I want to first tackle what the most natural question for me concerning pivots. How often are they hit?
First attached are some base line probabilities and numbers for eur/usd D1:
Every pivot that I attempt, if done for this type of statistic, is tested against these numbers. One of the interesting things about pivot work that I noticed as I worked on them was how many things were significant. By that I mean numbers too close to 50%, or too far from 50% are both significant in their own right. This makes the 20-40% and the 60-80% the areas that we want to avoid, but even these can be useful sometimes! Before I reveal the formula, take a look at the numbers for a minute and think about their potential usefulness.
For me, I see that for 20 pips, the probability is quite low that nother pivot is not hit. Therefore if I had some bias, I am safe to bet at least to that point. At the same time, if I randomly enter a position with a 20/20 tp/sl, I can expect to win about 50% of the time. On the other end of the spectrum, if price has already moved 60 pips in one direction, I expect it to turn around and move back down another 120 pips(60 to reach the original level, 60 to reach the other side of the pivot) about 5% of the time. I also expect price to be contained within a range of 120 pips 30% of the time. Hmm. Useful or not useful?
The pivot formula I used here is actually very simple. Current H1 close +/- 20, 40 or 60 pips, projected for the next 24 hours. That’s it!
Given the nature of volatility expanding and contracting not just over days or weeks but even months or years, sticky numbers like 20/40/60 are generally not good to use. However, a test for more recent data (perhaps 2014) will show if these numbers hold.
Ok, maybe this is not so impressive. But if the numbers can be skewed much further, would there be a use then? If we were able to change the probably that both pivots were hit from 50% down to 20% or 10% and bring the probability that neither pivot is hit down from 30% to 10% or 5% while retaining the size, would that be any better? When numbers are attached to the close and projected in the same manner for longs and shorts, The probability to go long is roughly the same as the probability to go short. I don’t think anyone expects pivots to create a directional bias.
However, the question to be asked is: When using double pivots, is there a point, if ever, that these set of probabilities become significant enough to be a good supplementary tool? Personally, my best results (the second attachment) are the best I can get for the moment, and my work and purpose for the thread will be on single pivots.