Rabbit Hole Series #2.1

Note: I’ve been through a lot of renditions of this already tweaking numbers and adjusting values; optimal numbers for frequency and success are still being looked into

4 Possibilities:
script
Either find a chance to jump in with price before the move is over, or anticipate the end of the current move.

Starting with the in-trend moves, or the oversold up moves and the overbought down moves. Within each wave, I’m looking at the number of times one of the above occurs within the set:

occurances
I was hoping that there would be much more of a cluster, rather than the range that we see (like 90% of the data contained with 0-5, but that’s clearly not the case). However, it’s just the case that sometimes you get something that looks like this:

seve
This is showing 8 occurrences when I think it should really be 2.. so filtering out the clumps..

1

Much better! My developing general intuition about using the number of occurrences before the “max” is that if you can’t cover at least 90% of the data within 3 or 4 points, it’s not very useful.

In these two scenarios(overfill/underfill), since movements are pinned against a wave, the only time it fails is if it gives a signal at the very end. In other words, The worst case scenario is if the price wave is down, and this TCD prints a overbought at the very end of the wave; inducing more sells at the worst price possible. For this TCD to have value, the number of these failures needs to be REALLY low. Trading TCDS is essentially statistically backed confidence, so I better have a good reason for being confident!:

post

A=The probability that the incorrect signal appears at the very end of the wave
B=The probability that the incorrect signal is the first time a signal is given
Therefore: probability that the first signal given is the incorrect signal= (.040*.348)=.014 or 1.4%. not bad..

currently there extra bars are in there are a sort of filter protecting (in live trading) against starting a new wave too soon, and also in hopes of potentially catching the new trend right as it’s developing. Each wave currently has an extra 2 dummy bars attached at the end of it. I think that I likely need to add a couple extra to help determine when the move is over, find a use for the other TCD, or create a new TCD (most likely) Example of one of the failures:

seve

What’s left is the biggest piece of the puzzle: How accurate are over/under fills in respect to price? Sure visually they consistently look similar, but how similar are they really? Tricky to answer..

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s