Statistical Edges In Trading

If you are going to trade using statistics, you cannot just use anything you want. You have to ensure the data has enough relevance and sample size. I wouldn't necessarily obsess over lack of objectivity of particular price patterns if you personally have a large degree of confidence that you can identify it. However, for the most part I would focus primarily on creating a system that starts with the largest statistical sample set that you can find. In my opinion, that is SECTOR seasonality, which have hundreds or thousands of stocks over decades of years every year trading over a particular seasonal date, and even more relevant if you are looking to capture a multiple month trend. This is not to be confused with individual seasonals which may only have 10 years of trading or 10 samples (a stock may have traded over only 10 months of October if it has been around for 10 years). The next statistical edge with a very large sample size is candlestick patterns. You have price patterns that may take weeks or months to form (although some take only a handful of days), but candlestick patterns are possibly only 1 day pattern plus confirmation.

So We start with the sector performance. i prefer to look for relative outperformance since the market has the ability to drag sectors up or down even though the current environment there isn't a lot of correlation.

From EquityClock

When you see a decline, this could mean the particular sector stays flat or goes up, but just that the S&P goes up at a faster rate. But for me that is important.

I typically like to hedge. That means that my hedge should go up less than the market, or go down more then the market. Overall it adds value and decreases volatility while reducing my exposure long.

For the month of October, I want to think about being long technology and maybe Materials and short financials and/or Utilities.

Now the Candlestick patterns...
For these I not only want a COMMON candlestick pattern for a large sample size, but I want one frequent enough to trade that also has an edge as either a bearish or a bullish signal.
My choice is the Marubozu pattern with the plan to wait for conformation to the opposite side. For example, a bearish Marubozu that breaks above it's high or a bullish Marubozu that breaks above it's low.
Note: The reason a Black Marubozu is called a bearish Marubozu is because it is a downward move and slightly more than 50% of the time you will have a confirmed downward breakout (close below the low). We are looking at BUYING a confirmed upwards breakout which means a "bearish failure", making it bullish. Similarly the White Marubozu is called "bullish", but we are looking to trade the failure.
====================================
Bearish Marubozu (Marubozu, black)
Bull Market Up breakout
Candle end + 1 day 1.91%
Candle end + 3 days 3.26%
Candle end + 5 days 3.86%
Candle end + 10 days 4.39%
10-day performance rank 24/103

Bear Market Up Breakout
Candle end + 1 day 2.57%
Candle end + 3 days 4.08%
Candle end + 5 days 4.96%
Candle end + 10 days 5.33%
10-day performance rank 25/103
====================================
Bullish Marubozu (Marubozu, White)
Bull Market breakdown (down breakout)
Candle end + 1 day −1.71%
Candle end + 3 days −2.81%
Candle end + 5 days −3.33%
Candle end + 10 days −3.55%
10-day performance rank 26/103

Bear market breakdown (down breakout)
Candle end + 1 day −2.20%
Candle end + 3 days −3.95%
Candle end + 5 days −4.53%
Candle end + 10 days −4.79%
10-day performance rank 36/103
 ====================================
While there are many better patterns, these still are among the better patterns and more importantly, it allows a very simple and very common pattern. Each day you will have a large number of ideas. I will filter it by sector since I am looking to trade seasonals as well. I also trade options so I am looking for optionable with enough volume. So here are the screens to create watchlists that should be run at the end of the day so you have confirmation of the pattern.

Bullish
Technology
http://finviz.com/screener.ashx?v=111&f=sec_technology,sh_avgvol_o50,sh_opt_option,ta_candlestick_mb&ft=4

Materials
http://finviz.com/screener.ashx?v=111&f=sec_basicmaterials,sh_avgvol_o50,sh_opt_option,ta_candlestick_mb&ft=4
-

Bearish
Utilities
http://finviz.com/screener.ashx?v=111&f=sec_utilities,sh_avgvol_o50,sh_opt_option,ta_candlestick_mw&ft=4

Financials
http://finviz.com/screener.ashx?v=111&f=sec_financial,sh_avgvol_o50,sh_opt_option,ta_candlestick_mw&ft=4


How To Enter Trades

This system is pretty simple. You are using two confirmed statistical edges to identify buy points. Since you will be getting ideas daily that confirm every few days for 1-10 day trades It is an easy system with a confirmed statistical edge. Beyond that, you probably want to use some discretion and when you really get the right trade, you might even use leverage.

Since the confirmed upward breakout/downward breakdown only occurs around 55% of the time for each trade, and the pattern is actually not very good at it's intended use, you may want to get long/short in anticipation of the entry if you feel you can identify any sort of bias at all. With this particular strategy, you are creating a greater upside, a longer time horizon, but a very quick exit. You basically will be entering the next day on the exit of the trade. Based upon the statistics, you should expect to lose more often than win, but the exit when you are wrong probably only costs you the average daily percentage move to the downside, while the exit when you are right gains you the size of the candlestick times two (target) about 75% of the time THAT it breaks out (about 33% of the time total). The remaining 12% of the time you will see somewhere between a small loss and a gain that falls short of the target. With leverage that probably means closing out the trade or seeing the contract expire.

