Tall candles

Today I'm trying Bulkowski's Tall Candle setup with intraday data.

Tests on daily charts were  disappointing at first. Bearish setups were loosing in bull markets, bullish setups were loosing in bear markets.

Tall candle trades on daily charts - bullish setups profitable in bull market

So, I had to ignore tall candles which indicate direction against market trend. First I have tried testing against short term trend, but it did not change results much:

Not trading against short term trend

And finally tests with long term trend in mind give very interesting results, at least for bull setups. Results are not perfect, but it is something to work on.

There were only 74 bullish setups in 2000-2010, average profit was two times bigger than average loss and 53% of winning trades made it a 136.81% average yearly profit. Drawbacks are too big, so I do not dare to calculate Sharpe ratio.
I believe I will move back to the daily setup some time later - tweak various parameters, play with stops and try other options.

Next step - moving to a smaller time scale (5 min. charts). Same setup as for daily charts recognizes too many candles as tall (1400 bullish and 1000 bearish setups in year 2012) and results are strange - bullish setups are consistently loosing, bearish setups generate are very unstable.

If fishing for candles 2.46 times bigger than average, 4 times less patterns are recognized, drawbacks are not that drastic, but it is still unusable:

I had to do something, so I have tweaked recognition algorithm big time and ended up with trading setup which invests after huge candle in the opposite direction. This way after huge bear candles I get 88 bullish trades per year and get consistent 10% yearly return. Bearish setups are still inconsistent. Of course I need more data to prove that.

Experimenting with pipe bottoms

Hey,

Today I'm analyzing Bulkowski’s pipe bottom setups (+ pipe top). I know these setups are best performing on a weekly scale, but I'm interested to view how they could perform in smaller time frames.

First I have attempted pattern on a daily scale with several selected securities and got -6% average yearly loss from 1990 till the end of 2012. Performed was good in 2000-2004, but bear market of 2007-2008 was  devastating. Anyway, I have suspected issues with the pattern recognition, so this result does not yet count.

I have fixed the algorithm a little and got 62% winning trades, but losses were too big and average yearly loss of 3% meant that half of trading capital would be lost in 20 years. So, setting fixed target and stop does not work very well.

If setting very loose stops (trail 135 last days), pipe tops under-perform (after 10 years trading capital is gone), but pipe bottom performs nice - 37.5% of winning trades, 27.5% average yearly profits. But profits are not consistent (most of the profits are from 2010-2012, in 2000-2002 performance was very bad. As it basically becomes buy and hold strategy, it would be nice to compare it to some benchmark, but I'll leave it for another time.
With tighter stops (trail 1-3 days) pipe bottoms under-perform (constant looses guaranteed).

I have adjusted algorithm several more times and it appears that search for very tall pipes gives better results (only 30 entries in 20 years using about 300 stocks). With 20 day trailing stops 1991-1999 years were quite successful (10000$ turns into 80000$), but later all the gains were lost in 5 years and there only few positive spikes in later years.

Anyway, I did not found any consistent and reliable setup while trading pipe bottoms in 5-15 min. scale.

Introduction

Hi,

This is my introductory post, so it will not be very entertaining.
I have started this blog for myself - so I could collect my toughs about loosing trades and tested strategies in a single place. My trading experience is not very rich:

• I have started in Lithuanian market, where I invested without any knowledge, but now I would call it "buy and hold" investment strategy. I did not know nothing about technical or fundamental analysis, but with luckily in 3 investments I have gained 50-100%, those profits covered the loses in other positions and on average I was profitable.
• Lithuanian market is not very liquid, commission costs are high, so I have started looking at US stock market. I have read a little about technical analysis, about options and other instruments and in my first week I have lost 40% of my trading capital playing with earning volatility (option straddles). This experience was stressful, but I have learned new interesting terms (for example, "volatility crash") and pushed me to learn deeper.
• After reading some books ("Reminiscences of a Stock Operator" and others), different swing trading blogs and sites I have started creating my money management strategy which I still use today and it helps me to control looses. Strategy does not let me loose more than 2000$ per year and as I have invested some money into bonds, bond yields cover those looses.
• In about a year I had some profits, but pair of unsuccessful trades (which again involved writing options) moved me into break even position. So, I'm not a successful trader yet, I did not find a good swing trade strategy to hold on, so there is not much to learn from me.

As trading with my money management strategy is not very time consuming, several months ago I have also started looking at algorithm trading:

• I have created some artificial intelligence algorithms which after learning could guess next intraday bar direction with quite good precision and I could earn 200-400% just by using this algorithm, if only there would be a commission-free broker. Otherwise commissions would take most of the profits, so I did not dare to test this algorithm yet.
• I have created an intraday strategy back-test platform and tested it with several different strategies (breakout, candle reversal, IBS, stochastics, etc).

As of today there is only one strategy which was very profitable in back-tests - using 2012 year 5 min. intraday data it generates 500-700% profits (depending on stop placement) even after deducting commissions. For earlier year I have intraday data only for few symbols, but with this data strategy is profitable, but profits are just 10-20%. I am testing this strategy on simulated account and it generated 80% profit last week (mostly because it invested in NFLX day before huge leap), but strategy results are not very consistent, so I'm not ready to give it a shot with real money.

So, that is my story. I will post more after testing other strategies.