Welcome Stockaholics!

We are a new and fast growing financial forum! Sign up for free and let's talk stocks!

  1. Do you want to help develop this community? We are looking for contributions from investors and traders like you! What stocks do you follow? What is hot right now? Sign up and get in on the ground floor of the newest, fastest growing financial forum!
    Dismiss Notice
  2. You will notice a live chat widget on the right. Click in to join us and lets hear about how you nailed that last UWTI trade!
    Dismiss Notice

Stoch's own mutual fund

Discussion in 'Trade Journals' started by Stoch, May 4, 2020.

  1. Stoch

    Stoch Well-Known Member

    Joined:
    Jan 5, 2019
    Messages:
    620
    Likes Received:
    305
    Having closed my business and becoming partially retired, I transferred my retirement account from Invesco (which has a very limited selection of funds) to my TD Ameritrade rollover IRA. I have been using a neural net software and previously tried several indicators, A/D. ATX, CCI, MACD, SMA, EMA, linear regressions (including the stochastics which I used for awhile but felt it didn't work as well as I would have liked). I was hoping to create an automated trading algo that would give buy and sell signals for a large number of stocks I selected, creating a trend following system using several moving averages and profit targets and stops. That way I would only have to look at the signals instead of reviewing every chart every day. I'm hoping this blog can help my put the trading rules down on paper and develop the system.

    I have been using NeuroShell Trader daily close software for may years. The programs are optimized to each security and gave great back-testing results, but the walk forward results weren't even close to those results. I have the results for a momentum strategy that was optimized for a 92% annual return but only generated a 2.2% annual return this year so far. The optimization selects and optimizes the inputs for each security and the trading signals are generated by a predicted result in the future by the neural net.

    PREDICTION

    AAAMomentum Chart.cht
    (ALL INSTRUMENTS AVERAGE) ((AVG))
    Optimization
    Percent Change in Open 10 trading days into the future from the next open

    INPUT CONTRIBUTION ANALYSIS

    # Contribution Input
    1 14.10 % AccumDist(High,Low,Close,5)
    2 5.90 % ADX(High,Low,Close,10,5)
    3 13.77 % ATR(High,Low,Close,14)
    4 12.59 % CCI(High,Low,Close,5,0.015)
    5 4.14 % DMI(High,Low,Close,10)
    6 21.49 % MACD(Close,12,26)
    7 23.12 % MACD Signal(Close,9,12,26)
    8 11.69 % RMI(Close,5,5)
    9 24.82 % RSI(Close,5)
    10 26.88 % Stoch%D(High,Low,Close,5,5)
    11 14.77 % Stoch%K(High,Low,Close,5)
    12 16.29 % StochSlow%D(High,Low,Close,5,5,10)
    13 26.80 % %R(High,Low,Close,10)

    PREDICTION OUTPUT

    Percent Change in Open 10 trading days into the future from the next open
    AAAMomentum Chart.cht
    Annual total return Dates
    (ALL INSTRUMENTS AVERAGE) (AVG) Optimization 92.4% 163.7% 1/3/2000 4/17/2020
    (ALL INSTRUMENTS AVERAGE) (AVG) Trading 2.2% 0.6% 12/28/2018 5/4/2020



    I decided to try giving the neural net a few moving averages to work with and adding stops and targets for a 30 day prediction. I trained it through the decline in February and March to make sure it could see a decent correction beside 2000-2003 and 2008 and adjust stops etc. The actual rules it uses are never known.
    .
    PREDICTION

    30 day stock prediction.cht
    (ALL INSTRUMENTS AVERAGE) ((AVG))
    Optimization
    Percent Change in Open 30 trading days into the future from the next open
    5/4/2020 9:03:08 PM


