bot based on chart patterns

I noted a couple of patterns that seemed to happen at the same time after looking at maybe 1000 or more charts. I backtested the strat and it was viable. I made a bot to handle the trades. Here are the results:

trade date ***** profit / L ****** wins
2/19/2009 ***** 1.23 *********** 15
2/20/2009 ***** -2.106666667 ****** losses
2/20/2009 ***** 1.02 ************ 1
2/20/2009 ***** 0.89 *********** win%
2/23/2009 ***** 1.22 *********** 93.75
2/23/2009 ***** 1.063333333
2/25/2009 ***** 1.566666667
2/25/2009 ***** 0.606666667 ***** total profit in L
3/2/2009 ****** 0.2 ************ 8.661933333
3/3/2009 ****** 0.431866667
3/3/2009 ****** 0.0149 ********* no trades found
3/3/2009 ****** 0.988533333 ***** 2/24
3/3/2009 ****** 0.066133333 ***** 2/26
3/4/2009 ****** 0.573833333 ***** 2/27
3/4/2009 ****** 0.593333333
3/4/2009 ****** 0.303333333

As L is the amount risked to lose for the last trade, the profit/L will change as I change L . It's a good start but this performance is not expected long term. I'll update this once in awhile.

 

What are you trading?

Trap

 

Several different stocks. I'll give this 30 trades to see how well it works. I think that is enough trades.

 

Clean this up...

trade date profit / L
2/19/2009 1.23
2/20/2009 -2.106666667
2/20/2009 1.02
2/20/2009 0.89
2/23/2009 1.22
2/23/2009 1.063333333
2/25/2009 1.566666667
2/25/2009 0.606666667
3/2/2009 0.2
3/3/2009 0.431866667
3/3/2009 0.0149
3/3/2009 0.988533333
3/3/2009 0.066133333
3/4/2009 0.573833333
3/4/2009 0.593333333
3/4/2009 0.303333333

wins
15

losses
1

win%
93.75

total profit in L
8.661933333

no trades found
2/24
2/26
2/27

 

sample size for days and trades is so small this is absolutely meaningless. Pure noise.

 

What is this post for?

 

IMO, even for a intraday system, you need to be backtesting on at least 12 months data.

 

Thursday - updated bot based on backtest results only. Big mistake.
date profit in L
3/5/2009 0.281906667
3/5/2009 -0.908733333
3/5/2009 -0.836666667

Friday - fixed false positives from Thursday's strat updates. Added some stuff to try to increase number of trades. Doublechecked these changes against charts, like I should have done Thursday.

3/6/2009 -0.728933333
3/6/2009 0.5626
3/6/2009 0.7189
3/6/2009 0.112466667
3/6/2009 0.766666667
3/6/2009 0.406666667

total profit : 9.036806667 L

wins: 21 losses: 4 84%

 

trackstar - this post's purpose is to be a journal of the performance of a bot trading non-trivial sums of money. It was originally in the journal section before being moved here. It would be great to see more non-strategy specific results for other trading bots. Faceoff - I only need to test it against recent data.

 

3/9 2 winners, 2 losers
3/10 0 winners, 3 losers
3/11 2 winners, 5 losers

final results, at my designated end of trading, for a bot trading real, 5 figure sums of $ per trade: 25 winners, 14 losers, all trades were short.

summary: The bot, which only performed short trades, started out 21-4 during the highly bear period of the market. It went 4-10 during a part of the bull run the market had from early to mid March. I'm currently testing a version which does long trades as well as shorts. Other upgrades might make this work better during choppy market runs. I hope to deploy it with real $ late March/early April.