Consistant profits

[MBender]

I'm finding today a descent trading day.

I currently feel I'd like to make $10/trade consistently rather than make $100 on one trade and -$90 the next. Thomas -- I suspect that means I'm buying into your approach.

That makes consistency more important than big-hits. That said, I understand the difficulty with consistency.

I anticipate the next challenge to be how to avoid throwing away the larger gains for the sake off consistency. A much simpler challenge when the time comes.

One war at a time.

Cheers,

[Doji3333]

I'm trying to revise my adaptive strategy to achieve an 80%+ accuracy ratio (or better). I spent all weekend (well most of it) trying to develop a means of identifying the best trading opportunities with different filters and conditions. I believe I may be on the right path now and am testing a new version of it today.

I am working on the code with the basic concept that targeting 1~2 pts per trade and targeting only the highest probability trades is the key to overall succes with this model. I've seen it hit 5~7 profitable trades a day, but then there are those days when the markets are really flat and it gets chewed up a bit. So I've installed a loss limiter that should help the overall results.

In backtesting on the 15 min chart, I've seen over 90% accuracy, but the real test will be LIVE/RT trading.

Let's try to keep these threads going because I'm sure there are many lurkers that are folling our efforts.

[arbifox]

The trick is . . .

I am working on refining the 7th version of my program after I had a terrible Q2 and flat Q3 this year. Interestingly enough, Q1 was awesome. The system is up about 5% YTD through the end of Q3 . . . but you can always do better is the motto of system traders.

This year, I had a bit of a revelation. Take a look at RYMFX. It is a managed futures solution from Rydex, down 10% or so for the year. Why? Their strategy simply did not work so well this year, but it might have been fine last year. I think that is why it is important to look at 5 year histories on Mutual Funds, since part of what the Market is and finding success is to find something that works in all types of markets. My revelation as a result of this came from understanding that it is simply not possible to find one strategy that works well in all markets. So how to work around this reality?

My approach is to develop a base system and to carefully monitor the trades that are taken, and the starting conditions of those trades as measured through the key indicators that determined whether or not you take a trade or not. This approach will eventually let you know which exact set-ups produced the highest probability of winners, and that I think is the best "trick" to consistent profits. You simply need to find the starting conditions that yield the highest probability trade, not the most number of trades. Remember, more trades, more commission.

All of this is also simply not possible without a back-tester, and one that you can rely on and be sure that they account for slippage on fills and commissions.

Too many times I have heard of traders second guessing their programs or thinking, "If I was trading, I would have never taken that trade", but what is missing is the understanding that a program sees only the set-ups that are programmed. I never compare the trades that the program makes with what I would do as the human brain adds so many layers of thought to a trade that it is impossible to simply program.

I have also tried "target trading". Never have I seen such disastrous results as to when I tried to get out of a trade early based on a pre-determined target, only to miss a huge continuation of the original trade direction, or the program takes a counter trade prematurely and wipes out the modest gain on the first trade by hitting the stop on the second counter trade.

So this is what I am working on. I rely on multiple time frames to sync up and give me the highest probability of success, and I then work on filtering those trades even further to be as selective as possible. Keep in mind, a computer program is infinitely patient, unlike their human counterparts that just want the thrill of the next quick win.

[merkro]

There are indeed 'Lurkers', and I am probably only one of hundreds. This is a very interesting thread, and one I would like to contribute to, however the fear of displaying one's ignorance is always daunting to someone who has been 'doing it alone' for a couple of years.

I've found it relativly simple to pick entry/exits in a trending market, and almost as easy in a sideways channeling market, but never have I been successful in devising a single method to cater for both. And the problem is how to tell the difference. I spend a considerable portion of my 'tinkering' time on trying to determine if the market is trending or going sideways. I find that by the time I have determined direction, it is most likely changing. (I prefer the shorter time frames)

Using ES as an example, I have had minor success using ATR(10) > .75 to indicate volatility for a scalp, however on some days it's possible to have a 2 or 3 point channel running with an atr smaller then .75 points.

Most of my experimenting is done on a 2 minute chart, using multi time frames of 5 to 30 minutes, however I favour the 10 minute frame. My jury is out on the longer time frame signal being confirmed, as if using an indicator of 10 minutes on a 2 min chart, then lots of noise can occur during the 5 by 2 min bars completed while the longer time frame bar is forming. If I wait til it is confirmed, then I can miss the entry by up to 4 bars of my working chart.

I'm currently playing with spaced out multi time frames for entry (10 and 20 minutes aligned) and times close to the 10 min for exit (8 or 9 minutes). I use a hard stop only for safety in case of power failure etc, and it is quite a long way from entry.

Have you guys used the $TICK for entry signal? My experience is that it's the only proactive indicator available. As an example, place the macd(5,34,5) on a 2 min ES, and a 2 min $TICK, and the $TICK will most times react a good bar before the ES. I have found it is good for scalping only though.

With all the work I have done, if you asked me about any particular indicator's worth, I could probably tell you why it is not the answer. I look forward to one day being able to say the reverse. I also look forward to reading the next post in this thread.