    INPUT CONTRIBUTION ANALYSIS

    # Contribution Input
    1 43.32 % ExpAvg(Time Series,ExpAvg Periods) 1 to 15 days
    2 39.15 % Avg(Time Series,Avg Periods) 1 to 15 days
    3 30.61 % Avg(Time Series,Avg Periods) 15 to 30 days
    4 29.27 % Avg(Time Series,Avg Periods) 30 to 80 days
    5 15.23 % Avg(Time Series,Avg Periods) 150 to 250 days
    6 19.55 % Close
    7 11.88 % TrailPrice%(Trading Strategy,Percent Trailing) 0.1% to 10% stop
    8 1.48 % PriceTarget%(Trading Strategy,Target percent change) 5% to 200% target

    PREDICTION OUTPUT

    Annual total return Dates
    (ALL INSTRUMENTS AVERAGE) (AVG) Optimization 179.2% 3157.6% 1/3/2000 3/31/2020
    (ALL INSTRUMENTS AVERAGE) (AVG) Trading 77.2% 8.6% 3/17/2020 5/4/2020

    The optimization for my selection of stocks and bonds was 179% annual return for 20 years (if you were a perfect trader) with a 6 week walk forward of 8.6% since March 17th, since it lagged the rally after the drop while the S&P was up 12.1%), I set it for a long only position and didn't include any bear funds in the mix.
    I decided to start allocating my funds based on it's recommendations.

    Since this is a retirement account, I decided to start with a 40%/60% equity/bond ratio. I identified 11 bond funds (TLT, IEF, SHV, SHY, and 6 other vanguard corporate and government bond funds of different duration) and Gold.

    I have been charting on 84 different equities I have on Amibroker, the charting software. Most are the usual large cap, with tech, utilities, Reits, consumer staples and discretionary, retail and defense stocks. I noticed I really haven't been following much in the industrial or banking space.

    Position sizing is set for 0.75% per full equity position and 7% per bond fund. As a second check, I will also use a simple stop and reverse(SAR) signal to limit positions to 1/2 size if not confirmed as up trending by the SAR.

    So I started on May 1st with my experiment with 53 long equities and 8 bond funds. The initial distribution is 29% bonds, 30% equities and 39% cash

    I will update the portfolio daily as needed and track the results. This will give me an objective trigger and remove the doubt and uncertainty that goes with selecting and closing positions. With so may positions, I'm sure it will seem to churn, luckily TD has eliminated most brokerage fees for active trading. Also with such a large bond and cash position, there is no way to keep up with the S&P in an up market, but it should easily beat it in a declining market. I only have to beat the 1.55% return from the online savings account anyhow..So after 2 days of trading...

    5/1 account S&P 290.63
    5/3 account (-0.4%) S&P 283.75 (-2.4%)

    Trades for tomorrow are closing J&J and Berkshire B and reducing position sizes in McCormick, Home Depot, Goldman and Disney due to SAR reversals after this mornings drop triggered them
     
  2. Syynik

    Syynik Active Member

    Joined:
    Aug 7, 2016
    Messages:
    152
    Likes Received:
    65
    The two you are closing are two I'm hoping to buy. And DIS? I'm watching DIS for an entry as well.
     
  3. Stoch

    Stoch Well-Known Member

    Joined:
    Jan 5, 2019
    Messages:
    620
    Likes Received:
    305
    It entered JNJ at 118.91 on 3/25 and rode it to 148.24, a 24% gain, I suspect its just booking profits on the trailing stop. BRK.B went long on 4/7 at 191.99 and out now at 177.95(-7%), probably a protective stop after yesterdays 2.4% drop.
     
  4. Stoch

    Stoch Well-Known Member

    Joined:
    Jan 5, 2019
    Messages:
    620
    Likes Received:
    305
    The first week of data is in and the model seems to be doing reasonably well. The high bond and cash percentages reduced the drawdown on the drop Monday (-0.8% vs -2.6% on the S&P) and the end result was close to the S&P for the week (0.6% vs 0.7%). The week ended 32% in bonds, 35% equities, and 32% in cash. The model is long in 49 positions out of 87 equities but only 47% are full positions, down from 54% last week as the majority are below the SAR..
    The model if up 9.9% since 3/17.

