stevehopwoodforex.com
https://www.stevehopwoodforex.com/phpBB3/
Print view

TMA Discussion Thread
https://www.stevehopwoodforex.com/phpBB3/viewtopic.php?t=587		 Page 2 of 17
Author:  NeoTrader [ Thu May 24, 2012 8:58 pm ]
Post subject:  Re: Don't trust visual backtests with TMA slope
garyfritz wrote:I understand that. I just thought that some people might see that appealing indicator, and NOT see NB's note, and think they could create their own approach. I wanted to make it clear that what you see in history is not what you'll see in realtime -- which was not what I expected given NB's comment in his book.
Since everyone who is involved with this system is strongly encouraged to read NBs documentation it is a little redundant information. But you are right, people who only see the indicator may get the wrong idea of it.
garyfritz wrote:NB's approach works. Stick with that.
Yes i know...and how it works.. ;)

may the Pips be with us all... ;)
Author:  garyfritz [ Thu May 24, 2012 9:25 pm ]
Post subject:  Re: Don't trust visual backtests with TMA slope
NeoTrader wrote:Since everyone who is involved with this system is strongly encouraged to read NBs documentation it is a little redundant information.
And as I said, NB's doc implies that the slope indicator uses "TMA TRUE" and DOESN'T repaint. So even reading the doc isn't enough to warn people away from it.

Honestly I'm not quite sure why you're so concerned about me posting a warning. I'm just trying to avoid somebody getting burned by a repainter, as many have in the past. If everybody already figured that out, no harm done.
Author:  astral77 [ Thu May 24, 2012 9:41 pm ]
Post subject:  Re: Don't trust visual backtests with TMA slope
Thanks GARY. It is quite right and proper that you have highlighted this issue. We all know from our own past foolishness that once we see an indicator which does one thing well then the temptation (in particular with the people lacking experience) to think the holy grail is found is just too much.

It does repaint, there is no doubt about it. BOB says so himself. Although he says in higher TF`s like 4H and D1 the repaint issue is significantly less. So under no circumstances should it be used just own its own as a trigger indicator.

Kind Regards
Author:  nanningbob [ Fri May 25, 2012 4:09 am ]
Post subject:  Do trust the indicator.
You guys are missing the point of the indicator and what it is telling you. In both of his illustrations in post 1 page 1 the bars are above the 0 line telling you it is moving in an up direction. So there is no t a repaint issue. The bars change because the strength of the move changed NOT ITS DIRECTION. Therefore what it is telling you is valid. In both pics in his illustration it showed the angle of the TMA is going up. With the movement continuing up it broke the last swing high so the bars moved above the 4 line correctly telling you its strength. If you understand the indicator and that it is not the typical histogram of MACD, and others where downward slope of the bars indicates change of direction, that is not true of this indicator. Being above the 0 line indicates upward movement and how far the bars move from the 0 line indicates the strength of the move. You must understand how the indicator works. This is not a repaint issue and I wish you guys would throw that term into the trash. The indicator did exactly what it is suppose to do. 1. price is moving up because it is above the line. 2. the move is getting stronger because the bars are moving up. 3. The present reading is correct at the moment of reading, first pic the move has x angle. 4. The back test reading is accurate because that is what it did over the period of time. You then can compare it with the S/R lines and see how strength of the bar shows how strong it will move away from the S/R/Pivot lines. That is how you use it. It has absolutely nothing to do with repainting both pictures tell you in real time the exact strength of the move at the moment and over the period of time. When price was below its swing high or under the R line it showed a low angle. When price moved above it showed it was part of a strong angle. Both pics are accurate.

In a back test you see exactly the move as it played out. In real time you see exactly that price is moving up and what its strength is at the moment. That strength can and does change and allows you to know whether to stay or leave the trade.
Author:  garyfritz [ Fri May 25, 2012 4:29 am ]
Post subject:  Re: Don't trust visual backtests with TMA slope
Makes sense. I believe I understand. Thanks for the clarification, NB.
Author:  NeoTrader [ Fri May 25, 2012 10:52 am ]
Post subject:  Re: Don't trust visual backtests with TMA slope
garyfritz wrote:Honestly I'm not quite sure why you're so concerned about me posting a warning. I'm just trying to avoid somebody getting burned by a repainter, as many have in the past. If everybody already figured that out, no harm done.
I am not concerned about you posting a warning.
You are a valued Member here and i do appreciate your work here and am grateful for what you do here.

