I think I’ve found my trick in the thread; now it’s time to try to turn it into a full system. The trick always involves me going counter-trend, but it’d be nice to go in the direction of the flow. Little scalp picks are great but knowing the general flow and going back to the ‘follow the trend’ idea is in general best if you can manage it.

There’s some important differences between transient bars that have their tips never touched, and mid bar transients. The first generally marks all relevant reversal points and the aim is to always trade counter the bar in question. The second marks the best points to place a moving SL, or an add on buy. Never forget that the worst place to trade is always the best place to trade the other direction.

Reversal transient bars tend to have much smaller transient zones than mid bar transient bars. Quite a big difference really.

Edit: Added filter. This looks good

### Like this:

Like Loading...

*Related*

Relativity‘How Convenient’

๐

4xBonesI suppose this implies that as soon as a narrow transient zone starts to form, we are likely to see a reversal. Also, a wide tz (mid bar) is more difficult to confirm until -/+ h bars has run it’s course?

4xbonesprev: was typing on mobile so being conservative with words

What I was trying to say;

1) Your stats suggest narrow tz associated with reversals

2) Wide tz with mid bar zones (= continuation?)

However, a wide tz is hard to predict before completion of “H” bars, because at any time it could become narrower. But…. a narrow zone cannot become wider. Are the stats well and truly on your side in this case?

lgtraderPost authorYes: small TZ=reversal, large TZ=continuation. However, it is difficult to tell the difference between the full reversals and the left-side transient bars (that will eventually become recurrent and also a continuation), at least for now. This is an issue!

Also, with the filter I’m using, I can judge my odds of the zone that is being witnessed is a mid-bar T-bar or not long before the h has completed. In other words, the TZ sizes being shown are not the true sizes (if true is the remaining TZ after h is completed)

Patrick M. WhiteTwo comments:

LG: I find it “interesting” that a “filter” can increase the # of occurrences. Don’t filters typically refine or reduce the population by some logic rule?

From my experimentation it appears that if you plot the increase of h against the increase of the “fill in” probability, such as the line from kprsa’s recurrence stats indicator, “Probability of fully resolving possible fractal bars”, it appears very similar to a logrithmic curve. You get diminishing increases in “fill in” probability for a linear increase in h. I bet this is consistent across timeframes.

For instance on a M5 chart, it takes an h of 10 to get a fill in probability of 70%, and an h of 35 to exceed 80% for fill in probability. An h of 100 only gets you to about 87% fill in probability.

lgtraderPost authorThe reason why filtering causes the number of occurrences to increase is because in filtered histogram I combined the top and bottom transient bars into the same one. To be more specific, The top left and top right chart are for transient zones/bars occurring above price, whereas the bottom two are for bars occurring below price. If you add all the numbers in the two ‘tip’ histograms in the first picture, you’ll end up with the sum of the numbers in the tip histogram in the second picture, and likewise for the TZ.

Concerning the “fill in” probability I agree with your findings. This is why I think in a lot of cases a small h value is a good thing. If we were to follow eurusdd’s advice on the 97/3 rule then I would vouch for picking the lowest h value that meets that rule. After all, if we’re trading based on the occurrence of left-side transient bars, the more frequent the better right?

Thanks for the comments (:

Patrick M. WhiteLG,

Thanks for the clarification – that makes more sense. Also thanks for the reply to the 2nd point. I guess I should be calling “fill in” probability, “hitting” probability as per Eurusdd. So there seems to be a probability that can be calculated, depending on how you want to think about it – the probability that price will hit (or “take out” / “fill in” a given transient area), or the probability that price will not hit. Both can be calculated within a certain # of bars. Of course, they are the logical opposite, so really it is just one calculation AFAIK. The “within a certain # of bars” is an interesting thing.

For instance, if you have a transient fractal bottom with an h of 20 that is both left and right side complete, you can compute the probability that price will remain above that bottom for the next x bars, where x may be a value less than or greater than h. And since on time charts bars ~= time, you could make a probabilistic prediction within a certain amount of time with regards to price. Maybe I’m alone but I think that’s pretty remarkable!

Thus it is possible to make calls such as: price will stay above y for the next x bars with p probability. That’s a rather interesting concept!

lgtraderPost authorThanks for the reply!

In regards to the ‘interesting concept’, now that you phrase it that way, it sure lead some insight to how eurusdd makes calls for his levels, each one with x probability. I have not done much work with the levels after the h has already completed, although this does seem worth looking into. Although given the current EU move, my guess is that most levels that are broken tend to stay broken for quite some time.

Patrick M. WhiteYes it makes more sense now how these calls are being made.

So we have all these TZs all over the chart for a given h. Each either acts to attract or repel price, and we can calculate the probabilities for each of them independently for a given timeframe and for a given # of bars into the future. And we can do the same on multiple timeframes to get the big picture for a given h. Some of the probabilities must be updated with each new bar because for each bar that elapses the numbers shift ever so slightly. Is it possible to combine all this info and have an accurate composite directional probability for a given market that updates dynamically?

I just saw a short term TZ on M1 that was supposed to repel price, get taken out. M1, M5 and M15 are all being attracted lower for now, and it seems to be going there! Now the low seems to be repelling on M1 M5, while M30, M60, M240 attract lower, who will win? Will M15 complete its bottom TZ fractal in time to stop this down move?! ๐

Patrick M. WhiteFYI, I’m using the term “TZ” to refer to all transient bars – both trend and reversal TZs.