    PREDICTION OUTPUT 1 yr est return on account
    (ALL INSTRUMENTS AVERAGE) (AVG) Trading 80.5% 9.9% 3/17/2020 5/8/2020


    Trading strategy S&P buy closed
    04/30/20 290.48
    05/01/20 -0.8% 282.79 -2.6% Amzn,duk,vpu,vz sold gd(-5.0%)
    05/04/20 -0.4% 283.57 -2.4% cl
    05/05/20 -0.1% 286.19 -1.5% vclt sold brk,b(-6.2%), jnj (+25.3%)
    05/06/20 -0.5% 284.34 -2.1% dis
    05/07/20 0.0% 287.68 -1.0% bep sold dedg, g(+24.3%), yum(+13.9%)
    05/08/20 0.6% 292.44 0.7% sold gis(-2.2%)

    Trades for Monday include getting back into Brk.b, adding utility PPL, and selling FB after a 33% gain, which must be a profit target as the stock didnt pull back Friday
     
    #4 Stoch, May 8, 2020
    Last edited: May 9, 2020
  5. Stoch

    Stoch Well-Known Member

    Joined:
    Jan 5, 2019
    Messages:
    620
    Likes Received:
    305
    The model is almost fully invested now, This week I also added metals to the mix, GLD, SLV, PPLT, PALL
    It is long 48 equities out of 84 but the SAR decreased to 29% positive of the long positions so overall equities position size declined for the week.
    Bond and metals SAR increased as 75% positive and the position size increased 28% to full investment in non equities

    The model lost -0.5% this week and is up 0.1% for the month while the S&P was down -2.1% and -1.4% for the month
    I still suspect it will lag on advances but outperform in declines.

    The distribution is currently
    60.7% bonds and metals
    32.2% equities
    7.1% cash

    PREDICTION OUTPUT 1 yr est return on account
    (ALL INSTRUMENTS AVERAGE) (AVG) Trading 87.2% 10.5% 3/30/2020 5/15/2020

    S&P (month to date)
    05/08//20 0.6% 292.5 0.7% bought brk.b, ppl Sold fb (+32.7%)
    05/11/20 0.6% 290.34 0.0% bought vcit
    05/12/20 -0.1% 286.67 -1.3%
    05/13/20 -0.5% 281.49 -3.1% bought are
    05/14/20 -0.1% 284.97 -1.9%
    05/15/20 0.1% 286.28 -1.4% sold so (+1.2%)
     
  6. Stoch

    Stoch Well-Known Member

    Joined:
    Jan 5, 2019
    Messages:
    620
    Likes Received:
    305
    A slight adjustment down in the bonds and up in the equities this week. The portfolio is up 1.4% vs the S&P 1.7% but with 1/3 the drawdown of the S&P so far. I added a few airlines and cruise lines but the model passed on most of them, only adding only carnival and Southwest..

    Week #7 has the following distributions
    53.8% bonds and metals
    33.3% equities
    12.9% cash

    long 51 stocks, 8 bond funds and 3 metals

    PREDICTION OUTPUT
    (ALL INSTRUMENTS AVERAGE) (AVG) Trading 67.1% 10.6% 3/17/2020 5/22/2020

    Portfolio S&P
    05/08//20 0.6% 292.5 0.7%
    05/15/20 0.1% 286.28 -1.4%
    05/22/20 1.4% 295.44 1.7%

    I'm reconsidering the equity sizing calculation. Right now they are modified 50% by the current daily SAR, but I'm thinking of adding an adjustment for the year to date and quarter to date relative strengths to the S&P, as well as an overall market indication using Dr Wish's general market indicator to reduce positions when the overall market is less favorable. This would result in smaller 25% position adjustments instead of the current 50% changes. The same would be true for the bond funds and metals, using the Vanguard total bond fund BND for relative strength.
     
