Saturday, April 28, 2007

BOB - We Have A Leader In The Clubhouse

I won't say "winner" because that would imply that the contest is over and done and nothing more can change and that it is finished. It will never be finished - the quest is never ending - but we do have a modification to BOB that I'm going to adapt.

Bullish Jim discovered it and you will have to read about that here and I suggest you read Jim's BLog frequently. There are two kinds of trading BLogs out there - those written by people who already know it all and wouldn't change a thing if their lives depended on it and those by people who actively seek change because they know growth only comes with change. Jim's is that kind of BLog - his is one of the good ones.

Anyway, long story short - Jim suggested that we constrain the first two candles to have a limited open to closing range. He did some tests and found some good results and I've followed up and my results are in this post.

I took my original BOB filter ("v1x")
show stocks where close is between 15 and 35
and average volume(90) > 500000
and ema(8) < ema(21)
and close 2 days ago < ema(8)
and close 2 days ago < open 2 days ago
and close 1 day ago < open 1 day ago
and low 1 day ago < low 2 days ago
and volume 1 day ago is more than 20% > volume 2 days ago
and close > open
and low > low 1 day ago

and made a change to it that I called "v31". ("v2" and mods are long gone).
and close 1 day ago is less than 1% < open 1 day ago

I then tested that version - we'll get to the results presently. Then I made a mod to "v31" named "v31x" -
and close 1 day ago is less than 1.5% < open 1 day ago

Followed by another test. Then I made an additional change (which is the change that Jim proposed) and I called this "v32" -
and close 1 day ago is less than 1% < open 1 day ago
and close 2 days ago is less than 1% < open 2 days ago

And tested that version. This was followed by one more change that I named "v32x" -
and close 1 day ago is less than 1% < open 1 day ago
and close 2 days ago is less than .5% < open 2 days ago

Notice that the difference is that the red bar two days back is constrained tighter than the red bar 1 day back. I tried it the other way around and achieved good win percentage but bad ROI. This way seems to work much better.

Here are the results -

And what now?

The tradeoff, of course, is that there is a variance in number of stocks selected as the filter becomes more constrained - here is how that looks -

You see that as the constraints become tighter you get a better return on the investment but the numbers of stocks selected goes way down. This is a dilemma - what to do, what to do?

Well here is what I'm going to do - I'm going to use every version and I'm going to access them daily in reverse order - "v32x" through "v1x" in turn and the first time I get a selection - that is going to be my BOB selection of the day.

Consequently I know that the risk is increasing as I progress through the filter set but we encounter risk all the time in this business. And the way we handle risk? Tighten up the stop.

Thank you Jim you are the Leader in the Clubhouse in the race to find the "better BOB".


And for any new readers - I use as my test bed. There may be better ones out there but this one is reasonably priced, allows me absolute freedom to write any strange method I can come up with and are really, really, really fast when responding to your plea for help via email.


Anonymous said...

Congrats, Bullish Jim! After reading his post last night, I came up with the following potential daytrades using his example of a BOB whose down candles have a hi-lo differential of < .01 (all on 10 minute charts):

4/26 ARRS 11:50
4/27 ARM 10:00
4/27 BKC 12:40
4/25 BLL 11:20

I wrote these down before looking at your most recent post, Marlyn, and I told myself, "Bullish Jim's idea appears to have some merit."

Very nice job Bullish Jim, and thank you Marlyn for taking the time to peform the backtest.


Bullish Jim said...

Thanks, Philip. This stuff is so much fun.

And thank you for the props, Marlyn. If I added anthing of value to the BOB filter please consider it a tiny payback for the thousand things I've learned from your blog.

My next goal is to learn to trade as well as the backtester does! I guess that's why I need to get around to implementing an automated system...

Bullish Jim said...

I just tested v32x using my standard backtest parameters including 1 stock per day and a maximum of 6 at a time. I got 47 stocks, a 135% ROI, no stop outs, and the highest reward to risk I've ever gotten in a backtest of 13.5. Unbelievable!

IIO said...

You guys have gotten me curious. I tried but all I get is a generic search page.

Marlyn Trades said...


den0 said...

Marlyn, Trying to set up your filter using my Telechart software. I don't seem to get the number of hits that you experience. I'm wondering if I am asking for the correct volume number. Is the average volume(90) stated in dollars or number of shares. All other criteria are easy to duplicate. Thanks for your input. Your analysis is outstanding and very appreciated. I continue to look at your blog every day for new insight. Thanks again.

kwan said...

I'm trying to figure out what's your exit criteria for your backtests? I went back in the archives and saw you have a 4 day minimum hold with a sell at the close unless a stop loss of 10% is triggered. Is this still what you're using for this particular backtest, or are you using something else?

BTW, love the blog. I've been a lurker for a few months now...

Marlyn Trades said...

kwan - here is my current exit criteria - 4 days, or, stop loss 8% or the function =
close 1 day ago > close 2 days ago
and close > close 1 day ago

That exit trigger is hit more than half the time - the stop loss is hit infrequently.

den0 - that is number of shares - let me know how telechart works.

den0 said...

Marlyn, I ran the scans with number of shares and all of the other criteria. I haven't figured out the problem yet, but I currently come up with no stocks identified for any of the five BOB options. Something is not working correctly for me.

Woodshedder said...

Marlyn, great stuff. I love this quantification of candles/price action.

I am getting ready to take on a system development quest, and am curious what you use to backtest ideas.

Also, Bullish Jim, if you are reading, what do you use?

Thanks guys!