Monday 24 February 2014

Week of 2014 FE 21

IWM rose by 1.3% this week, while my account rose 1.6%.  The loss-floor has risen to -2.9%, its highest value since last May.

US markets were closed on Monday for Presidents’ Day.

End of week allocations:

102% equities, 24% silver, 44% cash
% gain
size
Max loss
Sym  Buy  Fri  Mon  Tue  Wed  Thu  Fri Beg   End
 
URTY FE06 +1.2 12 +1.6 12 +1.3 12 +1.6 12 +1.7 12 +0.5 +0.8
AGQ FE07 +1.7 12 +2.2 12 +1.7 12 +2.0 12 +2.0 12 +0.7 +1.2
TNA FE11 +0.8 21 +1.5 22 +0.8 21 +1.5 22 +1.6 22 -0.9 +0.7
IWM   -0.9      +0.1   -0.9   +0.1   +0.4
me -1.9 -0.4 -1.9 -0.5 -0.3
floor -5.3 -4.5 -4.6 -3.6 -2.9

On Wednesday around 4:20pm, Yahoo! Finance was offline, so stop-updates for URTY and AGQ did not happen.  Yahoo came back an hour later, but the stops wouldn't have changed much anyway so I didn’t bother rerunning the update.

URTY: On Friday, the robot set the stop loss to +0.5, but my manual calculations came out to +0.8.  I raised the stop manually, but this needs investigation.

Stock-trading robot

No trades this week.

Double-TRIX trading model

Tuesday night, the robot generated a “buy” signal for UWM, but the signal was invalid because SMA(q) was less than CHANDELIER(+) so if the order had actually filled, it would have immediately stopped out.  I added a new rule: “do not buy if limit-price is less than stop-price”.  The new rule prevented a buy-order Wednesday night, but there was a valid “buy” generated Thursday night, which did not hit its limit during Friday’s session.  There is another ”buy” order for Monday.

Monday 17 February 2014

Week of 2014 FE 14

IWM rose by 2.9% this week, while my account rose 3.3%.  The loss-floor has risen sharply to -5.3%.  Seven weeks into the year, I have made back recent losses and my returns once again lag the broad market’s by only 1%.  The recent correction was not deep enough for my trading system to be much better than just staying fully invested all the time.

Equities and precious metals are both rising, while the dollar is falling.  Can you say “inflation”, boys and girls?

End of week allocations:

99% equities, 22% silver, 45% cash
% gain
size
Max loss
Sym  Buy  Fri  Mon  Tue  Wed  Thu  Fri Beg   End
 
SCHH JA31 +0.1 11 +0.1 11 +0.1 0
SRTY FE05 -0.3 10 -0.3 10 -0.3 0
ATML FE06 -0.1 10 -0.1 10 -0.1 0
URTY FE06 +0.3 11 +0.4 11 +0.6 11 +0.8 11 +1.2 12 +1.2 12 -1.5 +0.5
AGQ FE07 +0.1 11 +0.2 11 +0.3 11 +0.3 11 +0.7 11 +1.7 12 -0.2 +0.7
TNA FE11 -0.0 11 +0.1 11 +0.7 21 +0.8 21 -0.9 -0.9
IWM   -3.8   -3.6   -2.7   -2.3   -1.1   -0.9
me -5.2 -5.1 -4.7 -4.4 -3.1 -1.9
floor -7.3 -7.0 -7.5 -7.0 -6.7 -5.3

URTY: On Thursday, price rose above hourly SMA(180), causing a sudden jump in my stop-price.  This is the opposite of the sudden drops seen in the previous two weeks.  It might be better if I had four stop-price rĂ©gimes instead of two, as currently the two rĂ©gimes give sharply-different prices.

AGQ: Sudden jump on Friday!

TNA: Max-loss is back to unchanged, even though I doubled my position.

cash: My spreadsheet’s numbers now match Schwab’s; apparently there was something wrong with one of the recently-sold positions that settled on Tuesday.

me: My spreadsheet’s numbers now match Schwab’s, but didn’t earlier in the week.  I suspect that something is still wrong.