    #6 Stoch, May 23, 2020
    Last edited: May 24, 2020
  7. Stoch

    Stoch Well-Known Member

    Joined:
    Jan 5, 2019
    Messages:
    620
    Likes Received:
    305
    The first month of data is in and the model is 48% bonds and metals, 36% equities and 15% cash. It is long 51 equities, 8 bond funds and 3 metals. I would consider that almost fully invested. For the month my implementation of the model was up 2.7% vs the S&P gain of 4.7%. Not unexpected due to the large bond investment but the drawdown was almost 50% less as well.

    The equally weighted model is up an average of 12.1% annually since it started trading 3/17. it was up from 8.6 to 12.1 this month, a 3.5% gain for May. If we compare that to 60% BND and 40% SPY, we would have expected a 2.2% gain for May.

    PREDICTION OUTPUT
    (ALL INSTRUMENTS AVERAGE) (AVG) Trading 68.9% 12.1% 3/17/2020 5/29/2020

    I would expect a sideways too down movement of equities for the summer season, so I'm looking forward to seeing the performance in a choppy market, but the reason I started this experiment was to take out my expectations and trade a purely mechanical system, to remove the emotion from my trading decisions.
     
  8. Stoch

    Stoch Well-Known Member

    Joined:
    Jan 5, 2019
    Messages:
    620
    Likes Received:
    305
    The model has shifted more assets to cash this week. Interesting that the model has decided to close several profitable equity positions Friday and wants to add to the bond position. CCL(+47%), FDX(+17%), GD(+17), MCD(+14%), PSX(+71%) are are going to be liquidated Monday and I will add a 75% position on IEF, the 7-10 yr T Bond.

    Portfolio positions
    bonds 39% (-9%)
    equity 34% (-1%)
    cash 27% (+10%)

    PREDICTION OUTPUT
    (ALL INSTRUMENTS AVERAGE) (AVG) Trading 81.9% 15.5% 3/17/2020 6/5/2020

    since 5/1/20
    Model portfolio 4.8%
    SPY 9.9%
    60%/40% model 3.8%
     
  9. Stoch

    Stoch Well-Known Member

    Joined:
    Jan 5, 2019
    Messages:
    620
    Likes Received:
    305
    The model outperformed thus week as it continued to decrease equities and move into bonds and cash. It wants to move more into IEF bonds, and add defensive equity Coke KO and transport Fed EX FDX, while selling McCormick MKC and Marathon MPC on Monday. Its still holding 1/2 positions on CCL and LUV, but sold all the UAL Wednesday, Now up to 103 equities for the scanner by adding another 11 stocks including railroads, casinos, hotels and a software company VMC ware.

    Portfolio positions
    bonds 44% (+5%)
    equity 20% (-14%)
    cash 35% (+8%)

    PREDICTION OUTPUT
    (ALL INSTRUMENTS AVERAGE) (AVG) Trading 67.3% 13.0% 3/17/2020 6/12/2020

    since 5/1/20
    Model portfolio 4.0%(-0.8% this week)
    SPY 4.7%(-5.2% this week)
    60%/40% model 2.1%(-1.7% this week)
     
    #9 Stoch, Jun 13, 2020
    Last edited: Jun 20, 2020
  10. Stoch

    Stoch Well-Known Member

    Joined:
    Jan 5, 2019
    Messages:
    620
    Likes Received:
    305
    The model added just a little equity exposure this week, but added even more to the bonds and metals. It wants to go back into KO again after taking a 7% gain earlier in the week and has given up on the utilities SPDR VPU.with a 1.25% gain since May 1,

    PREDICTION OUTPUT
    (ALL INSTRUMENTS AVERAGE) (AVG) Trading 64.1% 13.7% 3/17/2020 6/19/2020


    Portfolio positions
    bonds 54% (+10%)
    equity 22% (+2%)
    cash 24% (-9%)

    since 5/1/20
    Model portfolio 4.9%(+0.9% this week)
    SPY 6.3%(+1.6% this week)
    60%/40% model 2.8%(+0.7% this week)
     