It was only that this topic of repainting comes on and on again and nobody really seams to grasp that this is not an issue with this indicator.

As NB tries to clarify it again and again the current value of this indicator is the only thing that really matters for trading the system as it shows the strength and direction of the current Movement.

So it shows us when to look for buy/sell opportunities based on the strength and direction of movement. This indicator is not meant as the only trade trigger (but Shelley shows us that it can be used as such).

But you are right that this indicator can not really be used for visual/manual backtesting as it repaints a lot of bars.

Happy trading,

NeoTrader
Author:  idempotent [ Mon May 28, 2012 11:03 am ]
Post subject:  I think we can make the TMA Slope not repaint
First, pardon me perpetuating the issue. I have a lot to learn.

By consensus of this thread and other doco, it seems to me the "repaint" issue is real when backtesting visually, but not when trading live.

But visual backtesting is commonly enough done that it's a sticking point for some of us including me.

The TMA slope indicator will cough repaint 20 bars back. See the first attachment for a strident example. It shows an M1 chart over the weekend gap.
repaint vs no repaint.gif
As it appear looking back with the normal TMA slope (first indicator), it indicated that things were going down 20 bars before the market closed. The second indicator is my first attempt at a non-repainting TMA slope, which shows an expected spike down in the TMA slope corresponding to the open gap down. Ignore the third indicator - that's a work in progress.

We can fix this. We can make the indicator not repaint. I've half done it; it works in "realtime" - once you apply it to a chart, every bar after that does not repaint. But it's not working for history yet. That is, if we apply it to a chart with bars in the past, no histogram is drawn for that. I'll look into that; I can't see why it can't be solved.

The second attachment just shows historically what a difference it makes. Again the top indicator is the unmodified TMA Slope and the second is the non-repainting version.. Again, look back, the TMA slope is brilliant, picking the movements seemingly before they happen. The realtime reality is somewhat different. It indicated "up" 14 bars later than the unmodified TMA slope, and had a false positive a bit later, that the unmodified one had erased.
repaint vs no repaint 2.gif
My point: if it's a big enough stumbling block, and we can fix it, then perhaps we should.

This is not a criticism of 10.2 or NB.

Tony
Author:  garyfritz [ Mon May 28, 2012 2:23 pm ]
Post subject:  Re: Don't trust visual backtests with TMA slope
Great work, Tony! I wanted to do this but just didn't have the time.

I think it's valuable to do this whether we use it in realtime trading or not. It shows us what we would have seen in realtime so we can have appropriate expectations when we start trading forward. It also lets us visually backtest the system, helping us to learn and understand it better, and possibly improve the system by seeing how it performed in the past.

So far we're working with realtime demo/live trading using the repainting indicator. Judging by peoples' delighted responses, that seems to be working well, in spite of the not-great realtime TMA values as shown in your indicator. But the TMA history is deceiving.

If this indicator shows the same values in realtime that the original TMA shows, and if it shifts those realtime values into history without change, then I'd strongly recommend using this indicator instead of the original TMA. It shows the same realtime value for trading, so nothing changes there -- but the accurate history is very valuable.
Author:  chriseel [ Tue May 29, 2012 6:24 pm ]
Post subject:  Re: Don't trust visual backtests with TMA slope
What could be really helpful backtesting NB10.2 system
is adding a "BarCount" feature to the slope that shows how exactly appeared before x count of bars.

Is that difficult coding wise?
Author:  nanningbob [ Tue May 29, 2012 11:21 pm ]
Post subject:  Re: Don't trust visual backtests with TMA slope
The back testing should be based on the pivots/S/R and the 5MA lines because those are your real trade triggers. They remain and dont move. That is how you backtest. The most important thing you will see in a backtest is when price crosses a pivot it almost always continues to move for a profitable trade. When you compare the position of the weekly pivot with the monthly pivot you will learn even more. Focus on your trade triggers not on the strength and you will comprehend the system much better. The strength indicator changes from moment to moment your trade triggers do not. The only thing the slope is going to tell you is this: this is strong stay with it or this is not strong bail out.
All times are UTC Page 2 of 17

Powered by phpBB® Forum Software © phpBB Limited