Stock-trading robot

Ticker
Symbol
Buy dateBuy priceSell dateSell priceAcct Profit
ModelActualMAModelActual M  A  M  A 
URTY FE 06 11:00 FE 06 12:34 $75.45 $75.05 (Not yet)
AGQ FE 11 FE 07 09:30 $66.11 $65.04 (Not yet)
TNA FE 11 12:00 FE 11 12:28 $69.56 $69.76 (Not yet)
FE 13 12:00 $71.64 $71.59

AGQ: Switched to new model, which had me buy on Tuesday instead of last Friday.  Meanwhile, due to yet another records-adjustment, the official results now say that Tuesday’s low was one penny too high, and stayed above SMA(q) for the rest of the week, so the latest numbers have me missing out on entire run-up in silver.  I noticed this because the robot grabs daily data each day at 4:20pm (and retains these values rather than downloading updates for previous days), while the above table uses a different database with data for the whole week downloaded on Friday (stockcharts.com and Yahoo both agree on the change, so apparently this was an official back-dated adjustment).  I changed the once-per-week data to match my model.  Anyway, it’s really lucky that I bought on Friday using the previous double-TRIX parameters!

TNA: Bought late because Schwab’s price-quoter got hung up and I didn’t notice for half an hour.  Bought the second tranche on time at noon, even though Rogers screwed up and I didn’t get the robot’s email until 5pm.  Good thing I have the robot set to do its own thing and not wait for my manual approval!

Sunday 9 February 2014

Week of 2014 FE 07

IWM fell by 1.2% this week, while my account fell 1.4%.  The loss-floor has risen to -7.3%.

US news: Friday’s jobs report was terrible.  Many fewer jobs were created in January than had been expected.  Markets rose anyway, but this could be a short-term fake-out.

End of week allocations:

33% equities, 22% silver, 68% cash
% gain
size
Max loss
Sym  Buy  Fri  Mon  Tue  Wed  Thu  Fri Beg   End
 
ATML¹ JA29 +0.0 11 -0.4 10 -0.3 10 -0.3 10 -0.3 10 -0.3 0 -0.7 -0.7
FNSR JA30 +0.2 10 -0.4 10 -0.3 10 -0.3 10 -0.3 10 -0.3 0 -1.3 -1.3
URTY¹ JA30 -0.2 10 -0.6 10 -0.6 10 -0.6 10 -0.6 0 -1.6 -1.6
SCHH JA31 +0.1 11 -0.0 11 +0.1 11 +0.1 11 +0.1 11 +0.1 11 -0.2 -0.2
SRTY¹ FE03 +0.0 11 -0.2 10 -0.2 10 -0.2 10 -0.2 0 -0.9 -0.9
SRTY² FE05 -0.1 11 -0.3 10 -0.3 10 -0.3 -0.3
ATML² FE06 -0.1 10 -0.1 10 +0.7 +0.7
URTY² FE06 -0.1 11 +0.3 11 -1.2 -1.5
AGQ FE07 +0.1 11 -0.2 -0.2
IWM   -2.6   -5.6   -4.9   -5.6   -5.1   -3.8
me -3.8 -5.3 -5.3 -5.4 -5.7 -5.2
floor -7.7 -7.6 -5.6 -6.0 -6.8 -7.3

[out]: Pulled out some more money to put food on the table.  I need a new job soon!

URTY²: Once again the max-loss value has dropped due to older, higher prices exiting from the averaging period.  New max-loss would have been -1.8 except for the change to the double-TRIX model (see below).

Stock-trading robot

Ticker
Symbol
Buy dateBuy priceSell dateSell priceAcct Profit
ModelActualMAModelActual M  A  M  A 
ATML¹ JA 29 09:30 $8.34 $8.31 FE 04 09:30 $8.11 -0.35% -0.28%
FNSR JA 30 JA 30 10:00 $23.22 FE 04 09:30 $22.49 -0.33%
URTY¹ JA 30 11:00 $82.20 $81.96 FE 03 11:00 $77.22 $77.12 -0.60% -0.63%
SCHH JA 31 09:30 $31.06 FE 06 09:30 $31.38 +0.10%
SRTY¹ FE 03 13:00 $49.78 FE 04 10:00 $49.17 $49.12 -0.13% -0.17%
SRTY² FE 05 11:00 FE 05 10:00 $51.49 $50.48 FE 06 11:00 FE 06 12:34 $48.53 $48.97 -0.55% -0.34%
ATML² FE 06 09:30 $7.60 FE 06 09:30 $7.31 $7.58 -0.39% -0.06%
URTY² FE 06 11:00 FE 06 12:34 $75.45 $75.05 (Not yet)
AGQ² FE 07 09:30 $65.53 $65.04 (Not yet)

SCHH: Profitable (barely) completion of first-ever real-estate trade!  Sold due to failed retest (hadn’t gone up four days after purchase).  Of course, it started rising immediately after I sold it, so I got a “buy” signal for Friday, but the limit-price wasn’t reached.  There’s another ”buy” signal for Monday, but I’ve decided to stop trading this ticker (see below).