  11. Stoch

    Stoch Well-Known Member

    Joined:
    Jan 5, 2019
    Messages:
    620
    Likes Received:
    305
    We got a look at the performance in a down week and as expected by the large bond and cash allocations, it limited draw-down and handily beat both the S&P and the 60/40 model. It plans to add utility DUK and renter VMW after taking the profit on the Dell news pop. It also adds IEF to the bond portfolio. It has had enough pain with UAL and will liquidate the position Monday after a 10% loss.

    PREDICTION OUTPUT
    (ALL INSTRUMENTS AVERAGE) (AVG) Trading 52.0% 12.3% $ 80.42017 3/17/2020 6/26/2020


    Portfolio positions
    bonds and commodities 47% (-7%)
    equity 21% (-1%)
    cash 31% (+7%)

    since 5/1/20
    Model portfolio 4.7%(-0.2% this week)
    SPY 3.3%(-3% this week)
    60%/40% model 1.8%(-1.0% this week)
     
  12. Stoch

    Stoch Well-Known Member

    Joined:
    Jan 5, 2019
    Messages:
    620
    Likes Received:
    305
    The model fell behind the S&P and the 60/40 model with its defensive posture after last weeks correction. It did start moving out of cash as the week went on. I added 2 index funds and 2 commodities fund to the screening list. The model went long on VMW again, and added IVOO mid caps and LGLV, the large cap low volatility fund. It passed on the soft commodities and livestock fund.

    Portfolio positions
    bonds and commodities 53% (+5%)
    equity 24% (+2%)
    cash 26% (-7%)

    PREDICTION OUTPUT
    (ALL INSTRUMENTS AVERAGE) (AVG) Trading 55.0% 13.8% 3/17/2020 7/2/2020

    since 5/1/20
    Model portfolio 4.9%(+0.2% this week)
    SPY 7.5%(+4.2% this week)
    60%/40% model 3.5%(+1.7% this week)
     
  13. Stoch

    Stoch Well-Known Member

    Joined:
    Jan 5, 2019
    Messages:
    620
    Likes Received:
    305
    The model continue to deploy cash as it moved out of the defensive posture it took after the drop at the end of June. It is almost fully invested at a 60/40 ratio, It added to existing equities rather than add new positions. I was a bit surprised when it sold NVDA this week, but it had a 40% gain since entry April 21 and I guess it hit the profit target.

    Portfolio positions
    bonds and commodities 57% (+4%)
    equity 37% (+14%)
    cash 8% (-18%)

    PREDICTION OUTPUT
    (ALL INSTRUMENTS AVERAGE) (AVG) Trading 52.4% 14.4% 3/17/2020 7/10/2020

    since 5/1/20
    Model portfolio 6.1%(+1.2% this week)
    SPY 9.4%(+1.9% this week)
    60%/40% model 4.4%(+0.9% this week)
     
  14. Stoch

    Stoch Well-Known Member

    Joined:
    Jan 5, 2019
    Messages:
    620
    Likes Received:
    305
    The model held most positions at an almost 60/40 ratio last week. It wants to add GD, V, and SO while taking a 6% profit on VPU, the utility ETF after a good day Friday. Last weeks NVDA sale looks like a good move this week.