Either way, you can do a risk/reward analysis pretty easily.
For risk, determine the average daily move over the last handful of days. Say it's $1.5
Then measure the current candle, say it's 1.6. Your risk is 1.5, your reward is 3.2.
55% of the time you lose 1.5 total value .55*-1.5=-.825
33% of the time you gain 3.2 total value .33*3.2=1.056
12% of the time you break even  .12*0=0
Total "expected value" = 0.231 If that is in a $50 stock that's a 0.462% per trade gain. 30 trades a year and it's 13.86% a year. If that is good enough you might make the trade.

Reward  1.056 / .885 = about 1.2 to 1
Trading the seasonality as well may increase your odds and upside as well since the bias on a given day is in the direction of your trade.

A way to analyze the typical case is taking the 1 day return if you are wrong to the 3 day average if you are right since you will let the winners run. In that case for a bullish trade on black marubozu you have
White Marubozu
Bull market down breakout
-.58%,  3.26%
bear market down breakout

-.91%, 4.08%

55% of the time lose .58 .55*-.58=-.319
33% of the time 3.26 .33*3.26=1.0758
12% of the time break even 0%=0
.7568%


55% of the time lose .91 .55*-.91=-.5005
33% of the time 4.08 .33*4.08=1.3464
12% of the time break even 0%
.8459%

Actually, that isn't entirely fair as the target will be greater and so the 33% if we are using 3 day average numbers should instead be 45%
55% -.319
45% of the time .45*3.26=1.46
1.148%

55% of the time =-.5005
45% of the time .45*4.08=1.836
1.3355%

And leverage?
Leverage is a complicated thing that is very dependent upon the security and it's specific expected move. But just for an idea, let's take goldman sachs and pretend it has just made a 4 point move forming a white marubozu to the downside. Let's say you really want to trade leverage to the upside.
You determine that you are aiming for at least a 4 point move from 161 to 165 strike price in option calls that have 7 trading days on them. The cost of the option will be about $.50 per share or $50 per contract.
OF those that breakout, 75% reach a target of 169 (measure rule of 4 points beyond the breakout price). However, it takes 4 days from the formation of the candlestick until you actually get a breakout on average. And that means HALF of the trades will take longer than 4 days... SO...
45% breakout, 50% of those that breakout do so in 4 days and 3 days are left to get the average performance of 3.26% in a bull market up breakout from the high which would be 165. That is about 170.4 very close to our measure rule target. I would just take 45% of 50% or 22.50% that hit $169 which would yield $4 on those calls. That is a return of $3.50/.50=7=700%.


Now the difficult part is determining the remaining 22.5% that breakout but take longer than 4 days. A fraction of them will take 5 and as a result have 2 days of performance. A fraction will take 6 days and still have 1 day of performance. And some on the last day of trading before expiration will end up with a confirmed breakout and result in you possibly gaining the cost back, maybe less, maybe more.

For now though we can just say that
-$.50 *.55=-.275 (in reality you would probably preserve a huge portion of this contract, but it isn't a given depending on price and liquidity)
$3.50*.225=.7875
$0*.225=0
+.5125 or over 100% per trade.
The problem of course with options is position size. You really have to keep the position size very small such as 1% but can put a handful of trades. So maybe with 5 trades that translates to 5% growth on your portfolio per trading interval if you have 5 trades on at a time and 30 trading intervals you can do 150% a year. But a huge predicament is that the trades will be slightly correlated and that means it is a tiny bit similar to having one large 5% position only not quite. as bad. Either way you have to be careful. Another problem is that although certain specific contracts have weeklies available, many do not, and many have illiquid options. Additionally, the optimal amount of time would probably be about 10 days with the plan to only use 7.

I think there simply won't be enough opportunity to use leverage with this method as frequently as you want and so I would consider trading any options even not in the right sector if it has weekly options available (even if you aren't using them, it means the options are probably more liquid than others).