SRTY², URTY²: Failed to swap SRTY for URTY on time due to a bug introduced by the new double-TRIX trailing-stop parameter (see below) which was implemented but not (supposedly) yet activated.  Didn’t notice for 1½ hours.

ATML²: Gapped down so hard that it opened below my stop price, then stopped out 30 seconds later.  I believe this is the first time that has happened since installing the new double-TRIX system last July — it was not clear that Schwab would actually accept the stop-order “so soon”.  The model has a special rule for such cases that assumes the worst: my sell-price = low of the day.  Actually I got a much better price than that, so the good news is that the “instant stop-out” thing actually does work, and with a smaller loss than the model calls for.

AGQ: Got a “buy” signal for Thursday.  My limit-price was close to the low of the day.  Official results at 4:20pm said the order should have succeeded, but it hadn’t (so robot tried to adjust the stop on a position it didn’t actually hold).  I decided to rerun the buy on Friday with the same limit price (Wednesday’s close).  This order went through within the first minute after Friday’s open.
      Meanwhile, due to records adjustments (or something), the official results now say that Thursday’s buy actually *shouldn’t* have gone through, so there should have been a repeat buy-order on Friday using Thursday’s close as the limit.  Well, I got a better price than that!

Double-TRIX trading model

I have added a new parameter to the model: trailing-stop.  The stop price is now constrained so it cannot be less than trailing-stop times the peak price seen since buying the ticker.  This allows other parameters to be retuned more aggressively, which improves profit and/or reduces average daily risk.

(daily) (hourly)
Model ticker: ATML FNSR YHOO SCHH FRED SLV IWM IWM
Trade ticker ATML FNSR YHOO SCHH FRED AGQ (2×) UWM (2×) URTY (3×)
TRIX(q) 3 3 2 4 18 4 26 27
TRIX(s) 14 17 26 17 59 29 54 97
SMA(q) 7 1 10 12 10 3 2 1
SMA(s) 270 270 120 210 150 250 150 180
CHANDELIER(c⁺,a⁺) 28, 4.5 26, 3.9 38, 4.9 6, 4.1 32, 3.4 11, 4.6 32, 4.8 312, 11.1
CHANDELIER(c⁻,a⁻) 22, 4.2 22, 3.8 21, 3.9 16, 2.4 23, 2.7 20, 2.2 25, 4.3 28, 9.6
RETEST(w,m) 4, 1.02 2, 0.97 3, 0.99 2, 1.06 3, 1.005 10, 0.98 5, 1.005 999, 1.0
Trailing-stop 0.84 0.74 0.84 0.80 0.89 0.84 0.79 0.88
Results for the last twelve years (or all available years, if less):
Old profit 32% 48% 19% ~6% 26% 27% 16% 60%
New profit 37% 38% 21% ~8% 26% 31% 23% 56%
Buy-and-hold 0% -4% +29% ~+1% +1% ~+9% ~+23% ~+34%
Old daily risk +0.5% +1.0% +0.1% +0.1% -0.3% -0.3% -0.4% +0.7%
New daily risk +0.1% +0.3% -0.2% +0.3% -0.4% -0.3% -0.6% +0.6%
New win:loss 8:4 7:5 9:3 4:4 9:3 6:2 11:1 9:2
(Yellow indicates a change from the previous model.)

Profits assume that tranches are 10% of my account.  Buy-and-hold profits for SCHH, AGQ, UWM, and URTY are estimated (from VNQ, SLV, and IWM) because these ETFs haven’t been around long enough.  Old/New profits for SCHH are also estimated for a similar reason.  “Daily risk” is the average daily difference between stop-price and purchase-price; if negative, then on an average day when I’m holding this ticker the stop is above the original purchase-price.  “Win:loss” is the number of profitable years vs. losing years.

Based on these results, I have decided to discontinue trading of FNSR and SCHH.  These have the worst win:loss ratios.  For FNSR, either the daily risk is terrible or the trading profit takes a big hit.  For SCHH, profits just aren’t very big regardless — and this model has had losses for each of the last three years.  Both these tickers have “buy” orders for Monday, but I’m not buying them.

Since there are now fewer tickers being traded, I have increased the tranche size from 10% to 13% for these double-TRIX models.  All profit and risk figures above should be multiplied by 1.3 to get the new values.  The TNA and TZA models are unaffected since they use PPO rather than TRIX (and they have “rebuy” rules to increase their position sizes).

The new model-parameters agree that I should be holding URTY and not holding ATML, FRED, UWM, or YHOO.  However, because of the new SMA(q) parameter for AGQ, the revised model says I shouldn’t have bought that yet — it has a “buy” signal for Monday.  Of course, the revised model also says I am to put 13% of my account in these tickers but I bought only 10%.