    Portfolio positions
    bonds and commodities 58% (+1%)
    equity 38% (+1%)
    cash 6% (-2%)

    PREDICTION OUTPUT
    (ALL INSTRUMENTS AVERAGE) (AVG) Trading 54.9% 16.0% 3/17/2020 7/17/2020

    since 5/1/20
    Model portfolio 11.8%(+0.9% this week)
    SPY 10.8%(+1.4% this week)
    60%/40% model 5.2%(+0.8% this week)

    I'm going to simplify things this week, consolidating 4 Bond ETFs into the Vanguard Government funds which have 1/3 the management costs of the iShares funds, Also the SAR and technical adjustment to equities result in too many moves so I reduced the position sizes to just 3 levels depending on the relative strength to the S&P and BND funds,
     
    #14 Stoch, Jul 18, 2020
    Last edited: Jul 25, 2020
  15. BermudianOption

    BermudianOption Well-Known Member

    Joined:
    Apr 6, 2016
    Messages:
    381
    Likes Received:
    253
    Does the model typically outperform or underperform relative to the S&P500?

    I read an article on Bloomberg that touched on the idea that the most benchmarks have a lot of their weighting behind the FAANGs and they outperform which leads the benchmark to be harder to actually beat.
     
  16. Stoch

    Stoch Well-Known Member

    Joined:
    Jan 5, 2019
    Messages:
    620
    Likes Received:
    305
    An interesting question. For the training period of 1/3/200 to 6/1/2020 the S&P went from 146 to 304. a gain of 1080%, the neural net model returned 2950%, (ALL INSTRUMENTS AVERAGE) (AVG) Optimization 170.3% 2950.0% 1/3/2000 6/1/2020

    However that was a training period and the neural net curve fits for the best return, so you need forward testing to validate the model. Since it started live trading in March it has returned 16% and the S&P is up 33%, so you would say it under-performed the S&P by half, but what it improved is on limiting the drawdown in a down market, so I need more data to evaluate its performance after a period of down movement.

    The way I use the model is to use its inputs maintain up to a 40% equity position, with the other 60% in the bonds and commodities positions. That will also reduce my performance in up markets, but add additional limits for drawdowns since at my age, I cant afford anymore 40% haircuts like 2000 and 2007. A slow and steady return would be fine, even if less than the S&P, as long as the downside is limited.
     
    #16 Stoch, Jul 25, 2020
    Last edited: Jul 25, 2020
  17. Stoch

    Stoch Well-Known Member

    Joined:
    Jan 5, 2019
    Messages:
    620
    Likes Received:
    305
    The model decreased its bond positions while maintaining its commodities and deployed some into equities while raising cash. It scored really big on the gold, silver and platinum positions and easily outperformed the S&P on a down week.

    Portfolio positions
    bonds and commodities 38% (-18%)
    equity 44% (+6%)
    cash 18% (+12%)

    PREDICTION OUTPUT
    (ALL INSTRUMENTS AVERAGE) (AVG) Trading 51.2% 16.0% 3/17/2020 7/24/2020

    since 5/1/20
    Model portfolio 8.8%(+2.4% this week)
    SPY 10.5%(-0.3% this week)
    60%/40% model 5.3%(+0.1% this week)
     
    #17 Stoch, Jul 25, 2020
    Last edited: Jul 25, 2020
  18. Stoch

    Stoch Well-Known Member

    Joined:
    Jan 5, 2019
    Messages:
    620
    Likes Received:
    305
    Three months have been completed and the model has a slight decrease this week, the third since June. It was up 10/13 weeks and currently stands up 15.8% since starting trading.

    Portfolio positions
    bonds and commodities 50% (+12%)
    equity 40% (-4%)
    cash 10% (-8%)

    PREDICTION OUTPUT
    (ALL INSTRUMENTS AVERAGE) (AVG) Trading 47.6% 15.8% 3/17/2020 7/31/2020

    since 5/1/20
    Model portfolio 9.1%(+0.3% this week)
    SPY 12.4%(+1.9% this week)
    60%/40% model 6.3%(+1.0% this week)
    40%/60% model 8.3%(+1.3% this week)

    Moving forward, I will continue 1 account with the model recommendations, but reduce the commodity allocation due to the increased volatility of the Gld, Slv and Pall positions. In the second account, I plan to add an additional step of screening the recommendations by using a relative strength filter to the position, avoiding recommendations that are lagging the SPY.
     

Share This Page