Cheers

Merv

[arbifox]

Trading vs. Trending . . .

I can understand Merv's comments about determining whether the market is trading or trending, and adopting a strategy to work with that type of a market. I have tried hard to avoid trading during times when nothing is going on, but it is really a pickle to get that logic to work right.

I have not really worked with $TICK too much, as I find that it is really most useful for day trading very short time periods. As far as time frames go, anything past 5 minutes in this market I find too slow and leads to way too much lag. To solve the issue of timing lag, I usually allow slower time frames to be positioned in such areas in respect to overbought and oversold that they justify the potential of the trade, even if they have not yet turned and verified the signal.

If you use multi-time frame trading, then I have found that the agreement of all time frames will keep you out of most of the worst trades. I use 150T, 450T and the 3M on the Russell Futures, and I also use overbought and oversold zones on my base indicators which keep me out of trades as a result of low probability of success determined through backtesting. This methodology helps a lot, but is not the end all cure all.

The problem that I have with knowing which market is which, is that a virtually flat market can start to trade all of a sudden, and then you may get a great day. So I always approach each day as the potential for a great day, so I am always careful to not miss a day of trading. If I had a dime for all the days that I have missed as a result of one technical issue or another and has turned out to be one of my better performing sessions. The message here is that if the market is open, you should trade the system and let the chips fall where they may.

Lastly, you must have some type of risk management. I allow the system to make 2 trades a day that can go bad, and each trade has an initial stop of 2 points on the Russell E-Mini. This means that at worst I have the potential to lose $400 any day, and I usually trade $10K per contract. I have tested these settings and have come to realize that they work the best for my drawdown tolerance, but this is a concept that can be developed by anyone to fit their needs specifically.

As a final resort, you can always develop trade filtering based on either the time of day, or starting condition of a trade signal. I use this to some extent, but I have found that it can limit the trades you take under normal conditions and may prove less profitable over the big picture. However, if the lower profitability results in greater program stability, I am all for it.

So given these concepts, so far these are some of the rules I follow:

1. Let the selectivity of entrance criteria control trading frequency.
2. Watch not to trade into obvious overbought and oversold zones of base indicators.
3. If the system is on, it should be on as long as the market is open.
4. Measure the performance over the course of a year, and do not let bad weeks or months affect judgement of the performance of the system. Back-testing in this area will greatly increase confidence in the system and prevent changes which may help shorter periods of performance while harming larger periods.
5. Use stops to limit ultimate losses.
6. Investigate the use of trade filtering.

Just my musings in the middle of the night, all comments are welcome.

Thomas

[merkro]

Hi Thomas, I compare your notes with what I do, and find some similarities, and some differences.

"1. Let the selectivity of entrance criteria control trading frequency."

Agree, however part of that entrance criteria must be current market tradability. There is a version of the macd on the forum which takes the average of the last 5 histogram values as the signal line. During periods of low volatility, the histogram narrows considerably, and on my to do list is to take a sma of the histogram, and compare the last sma value with that of maybe the middle value to see if the histogram trend is holding or declining. (Confused? Say a sma(30), and (0) < (-15). Have used this idea with success on sma's on Tick for day trading direction.)

"2. Watch not to trade into obvious overbought and oversold zones of base indicators."

Haven't used this as strictly as I should have. Have always adopted the view that as all indicators are reactive, they change after price has changed, therefore if they are still indicating up then it's a good thing regardless of where they are on the scale. Reading your line, and putting a bit of thought into it, my view has been way off mark, so I'll be hitting the keyboard tonight.

"3. If the system is on, it should be on as long as the market is open."

If "current market tradability" can be gauged with a moderate amount of success, then the system should be on 24 hours a day.

"4. Measure the performance over the course of a year"

Ahhhh, my weakpoint. I have never had any joy with back testing, having made many billions in backtest mode over the years, only to have everything come undone in real time. Have not devoted much time to eSignal's back testing mode, however at first glance it does not seem as easy to use as that in MT4 or Market Analyser, and they did nothing for me. Think I'll continue with 'live testing'.

"5. Use stops to limit ultimate losses."

Here we differ. I have found losing trades average much less when using indicators for exit criteria, as opposed to hard stops. Many times I find hard stops have cost me much profit. Perhaps this is a reflection on my lack of refinement on entry point, and if I can improve the entry, then I may reconsider. Something I have had both success and fun with, is having a hard stop some distance from entry, and dragging it up tick by tick behind price. On some occasions, I have left it at break even once reached, and I have also played with dragging it at a 50% level of profit reached. Mostly though, I prefer to have a hard stop some distance away for the sole purpose of loss of internet, power outage etc.
I use eSignal only, and do not have the luxuries of entry confirmation, entry price etc. from within my program, which makes close stops and profit levels a bit hit and miss. I'm told now that as long as priorities don't again change, I may have those little luxuries by last quarter 2011.

"6. Investigate the use of trade filtering."

