[swansjr]

Originally Posted by Lynnyhoot »

clayburg has also added a self adaptive indicator. have anyone looked to clone it also?

I've been watching his nightly videos where he has been demonstrating this new forward testing method. It's a great idea and I can easily see how this could dramatically improve the reliability/stability of the Universal. It address the core issue of Universal, the sensitivity of the four critical inputs.

For those not familiar, Clayburg has developed an indicator that essentially runs multiple virtual trading systems in real time. Each system is identical except for the four critical input parameters (B1,B2, S1, S2). He turns-on this indicator during the premarket session for about 60-90 minutes. He then can pick the best performing system's input parameters and uses them to trade live. All of this is contained within one indicator. The premise is the market's overall tone will be set early and by testing 20, 40 or even 50 Universal system settings on the live pre market data, you can determine which system settings will be more likely to perform well.

You can find the guys at TTM giving a brief demo here: Forward Testing for Tomorrow

It looks interesting and may have potential. My current opinion is the Universal is extremely difficult to trade right now. It may be done, but I have yet to see it. I can't believe they charge $7,500 for it!

Today Clayburg is giving a wibinar on the new "self adaptive" indicator. I'm recording it right now. Looks like the indicator will be released today.

Will there be a clone? Who knows?

[windsurfer]

The add-on is neat, but is an additional $4500! This price is ridiculous.

I'd think it could be coded by a TS user who is familiar with parallel functions. It merely runs the equity curve of 20 or so variations of the Mo and RSI variables on a single chart (e.g. the 5 min YM) so you can visually see which are performing best early in the trading day.

It relies on a major assumption- that the variables at the beginning of the day that are the best for a daytrading system will continue to be profitable the rest of the day. I am unaware of any research on this assumption. Any thoughts?

[swansjr]

Ah, yes! $4500!!!

Well, naturally Clayburg has been showing videos of running the add-on on EC, ES and a few other markets with noticeable improvements. But who knows how well it really works.

A poor mans way of duplicating the add-on is simply run 20 copies of the Universal with slightly different parameters. Start them all 60-90 minutes before market open then pick the best-looking system at market open. A parallel function is "simply" a copy of the Universal system. So, you can get the same result by running many copies with different B1, B2, S1, S2 inputs.

[insideday]

In my opinion the premarket solution isnt really a improvement, because you only get 1 or 2 trades in this premarket time range. Why dont he use the systems as meta indicator? If x % of the 50 systems give a buy signal then buy...

[swansjr]

Originally Posted by insideday »

In my opinion the premarket solution isnt really a improvement, because you only get 1 or 2 trades in this premarket time range. Why dont he use the systems as meta indicator? If x % of the 50 systems give a buy signal then buy...

The video's he demonstrated across several markets showed noticeable improvement. Of course he will be showing the best system impovements, but it is possible. Nothing is stopping someone from also extending the time earlier and/or later in the morning to capture more trades. But apparently even a few trades are good enough to determine the overall daily "mood." Furthermore, nothing is stopping anyone from updating the settings several times during the day.

I like your idea as well. Currently, it's not set up to function in that manner but it's certainly another idea. Kind of a majority rule. On the downside of such an idea, what if most of your systems are not performing well on a given day. Say 45 of the 50 are in the red. But only 5 systems are in the black. If you simply take the majority rule - well the majority can be wrong and you will get a poorly performing system.

I guess an alternative would be to take the top 3-5 performing signals and use their signals as a vote for when to buy/sell.

[indextrader1]

If you view Clayburgs free video each day you will see that the self adaptive indicator does not consistantly provide an improvement. Intuitively, it doesn't make a lot of sense to me because trading patterns during the regular session are usually quite different than the premarket patterns. Another problem with the Universal Trading System, besides the high price and the fact that even for that high price you don't get the code, is that if you turn on the interbar order generation option(IBOG), which you need to do to get the best results, each time you refresh the data or turn your computer off and then on again the next day, you will see that the historic results have radically changed, usually showing a marked improvement over the real time trading results. That makes optimization results questionable, although in my testing of the system it still seems to work OK, so far.

[windsurfer]

Using the clone and looking at optimized backtest results, my conclusion is that it will not be sufficiently profitable for the ES. For a 5 minute chart, I used 9,13,6,7 as the inputs to get a result that would barely cover slippage and commissions.

I had better results on the 113 tick EC using 23,11,5 and 20. I changed other inputs (session time, stops, etc) where appropriate.

Of course, what works now often will not persist for long. I'll autotrade it on 1 contract and see what happens. Today it lost $250.

For the time being, I'm not going to bother trying to adjust the inputs to the supposed best inputs for the day. In fairness to Dr. C, making these adjustments is in his view a necessary part of his system, and therefore my results cannot be validly used to compare with those of his clients.

Comments or suggestions?

[barryd]

I'm playing with the Universal Clone too and I'm seeing some strange results - can anyone else see if they get the same thing or give me an explanation of what I'm doing wrong?

I'm testing using a 5 min EC chart starting from 5/11 until the start of 5/18 (last 5 days) on 'natural' exchange hours.

When running parallel optimizations for the 2 Buy and 2 Sell inputs I get a max net profit (accounting for $5/contract commissions) of $4027 using the following:

B1:5
B2:2
S1:5
S2:4

Target 1: 5
Target 1 contracts: 1
Target 2: 15
Target 2 contracts: 1
Target 3: 50
Target 3 contracts: 1
Stop Loss: 150
=> No other trailing/day/equity stop toggles enabled (all set to 0) and start/stop hours left as default.

This is amazing for 1 week (unlikely to ever happen, of course) however when I looked at the actual trades I could see most of the trades taken were stopped out using the '$ Trailing Stop' Sell/Buy to Cover - surely this should not have been enabled if all the input toggles actually work?

Also, I also noticed a few anomalies when I played with the target inputs - I tried trading changing the profit targets as follows:

Target 1: 1
Target 1 contracts: 1
Target 2: 2
Target 2 contracts: 1
Target 3: 3
Target 3 contracts: 1

i.e. I'm only looking to capture a maximum of 6 points/ticks on the entire trade if I'm not stopped out.

In the EC, 6 ticks would equate to 6*$12.50 = $75, however in the 'Trade Log' I never saw any results for this. None of my trades ever achieved Profit Targets 1, 2 or 3 and all my trades were still stopped out using the '$ Trailing Stop' exit.

Most peculiar of all was that my Total Profit remained at $4027 and I found the only parameter that changed this was my 'Stop Loss'. Increasing the Stop from $150 to $200 increased the total profit to $4140 - plausible I guess.

Most interesting though was that putting in zero for the Stop Loss meant that NO trades were generated - why would this be the case???

Course, without being able to view the Easylanguage code I can't even begin to troubleshoot so I'm going to have to mail the developers, unless anyone else can help?

Thanks,

Barry