior testing it helped a bit with drawdowns, even though Gary Antonacci says there is no need for a MA as price momentum itself is basically the MA, I've found that it skews risk a bit when adding an MA filter to trading long.

Lastly I guess the debate is why is in / out better at squeezing performance than just straight dual momentum type of strategies, and how can we backtest this say 100 years to really check it's robustness, or if it's just overfitted the last 10-20 years. I haven't been able to really find any research papers on extreme price moves and their accuracy. It does make sense as a hypothesis of an extreme move meaning something broke and caution. 

Guy Fleury , I believe you are incorrect about the differences between SPY and QQQ. https://www.invesco.com/us/financial-products/etfs/holdings?audienceType=Investor&ticker=QQQ

QQQ is not simply the top 100 companies. Its taking a more concentrated position in tech. SPY by construction is more diversified. It contains 500 largest companies. NASDAQ by construction ignores banks and energy companies. I know that last 12 years have great for tech, but there is no fundamental reason to believe that tech will always do better in the future. SPY is more diversified and therefore I would argue that its safer investment to hold. But if you do chose QQQ do this with your eyes wide open and understand the risks.

Other arguement that you have made is that since SPY has 400 more smaller market cap stocks therefore it will always underperform. This is patently false. Just because a company has a lower market cap does not imply it will have lower growth. In fact I would argue that smaller companies are more likelier to grow. 

But I will accept that by construction QQQ is more likely to have higher volatility. But I am not totally sure that QQQ vs TLT is a better bet to make than SPY vs TLT. I know that backtest will show that QQQ does great but I will argue thats its just luck and not by construction.

Tristan F also says defensive momentum is set to 12 months, but self.TD shows 1? is it set to 1 month or 12? Trying to figure out why your version goes into TLT during 2022, whereas the last version i posted stays in BIL and goes into QQQ for one day or so. 

Jack Pizza ,

This is a better way than some of the other modifications of adding momentum, question is how can we add an SMA filter to trading in? It seems the trading logic is lumped for both in / out with the weights logic.

I'm not sure I understand.  Are you suggesting determining momentum using a low pass filter (fast SMA) instead of the latest price?  For example, check that the average of the last 2 weeks is trending higher over the last 6 monhts:  SMA(10) / SMA (126), instead of the last price P(-1)/SMA(126)?

defensive momentum is set to 12 months, but self.TD shows 1? is it set to 1 month or 12?

The paper from Keller (here:  https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4166845) has all the definitions.  LO is lookback on offensive, and LD on defensive (6 and 12 months respectively in what was shared above); TO is the number of offensive assets to choose that are trending up (equal weight), same with TD for defensive (1 for each in this case).

Tristan F No saying use a an additional broad filter such as if SPY <> SOME MA TRADE LONG when IN. So if IN first check if overall market is above MA or else don't place any long trades. Then run the typical momentum calculations if it's ok to trade.

Ok let me retest with different lookback periods, from past testing and what we've witnessed I believe we want quicker lookback periods for defensive given if interest rates go up rapidly instead of a typical .25 it happens rapidly a 12 mo lookback is a bit too slow to probably react quickly enough. 

Another option is probably manually just removing all bonds from the strategy during rapid rising interest rates like now, I did a rough study of interest rates, when they rise in orderly fashion say .25-.50 there isn't massive hits to bonds, when they rise rapidly like now bonds take big hits.  Or maybe we can use Santa24 bond inversion to just use CASH and remove any bond funds for defensive until yields stabilize. 

Jack Pizza , I see.  The attached has an additional SMA check that can flip an “in” flag to “out".  Performance is much worse.  This is not too surprising given we are overriding a key value proposition of this strategy: getting back in quickly after a tail event.

See self.ma_eq, self.ma_prd variables that control what you were asking for.

Tristan F yeah that didn't work to well not that it added changed much in the other version i had just a tiny bit, this just made it brutal lol