MACD-based trading model

This thing really doesn’t work very well.  It needs replacement.  However, I do not yet have a suitable replacement model for SRTY.

Sunday 2 February 2014

Week of 2014 JA 31

IWM fell by 1.1% this week, while my account fell 1.2%.  The loss-floor has fallen to -7.7%.

Business news: On Tuesday, Turkey hiked their overnight interest rate from 7¾% to 12½% (the US and Euro rates are approximately 0%).  Markets rallied, but within 12 hours had sunk back to where they were before.  This refusal to rally on inflationary news is a bearish sign for world markets.
      On Wednesday, the US Federal Reserve announced that they would continue their “taper” and print up only $65 billion to support the ultrarich in February.  Markets shrugged it off.

End of week allocations:

51% equities, 10% real estate, 59% cash
% gain
size
Max loss
Sym  Buy  Fri  Mon  Tue  Wed  Thu  Fri Beg   End
 
FRED OC24 +1.0 11 +1.0 11 +1.0 11 +0.7 11 +0.7 11 +0.3 11 +0.4 +0.4
ATML¹ DE18 +1.2 12 +1.1 12 +1.2 12 +1.2 12 +1.2 12 +1.2 0 +0.6 +0.6
AGQ JA13 -0.2 10 -0.5 10 -0.5 10 -0.5 10 -0.5 0 -0.5 -0.5
UWM JA21 -0.5 10 -0.5 10 -0.5 10 -0.5 10 -0.5 0 -0.6 -0.6
SRTY JA24 -0.0 10 +0.2 10 +0.2 10 +0.2 11 +0.2 0 -0.9 -0.9
URTY¹ JA28 +0.0 10 -0.5 10 -0.5 10 -0.5 10 -0.4 -0.4
ATML² JA29 -0.1 10 +0.0 11 +0.0 11 -0.7 -0.7
FNSR JA30 -0.0 10 +0.2 10 -1.3 -1.3
URTY² JA30 +0.0 10 -0.2 10 -0.7 -1.6
SCHH JA31 +0.1 11 -0.2 -0.2
IWM   -1.5   -2.9   -1.9   -3.3   -1.9   -2.6
me -2.6 -3.0 -2.8 -3.7 -3.5 -3.8
floor -4.7 -4.0 -3.8 -4.5 -7.1 -7.7

All leftover trades from 2013 have now been sold.  My “loss YTD” figure is still 0.11% less than what Schwab claims.  Schwab’s notion of cash-on-hand is 0.01% greater than predicted by my records.  Still not sure what to do about these.

AGQ: Due to last week’s 1:4 stock split, I ended up with some fractional shares.  After Monday’s sale, the fractional shares were credited as immediate ”cash in lieu”, while the rest of the shares were credited to cash on Thursday as usual.

FRED: My first dividend since 2011!  FRED declared its dividend on NO 27, paid DE 16, but I didn’t notice until this week.  The dividend adds 0.04% to my effective account-profit from this trade, so it’s small enough to get lost in the round-off error.

URTY²: The sharp drop in the max-loss value is due to an artifact of the “chandelier exit” system: we use the last 32 hours of price-data to compute the peak and true-range, so during Wednesday the higher prices from last week dropped out of the calculations.  The new lower stop went into effect after Thursday’s close — just *after* I bought!  Maybe there is an extra day’s delay here that could be removed.

Stock-trading robot

Ticker
Symbol
Buy dateBuy priceSell dateSell priceAcct Profit
ModelActualMAModelActual M  A  M  A 
FRED OC 18 OC 24 09:30 $16.11 $16.59 JA 31 JA 31 09:30 $17.38 $17.09 +0.75% +0.30%
ATML¹ DE 18 DE 18 09:30 $7.32 JA 28 09:30 $8.18 +1.19% +1.18%
AGQ JA 13 JA 13 09:30 $66.68 JA 27 JA 27 14:59 $63.67 $63.58 -0.46% -0.49%
UWM JA 21 JA 21 11:09 $86.22 JA 27 09:30 JA 27 10:18 $82.68 $81.44 -0.45% -0.69%
SRTY JA 24 13:00 $44.47 $44.50 JA 27 15:00 $45.29 $45.33 +0.16% +0.15%
URTY¹ JA 28 15:00 $82.30 $82.39 JA 29 15:00 JA 29 14:09 $79.18 $79.01 -0.40% -0.45%
ATML² JA 29 JA 29 09:30 $8.34 $8.31 (Not yet)
FNSR JA 30 JA 30 10:00 $23.22 (Not yet)
URTY² JA 30 11:00 $82.20 $81.96 (Not yet)
SCHH JA 31 JA 31 09:30 $31.06 (Not yet)

FRED: An okay trade.  Bought several days late (this was the first trade using a new model), which lost me 0.30% of account-profit.  Stopped out during a panic-open, so I got the lowest price of the day (opened at $17.38, dropped to $17.09 within the first minute, then recovered and stayed above opening price for rest of day).  Got an extra 0.04% of account-profit from the dividend.
      On JA 24, FRED’s price was up 14% after three months, but over the next ten days it dropped to only 5% up.  Quite a fall!  My model parameters for FRED are on the large side, causing long slow trades; the downside is that I will stay too long after the stock goes sour.