Marubozu white
http://finviz.com/screener.ashx?v=111&f=ta_candlestick_mw&ft=4&t=OEX,XEO,SPX,DJX,NDX,RUT,AGQ,DIA,DUST,DXJ,EEM,EFA,EWJ,EWZ,FAS,FAZ,FXE,FXI,GDXJ,GLD,GLD7,GDX,HYG,IWM,IYR,QQQ,NUGT,SDS,SLV,SPY,SPY7,SSO,TBT,TLT,TNA,TZA,USO,UNG,UVXY,VWO,VXX,XHB,XLB,XLE,XLF,XLI,XLK,XLP,XLU,XLV,XLY,XME,XOP,XRT,AA,AAPL,AAPL7,ABT,ABX,ACN,AET,AGNC,AGU,AIG,ALXN,AMD,AMGN,AMRN,AMT,AMZN,AMZN7,ANF,ANR,APA,APC,APOL,ARNA,ATVI,AXP,BA,BAC,BAX,BBRY,BBY,BIDU,BIIB,BK,BMY,BP,BRCM,BTU,BX,C,CAT,CELG,CF,CHK,CL,CLF,CMG,CMI,COF,COH,COP,COST,CSCO,CREE,CRM,CRUS,CVX,DAL,DD,DDD,DE,DECK,DELL,DIS,DISH,DNDN,DOW,DVN,EBAY,ELN,EMC,EOG,ETN,F,FB,FCX,FDX,FFIV,FSLR,GE,GG,GILD,GLW,GM,GMCR,GME,GNW,GOOG,GOOG7,GPS,GRPN,GS,HAL,HD,HES,HLF,HON,HPQ,IBM,INTC,IOC,IP,ISRG,JCP,JNJ,JOY,JPM,KBH,KGC,KMB,KO,KORS,LCC,LINE,LLY,LNG,LNKD,LOW,LULU,LVS,M,MA,MBI,MCD,MCP,MDLZ,MET,MGM,MMM,MNKD,MON,MOS,MPEL,MRK,MRVL,MS,MSFT,MTG,MU,NAV,NE,NEM,NFLX,NKE,NLY,NOK,NTAP,NVDA,ONXX,ORCL,OXY,P,PBR,PCLN,PFE,PG,PHM,PM,POT,PSX,QCOM,QCOR,QIHU,REGN,S,SBUX,SCTY,SINA,SIRI,SLB,SLW,SNDK,SNE,SODA,SRPT,STX,SU,T,TAP,TGT,TIF,TIVO,TMUS,TOL,TSLA,TSO,TXN,UA,UNH,UNP,UNXL,UPS,USB,UTX,V,VALE,VRTX,VHC,VLO,VMW,VOD,VVUS,VZ,WAG,WFC,WFM,WLT,WMB,WMT,WYNN,X,XOM,YELP,YHOO,YUM,Z,ZNGA,ZTS

Marubozu black
http://finviz.com/screener.ashx?v=111&f=ta_candlestick_mb&ft=4&t=OEX,XEO,SPX,DJX,NDX,RUT,AGQ,DIA,DUST,DXJ,EEM,EFA,EWJ,EWZ,FAS,FAZ,FXE,FXI,GDXJ,GLD,GLD7,GDX,HYG,IWM,IYR,QQQ,NUGT,SDS,SLV,SPY,SPY7,SSO,TBT,TLT,TNA,TZA,USO,UNG,UVXY,VWO,VXX,XHB,XLB,XLE,XLF,XLI,XLK,XLP,XLU,XLV,XLY,XME,XOP,XRT,AA,AAPL,AAPL7,ABT,ABX,ACN,AET,AGNC,AGU,AIG,ALXN,AMD,AMGN,AMRN,AMT,AMZN,AMZN7,ANF,ANR,APA,APC,APOL,ARNA,ATVI,AXP,BA,BAC,BAX,BBRY,BBY,BIDU,BIIB,BK,BMY,BP,BRCM,BTU,BX,C,CAT,CELG,CF,CHK,CL,CLF,CMG,CMI,COF,COH,COP,COST,CSCO,CREE,CRM,CRUS,CVX,DAL,DD,DDD,DE,DECK,DELL,DIS,DISH,DNDN,DOW,DVN,EBAY,ELN,EMC,EOG,ETN,F,FB,FCX,FDX,FFIV,FSLR,GE,GG,GILD,GLW,GM,GMCR,GME,GNW,GOOG,GOOG7,GPS,GRPN,GS,HAL,HD,HES,HLF,HON,HPQ,IBM,INTC,IOC,IP,ISRG,JCP,JNJ,JOY,JPM,KBH,KGC,KMB,KO,KORS,LCC,LINE,LLY,LNG,LNKD,LOW,LULU,LVS,M,MA,MBI,MCD,MCP,MDLZ,MET,MGM,MMM,MNKD,MON,MOS,MPEL,MRK,MRVL,MS,MSFT,MTG,MU,NAV,NE,NEM,NFLX,NKE,NLY,NOK,NTAP,NVDA,ONXX,ORCL,OXY,P,PBR,PCLN,PFE,PG,PHM,PM,POT,PSX,QCOM,QCOR,QIHU,REGN,S,SBUX,SCTY,SINA,SIRI,SLB,SLW,SNDK,SNE,SODA,SRPT,STX,SU,T,TAP,TGT,TIF,TIVO,TMUS,TOL,TSLA,TSO,TXN,UA,UNH,UNP,UNXL,UPS,USB,UTX,V,VALE,VRTX,VHC,VLO,VMW,VOD,VVUS,VZ,WAG,WFC,WFM,WLT,WMB,WMT,WYNN,X,XOM,YELP,YHOO,YUM,Z,ZNGA,ZTS

I don't expect the results to be typical, the information was found or derived from Encyclopedia of candlestick charts by Thomas Bulkowski.

You certainly could use this as your only trading method, I use it to find hedges primarily and the occasional trade if the setup is right.