Not entirely sure what you mean here, as trade filtering is the act of entry criteria. Your trading time frames are much shorter than mine, as I rarely work below 2 minutes. In fact, trading 2 minutes with confirmation for entry from 5 min, and direction supported by longer time frames as far out as 30 minute, suit my style well. A moving average on a 2 min Tick will generally give an indication of where price may go. (May being the key word here). The direction of the ma is important. The fact that the Tick is averaging +400 means nothing if 15 bars ago it was averaging 600.

Hope I haven't rambled on too much here, or showed my lack of knowledge in too great a detail. I enjoy the challenge of system design immensly, and can talk about it indefinately.

Cheers

Merv

[Doji3333]

I made some changes to my new adaptive system and posted a new chart today. Check it out if you like.

I'll be posting more as I continue testing. I'm pretty happy with what I have now though. Just want to make sure it does what it's supposed to in RT/LIVE mode (auto-trading).

Pretty happy that is held it's own today in a very flat market.

[merkro]

Brad, I've studied each of your charts closely, and am impressed, especially when you start mentioning 80+% hit rates. Since your first post on the subject, I have had the word 'adaptive' at the back of my mind. Though it's great to see your successes, I hope you find a way to get us started on 'adaptive strategies' ourselves, without giving away any trade secrets. Can you throw me a bone?

Cheers

Merv

[Doji3333]

I posted an example of how to make an efs "adaptive" over in the EFS Studies thread. It is in the topic "variable question".

It shows a simple example of how to accomplish an adaptive strategy with RSI. Nothing too complex, just the basics.

[merkro]

I'm not much for keeping records, and if I do run something thru a backtester, I normally commit the results to memory and forge ahead on the latest idea for improvement. I'll never be a success because of this, however as long as I'm having fun......

Hope the spreadsheet I have attached works. This is the result of 2 months of a simple 'Direction forcasting' idea I had. It attempts to pick the direction of the ES between close of one day, and open of the next. The entry is calculated at close at 1600 hrs, and exit at open at 0930 the next morning.

It uses 2 indicators of 3 time periods, giving a total of 6 points if all indicators are positive. Where you see a signal of 'none (WU)', it means that there is no signal, but 5 of the 6 pointed to up (or WU for down). I have been keeping a running total of both weak signals and confirmed signals. The first few days I just jotted the result on a piece of paper, hence no open/close values.

There was no sound basis for the theory other then the expectation that the bulk of traders would possibly carry thier sentiment overnight from close to open. That is, if they felt bullish as close, they would maybe wake up still feeling bullish the next morning. In any case, I still suspect the results are simply a coincidence.

But at what point do I consider it beyond a coincidence? Close to a grand a week for one trade a day of one contract is not a bad return for effort.

Cheers

Merv

[merkro]

This attachment may be a bit easier to read. It wouldn't let me upload a .xls file......

Cheers

[MBender]

Joining a group

BTW: I'd love to join a more private group. There are many discussions I'd like to have, but shouldn't in a public forum like this. I have been trading for over a year, but I would miss Merv's involvement. You can reach me personally at [email protected].

Cheers

[arbifox]

I have not forgot about this thread, I just got swamped with work. I want to update my spreadsheet, and upload, but I think it might be futile to try to convert into a PDF.

Does anyone have a blog space that we can use instead of this thread . . . I too have some ideas I do not mind sharing, but I rather do it in a more controlled and private space.

I can also send out my spreadsheet, just send me a mail to arbifox <at> hotmail, or we can just continue this through mail. Brad and I have already chatted about private work a few times, I would just love to have an area where we can post up collective ideas and see how they work.

If not, later in the day I will have a chance to go through all the threads and post more detailed responses.

T

[Doji3333]

I tried to get a think tank of skilled investors together years ago. The concept was that many people have great ideas, but very few know how to program those ideas into actual code (systems).

There are also so many tricks to running a system. I've heard that active ES managers are targeting 10~15 pts a week. I don't think this is impossible, but I do believe you have to take what the market will give you each week and not be greedy.

I've developed my systems to try to "take the easy way to profits". In other words, I'm not targeting HUGE gains every week. I'm just targeting the easy profits and calling the day "done".

For example, I would not have traded yesterday (as it was a holiday) and my system proved I was correct (taking two losses). But today, it's back on track with two winners and done for the day (+$200 trading only one contracts).

I've sent some PMs because I have a solution for a private room to discuss our systems (if needed). I still like the idea of posting more general information here on the esignal forums. There are many lurkers that want to learn from our experience.

BTW, my new adaptive system was flat yesterday (+$30 roughly) and is so far +$80 on the ES with one long trade. The objective of this system is nice, easy, consistent gains (I hope).

More later.

[MBender]

I agree ...

While I'm not knowledgeable on what ES managers want, I agree with you on all other counts.

The system plus the strategy -- that's the trick isn't it? You have to know how to trade, and you have to know how to develop a system. For me, developing the computer system is easy -- I've been doing that all my life. I just need to figure out trading.

I set up a private, invitation-only blog. e-mail me and I'll send you an invitation. I also completely agree that public information should be done here (note my post earlier today).

I'm playing with my mud indicators today. I'd really like your comments.

Cheers,
Michael.