[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Oddball Directional Bias



PureBytes Links

Trading Reference Links


the assumption of oddball,

1. advance issues's momentum change is a leading indicator
2. you are willing to take wide stop :)

the statistics,

1. you will want to locate the lowest time frame that
   the advance issues has highest correlation with 
   the price data, given that the data before 10 am are removed
   from the series.

2. you will want to locate the time to cut the morning part of
   the price data and the closing part of the data which can make
   the system more profitable, as instability to a breadth driven
   system is greatly affected by the opening and closing movements.

try to see if you can locate the best time frame and 
what part of data to cut off :)

you will probably have a hard time doing that in TS ...

Lawrence



--- David Folster <mr_bond@xxxxxxxxx> wrote:
> Oddball Followers,
> 
> (Following results are for the time period Nov 30/98 - July 26/2001)
> 
> Here are some results on Oddball using a simple directional bias (i.e. only
> taking trades in the direction of the trend).  The directional bias used was
> just a 25 period MA of the hourly close (see code variation at the bottom).
> I invite comments on the results, as I have trouble figuring out what good
> tradeoffs are in tweaking systems.  Let me first say that I have not been
> able to reproduce Mark's results exactly, and for some reason, I get wildly
> differing results depending on what time frame my data is downloaded in (I
> tried with 30 min., 5 min. and 1 min. downloaded as txt or csv files from
> quote.com).  My best results were using the 30 minute file, where I  got net
> profits of about 640K, then with the 5 minute file, it was 430K, and with a
> 1 minute file, the worst results of 339K.  I am taking these worst results
> to be the most accurate.  Having said that, I know I'm still not getting the
> same signals, as I plotted the system on ES1Z and compared it to the gif on
> MArk's site -- mine has more signals and more whips.
> 
> ***PS (Can anyone send me several years (as many as possible) of data for
> back adjusted Intraday (preferably tick) snp and NYSE advancing issues in
> OMZ or XPO format?  The accuracy of the data and how it lines up is one
> problem I think.  Thanks in advance!)
> 
> RESULTS -- ORIGINAL ODDBALL
> 
> Total Net: $339,375
> Total Gross: $1,232,125
> # Trades: 707
> Wins: 333
> Losses: 374
> % right: 47.1%
> MAX Intraday DD: $61,325 (ouch)
> Maximum EOD DD: $64,667 (48%) (ouch)
> Maxiumum time betwen new equity highs: 10 months (279 trades)
> 
> 
> RESULTS -- ODDBALL DIRECTIONAL BIAS
> 
> Total Net: $282,075
> Total Gross: $831,000
> # Trades: 486
> Wins: 254
> Losses: 232
> % right: 52.26%
> MAX Intraday DD: $$32,475
> Maximum EOD DD: $34,775 (25.24%)
> Maxiumum time betwen new equity highs: 5 months (68 trades) -- (However
> immediately preceding this there was another flat period of 4 months/74
> trades)
> 
> 
> Good things:  221 fewer trades, HALF the EOD DD and Intraday DD, slightly
> smoother equity curve.
> Bad things: 60K less in profits, Anything else???
> 
> Also note, tested over the last 2 months, Original Oddball had WAY better
> bottom line by comparison to mine (58K as opposed to 38K).
> 
> 
> 
> ODDBALL DIRECTIONAL BIAS
> 
> { Data1 = S&P 500 futures contract.
>   Data2 = NYSE Advancing Issues.}
> 
> Inputs: RL(7), BZ(3), SZ(1), MALength(25);
> 
> if time > 930 and time <= 1600 then begin
> 
> If c data1 > average(c data1, MALength) and RateOfChange(C Data2, RL)>BZ
> Then Buy;
> If c data1 > average(c data1, MALength) and RateOfChange(C Data2, RL)<SZ
> Then ExitLong;
> If c data1 < average(c data1, MALength) and RateOfChange(C Data2, RL)<SZ
> Then Sell;
> If c data1 < average(c data1, MALength) and RateOfChange(C Data2, RL)>BZ
> Then ExitShort;
> 
> End;
> 
> 
> 
> 
> DAvid Folster
> 


=====
Lawrence Chan   http://www.tickquest.com    
Transform market data into opportunities