ATML¹: Great trade!

AGQ: Stopped out in the middle of the day, so I’m not sure why there was stop-slippage.  Anyway, it’s close enough — and silver’s price has dropped another 2% since, so I’m glad that I’m out of it.

UWM: Something went wrong with this trade.  The model now says that UWM should have been sold at the open via the “retest” rule (failure to be above purchase price after three days).  What actually happened is that it stopped out — and with slippage: the stop price was supposed to be $81.51.

SRTY: End-of-week price is roughly the same as where I sold it, although a price of $47.00 was available several times during the week.

URTY¹: Yet more stop-slippage.  What was up with the market-makers this week?

ATML²: Opened below my limit price.  For some reason I got a slightly-better deal than the open.  Still, bought it back for more than I had sold it for.  Hasn’t gone anywhere since.  Could still end up being a good trade.

FNSR: Had sold JA 22 for $23.75, bought back JA 30 for $23.22.  And now its price is up to $23.72!  So basically I gained 0.2% account-profit simply from *not* holding this ticker during a particular week.

SCHH: First-ever trade of a real-estate ETF!  This was one of the original tickers included within the new system installed last July.  Every night for six months the robot looked at the market and said “no” to buying this stock the next day, until JA 30 when it finally said “yes”.  The ticker opened JA 31 at $31.08 and dropped to my limit-price within the first minute.  It then spiked down to $30.91 at 09:32, then ramped up to $31.40 by 10:12 and stayed there for the rest of the day.  Looking good so far!

Trading model review

Now that the FRED trade has closed, let’s review results for the trading models over the last 3½ months.

REGULAR DAILY TRADES

Atmel: Pretty good!  Two gains, no losses, total +1.4%.  Trade in progress = -0.01%.  Buy-and-hold would have yielded +1.6%.
Finisar: not so hot.  Three losses, only one gain (but bigger than any of the losses), total -0.6%. Trade in progress: +0.2%.  Buy-and-hold would have yielded -0.1%.
Fred’s: disappointing.  One gain, total +0.3%.  Model says I should have received +0.75%, but I bought several days late and the stop-slippage on the sale was nasty.  Buy-and-hold would have yielded +0.9%.
Yahoo:: poor.  I missed out on the big Nov/Dec run-up because this ticker spent most of its time above the blue SMA line so all buy-limit orders failed.  Three losses, one gain, total -0.4%.  Buy-and-hold would have yielded +0.7%.
Real estate:: It’s nice that I avoided the Nov/Dec run-down, but I also missed out on the Jan run-up because of that spike on DE 27 which pushed the dotted red line skyward, preventing any trades.  Result: no gains or losses, total 0%.  Trade in progress = +0.1%.  Buy-and-hold would have yielded -0.1%.

LEVERAGED DAILY TRADES

Russell 2000: (2× bullish via ticker UWM): terrible.  This model’s timing did not match the market’s timing.  Four losses, total -1.5%, including two nasty stop-outs.  Buy-and-hold would have yielded +1.0%.
Silver: (2× bullish via ticker AGQ): Not good.  Four losses, no gains, total -2.0%.  Three trades were stop-outs, one was a same-day stop-out.  About the only good thing was that buy-and-hold would have yielded -2.3%, so I did better than that at least!

LEVERAGED HOURLY TRADES (Russell 2000 derivatives)

URTY: (3× bullish, using TRIX model): Pretty good!  Five gains, three losses, total +1.5%.  Trade in progress = -0.2%. Buy-and-hold = +1.5%.
TNA: (3× bullish, using PPO model): Excellent!  Two gains (both with rebuys), one loss, total +2.0%.  Buy-and-hold = +1.5%.
TZA: (3× bearish, using PPO model): As expected.  This is a “correction” trade and there haven’t been any corrections of note.  One gain, no losses, total +0.1%.  Buy-and-hold = -1.5%.
SRTY: (3× bearish, using MACD model): Disappointing.  Three gains, three losses, total -0.5%.  Buy-and-hold = -1.5%.