Adaptive Strategies

[Doji3333]

Has anyone played with adaptive strategies in EFS? These types of strategies are designed to adapt to market conditions.

I've been playing with my newest adaptive strategy and just converted it to autotrade into IB yesterday. Today was spent fixing things and working on other's code - so no real progress.

But what's interesting is how the adaptive strategy is deployed. It tracks trends and noise as well as other features of price action, then adjusts certain technical indicators as needed based on the results.

I'm working with this as a scalping system and using some of my advanced indicators as part of the decision making process.

What I'm seeing is that the NQ adaptive model reacts differently than the ES. Thus I'm going to work over the weekend on a simple fix for this. Additionally, expanding the adaptive model to target only the initial 1/2 of the move appears to be better than letting it target 70~80% of the move. The objective is consistent profits, not huge profits.

Just curious if anyone has learned that efs can be build into adaptive models and if anyone has ever tried it? I may post some charts later.

[Doji3333]

OK. here's some charts.

Remember, this targets trades above the displayed levels at certain times. The objective is the the Min Point.

My adapting, this system is designed to target 50~80+ percent of the trend (in theory). The test is realtime trading.

So, I've modified the system and am realtively ready for Monday. But it all depends on my theory.

My theory is that the markets more is 0.5 to 2.0 pt ranges over time and that the breakouts can be identified based on an adaptive method with only 3~4 indicators.

If this works, it shoud grab $200~300 pts per bigger move (or more) and risk only 25% of the potential move (or less).

The trick is to figure out the relationship of my indicators and the price action. Quant trading at it's best.

[Doji3333]

more charts...

[Doji3333]

Update:

I've converted it back to a trend following system because it's easier to contain the logic and trades. Plus, this type of system should be able to do really well when the markets trend (as they've been doing recently).

Here are some charts and you have to remember this system adapts and should trade differently in realtime vs the 15 minute charts I'm looking at now. The biggest difference is where the system ENTERS the market. It should be much earlier than projected on these charts (meaning at a better price than shown).

The only mods I've made today are:
_ protection for runners. This was a simple fix that kicks in if the market runs more than 10 pts (for the ES and NQ). The idea is to try to lock in a protective stop if this happens and try to let it run.
_ mods to the adaptive theory that attempt to provide solid entries and avoid congestion areas.
_ converted it back to a trending model because it should be easier for testing and trading live.

Next, I'm going to try to add a "stamina bias" whereas the system will know to stop trading on flat/rotational days (or after a few losses).

Then, I think I'll have my new adaptive system ready to go.

I've decided to run it on 15 minute charts because it appears to provide good results and a limited number of trades. My theory is that I want to catch the best trades of the day, not hundreds of trades a day.

Here are more charts for you to review.

[Doji3333]

more charts.

[Doji3333]

Just added a simple fix to the parameters to handle the forex markets. How does this chart look to you?

Getting really excited about this and hope next week tests out the way I expect.

I love it when a plan comes together.

[Doji3333]

Well, after working on it most of the weekend, I had one major problem with the code today. It failed to handle the orders properly after the target was hit. Additionally, I had to alter how the entry orders operate so I could accurately mirror the historical results.

Right now, It's running LIVE with IB and firing orders (in simulated mode). It looks really good still - even though this is a bit of a flat day.

It got short right after the open and hit the target before getting stopped out near the entry price on both the ES and NQ markets.

Currently, the NQ is long and it just hit the target with one contract trailing. The ES is still flat from the morning trade. If the markets run UP this afternoon, I might see the ES enter a trade, but if not then it does not matter because I built this to try to capture the bigger moves and stay out of the markets when flat (like today). So this is a good test of the system's features and abilities.

Let's just hope we see some big moves this week to see if I've done my job well.

Good trading all and I'll post a pic or two later.

[Doji3333]

As with any new development, I've been making modifications to the system and trying to improve it. In doing so, I've created two different version of the code that I'm testing. The first is a minor mod to the original theory. The second is a new version of the trading logic with new filters.

Recently, I've added a simple "overall trend" filter to help prevent issues with flat days (like today).

I'll post more charts later today.

[Doji3333]

Here's the NQ chart for today.

[Doji3333]

Here's the ES chart for today. Still working out stop issues.

One of the biggest issues with developing automated systems in any platform is to make sure the system operates properly in REALTIME. When you test something historically, you see the END OF BAR results. But in REALTIME, things can operate much differently. This is why most automated systems require many days of testing and debugging.

The good news is I believe I have finally worked out all of the stop issues and tomorrow will be my test to see if I'm right. If so, then I'll watch this for a few more days and possibly turn it on LIVE later next week.

I have a few other minor modifications to make to it. There are some settings that are hard-coded into it and I need to make those inputs or parameters. Other than that, it looks pretty good.

Another thing to consider when running a robot or automated system is how you want to deploy it each day. Some days (like news days) can be really big winners whereas other days can be very flat. Simply searching Yahoo News or using the esignal news service can help you determine which days show greater opportunity.

Mostly, I like to try to run my systems every day. The only times I'll shut them off is IF or WHEN they start taking drawdowns. I believe I've mentioned this before, but any DD over 5% will cause me to rethink how I'm deploying the system. Anything over 10% and I may consider shutting it off till it starts hitting profits again.

All systems go through drawdown periods. I've seen people scale up as their system profits only to have the system BLOW OUT thru a drawdown period. The problem was the user did not understand that drawdowns happen and you have to be prepared and understand how to manage them.

Best of luck tomorrow all. I have a feeling it could be a big day.