The quest for October Profits

[MBender]

Okay. I have an adequate test lab. I have a program structure that allows me to plug in a new trading algorithm in about an hour. So, here's that I have so far:

I can't get an XTL strategy that makes profits an any of the preceding 6 months (more work to do on this -- testing has been light).

I tried a simple EMA/SMA crossing strategy with little luck. I'm somewhat convinced this makes an excellent confirmation strategy but not a trading strategy.

E-signal textbook stochastic trade shows promise, but testing is still early on this (I wrote the algorithm at 7:00 this morning).

Standard textbook (well, almost) Donchian shows the most promise (75+ success rate in the Summer months, Profits May - September). Understand that 3 weeks ago I didn't even know there was a standard Donchian strategy.

STILL!!!! I can't find an algorithm that makes money on October. I break even at best.

I'm convinced trending strategies won't work. Channeling strategies (more stochastic work) is next on my list.

Suggestions welcome; well, actually, begged for.

[Doji3333]

Well, I believe the best solution is a combination of strategies whereas both trending and rotational strategies combine for the best results.

I've built a number of new things over the past month or so. First, my adaptive strategy is finally operating and trading in realtime as I expected (almost completely debugged). Limiting it to 2 loses a day seems to work very well and keep me out of trouble.

Additionally, I've build a new Filter that I'm playing with. This new filter is designed to catch trend reversals as quickly as possible. Actually, I now have 3 versions of the filter and am considering developing a scalping system based on a combination of two of them. I'll try to post some more images later this week.

One issue I've found to be concerning is the scalping mode of my adaptive system. I'm seeing that it works really well at containing unwanted risk, but that at the end of trends, the system can get into one extra trade that generates losses (not always, but sometimes). Wish I could find a way to identify that last trade and turn it off.

Here's a chart of the ES with my adaptive system for today. Getting pretty excited about the potential here.

[Doji3333]

Here is the NQ (scalper mode) chart for today.

As of right now, the NQ system is +$84.00 in completed trades and -$12 in the current trade (trading only one)

The ES system is +$71 from it's first target and +$335 in the trailer. Remember, I'm running the ES as a trend following (trailer) system and the NQ system as a scalping (in and out at target) system. This way I can see/test the benefits of both model types.

[arbifox]

Not all months can be good . . .

Today seems to be a great trading day, I am up about $300 on version 7 but the same program lost 26.4 points in October. I would have killed for a flat . . .

Not sure what to say in terms of what works and does not work, but I would just continue to backtest on several months and if you make money over the course of the previous 3 months, I would say that is a positive goal to shoot for. Sometimes a flat is fantastic especially if the market either fell, or is not doing anything.

The other thing I started to try to control is trading frequency. I have made over 900 trades this year, and I think I need to cut that down. Too many times I try to design the system to trade like I would mano-a-mano with the market, where the goal of automated trading should be to do what humans can not . . . wait patiently and take only the trades that are proven to be statistically winners.

With that in mind, I record the starting conditions of trades, and after I have tuned the system to yield the most amount of points over a test period, I try to weed out trades that might have gone short in an oversold state on one of the charts and see if those trades collectively are negative. Saving commission is one way of increasing results.

These days, I am just trying to focus on statistics of trade probabiltiy, controlling trade frequency, as well as looking at my P/L to make sure that the swings in capital are minimal.

More to follow . . .

T

[MBender]

Nice Charts

Brad:

You should be excited. Those are nice charts.

[Doji3333]

I am excited. I added one feature today called the "Range Limiter" that helps prevent sideways market trades (like today) outside of known boundaries/limits.

My system is designed to trade regular hours, so when we get a day like today where 90% of the move was premarket, I don't want it to be as aggressive if it's already beyond known limits. On days where we have more normal activity, then fine, let it trade.

Additionally, I've finally fixed the entry and stop issues and have cleaned up the code considerably. Hope to have it ready to go live before too long.

I would like to add though that I'm still a big believer of the "shut it off when it starts taking unwanted losses". I can't tell you how many systems I've seen that work great for weeks or months, then fall onto the other side and take losses for weeks or months.

My general rule is if you take a drawdown of 10% on your system, you might want to shut it off till it starts winning again.