PatternCast

2022 January 12, Wednesday

(1) Question: When you indicate the number of positive and negative trades, their percentages are calculated assuming to open the position based on the signal until the close of the session. On a day like yesterday there were several opportunities to exit the position with a great gain but waiting for the closing the gain would have been different (in this case better). There could therefore be big differences between theoretical statistics and real operations. Do you have statistics on real operations? Or, in other words,do you have any statistics assuming you close the trade for example if you gain 1% or if you lose 0.5%? They could be useful for set a correct stop loss.

Answer: This question is so good, so deep. This gets to the heart of why PatternCast exists. The very foundation of PatternCast is built on the answers to these questions.

Let me start with a dry, “by the numbers,” answer: All PatternCast back-testing shows it was less-profitable to cut winners short. There is no other way to put this.

Over and over, we found two things were true.

The first truth we uncovered was there was no way to cherry-pick times to get out before either a target was hit or the close. For every time we found a market gap-up – with a PatternCast “buy” signal – where our intuition would say, “Take profits early to keep our gains after an overnight gap,” – and it seemed to work, we’d find another time where letting a trade run put the most money in your pocket. The truth we found is: there was no non-arbitrary method to cut trades short of the “max potential,” and return anything close to a “hands off” approach.

Does this mean the future will be the same? No. Clearly, not. Maybe the decades of data were simply a reflection of those times. Maybe the signals now only SEEM to work best with the parameters we use. But, these are not concerns we can solve without them being solved in hindsight.

The second truth we found was the system, algorithms, formulas, and, the very essence of whatever it is that seems to make PatternCast work, do NOT include stops. We tried to find a stop-loss methodology that was not arbitrary. We tried to find one where there was a percent drawdown the system would accept. Yet, like the target issue described above, what we found was only an illusion – what seemed to work over any period seemed to break down over time.

Then, we had what we thought was a breakthrough: We found that if any signal was under water at 10:00 AM EST, it was a better idea to stop that trade then hold for a “max potential” return or the close. This was a difficult thing to test, so we first tested about 200 trading days, and, it seemed like a serious contender for that Holiest of Grails; the “what is the best stop loss” Grail quest looked like it was complete. The overall equity curve on returns moved up about 10 to 15% over the period, and, most serious drawdowns were snuffed-out.

That was all until we tested the next 300 trading days. After that test, the equity curve not only reverted back to the original, but, actually lost some more and would have turned into a less-profitable set of returns.

Why was that? Well, come to find out, there were many days where the early moves against the original signal reversed. And, on THOSE REVERSAL DAYS, big profits were left on the table with that 10AM stop.

Let me stress how important I believe our findings are. If all this was purely digital; if all this was just numbers; if chaos could be solved at every level; there might be way to figure out how to stop the progress of a moving thing like the market. Yet, after thousands of tests, over millions of patterns, and BILLIONS of potential setups it’s clear to me some things are unknowable.

Well, maybe not unknowable – more like unstoppable.

If PatternCast finds a pattern indicates a direction, well, try as we might to find ways to cut short that progress, it will go the way it goes. If we choose to cut out early, it looks like we miss the big moves.

See, PatternCast finds patterns that indicate where things will END UP on the day, but… and, PLEASE read and understand this… BUT, HOW it gets there is a mystery. Like water to the sea, that flows down a mountain, we KNOW the water will find a path of least resistance to get to the ocean, but, we CANNOT know every rock in the way. To try and predict the PATH water takes may mean more chaos than all the “super computers” in the world can solve.

Yet, we CAN find evidence of the DIRECTION the water will take.

If you have further questions, please email patterncast@gmail.com. Thanks!

These questions were sent about the PatternCast “Quad Scan”. These answers specifically apply to the “Quad Scan,” and generally apply to other PatternCast scans, insofar as we understand them.

(1) Question: Does the PatternCast Signal give a simple, straight-forward “buy at the open” because PatternCast expects the market will rise from the open and close higher?

Answer: Each pattern seems to indicate properties for the next trading session. Our interpretation is “open to close.” The use of the word “expects,” adds a certain human quality to the equation, though, that we don’t use. There is no “expectation” by these algorithms because they are not psychic, have intuition, or are imbued with hope or fear.

The way we look at the algorithms’ output is historically-significant evidence that a statistically-likely set of parameters are probable. This is our best attempt at the moment to describe what PatternCast does.

Consider this, “If a pebble is dropped into a pond, can you predict how far, and where, the ripples will spread?” You could spend a lifetime on that question. PatternCast attempts to quantify what come before the pebble hits the water – the size of the pebble, the distance to the water’s surface, the wind speed and direction, the movement of the water below. The result is output that describes most-likely outcome scenarios.

(2) Question: Is there experience from prior signals the market will move down near the suggested entry point before the move up?

Answer: There is evidence and a great desire to find more. With our current testing the best results with these signals comes from an “at the open of the next session” approach, though.

We find net returns overall, and winning percent, favor a “wait to make a decision” for the open the next session. Yet, the difference we find for trades taken at the close of one session, or the open the next session, is slight enough that it could be statistical noise to assume the “at the open” scenario is best.

More testing should be done to explore this concept.

PatternCast Quad Scan “Raw” Versus “Filtered” Results

(3) Question: How significant is the suggested entry point?

Answer: The word “suggested,” needs to be defined. “Suggested,” is not used as in, “It is suggested you take vitamins for optimal health.” Rather, “suggested,” is used as in, “It is suggested – by data – that over the last 100 years, people who took vitamins got sick less often.” The distinction is, we don’t make suggestions for you to personally do anything, rather, the data (and, our interpretation of it) suggests statistically-relevant parameters to consider.

Thus, we now look at “suggested entry” prices in context: The “suggested entry” prices are based on filter sets we apply to the “raw” PatternCast signal. This is where we get the terms “raw signal” and “filtered signal.”

And, yes, we do find significance with “suggested entry” prices. We find higher win rates, larger returns, and smaller losses per trade. The tradeoff we find in the data is less trades triggered in our current understand of these signals.

The gross returns were larger with “raw signals,” and “at the open,” as they were created, but the more efficient trades we saw utilized “suggested entry” prices.

Either way, over time, we saw a move higher in the total returns. It seems to us the decision to take one at the open, or wait for an entry is personal, and, goes beyond the scope of PatternCast.

(4) Question: Does the market tend to gravitate down to these areas in the pre-market? Is there a futures opportunity to the downside before market open?

Answer: That is something we tested, but not enough to make a hard, definitive answer. We see some people who seem to have good results with pre-market, or overnight, trades – where they either sell into a buy, or buy into a sell. But, to be very, excruciatingly clear, there is no hard data we have to back those ideas.

If you have further questions, please email patterncast@gmail.com. Thanks!

Questions #002

Questions #001

These questions were sent about the PatternCast “Quad Scan”. These answers specifically apply to the “Quad Scan,” and generally apply to other PatternCast scans, insofar as we understand them.

Pages

Recent Posts

Categories