One way to build an indicator from the COT data is to combine the two non-commercials (long term and short term specs) and plot them against the commercials. Then it is quite simple to turn those two plots into an RSI oscillator, probably looking for rough cycles to use versus a random 9-bar period or something. In many markets you do often see the Commercials coming in heavy towards the end of a sell-off and they are often the first to start easing out during a large run-up as they gradually distribute what they accumulated earlier and gradually begin accumulating short positions. I find they give the most helpful readings when they are at extreme bull or bear but in the middle they can waffle around for many months, sometimes a year or so, as there is no big wave on their part one way or another. But when there is a big wave, it is best to be aware of it and pay attention because nearly always there is a sustained move - multi-month - afterwards.
@shodson yes true...not sure how at this point how this is being used. There seem to be very complex formula's revolving around the same.
This is something new to me as well. But again strongly believe that everything in the markets is inter-related & so must be COT in some level as data is available across all asset classes. Well will report back if I find something.
Broker/Data: Mirus (Broker), Continuum (Data), Dorman (Clearing)
Favorite Futures: Futures
Posts: 202 since Mar 2013
Thanks: 426 given,
I've been reading a bit on how Larry Williams trades the COT data, and although I am not quite done with my reading on this, he puts the COT movements into a relative picture by contextualizing the weekly values as a function of where the weekly values fall within a 3 year max to min band. So for instance, if this weeks value was 65 and the 3 year max was 150 and the 3 year min was 15, he would say that 3 year range was 150 - 15 = 135 and the current week's value in relation to the 3 year min would be 65 - 15 = 50, so the current week's relative position in the last 3 years would be 50 / 135 = 37%.
He then looks for extreme COT positions. So for instance, when the Commercials current COT value yields a relative value that is 75-80% of the last 3 years, he says to change your bias to longs only--and less than 20-25% of the previous 3 years as bearish bias.
He uses the Nonreportables and Non-Commercials as contra-indicators (ie, when they are 75-80% of the 3 year range, change your bias to shorts--and when they are 20-25%, change to long bias).
His idea is that the Commercials are the ones that really move the markets (ie, start the trends), that the Nonreportables (ie, retail traders) are wrong 90% of the time (ie, 80-90% retail trader failure rates) and that the Non-Commercials are the large funds that typically trade trend following systems with volume--with the trends only being initiated by the Commercials.
From the 10,000' view, this is how I understand his idea for how to use COT data:
Commercials enter positions for non-speculative reasons (ie, because they use the instrument in their daily business). After the Commercials demand starts to materially eat into available supply (supply/demand economics), prices start moving up, which triggers the Non-Commercials' trend following systems to kick into gear, which further drives prices up. But the Non-Commercials don't enter all at once. They pyramid their positions as various X-day or X-week highs are hit (--although he doesn't come right out and say it, he's basically purporting that the big funds follow a Turtle style system), at which point they eventually eat all of the available supply. The Nonreportables see the price moves initiated by the Commercials and moved by the Non-Commercials, so they get on the train (late). Prices then fall because the Commercials are no longer buying (because they've already bought what they needed to at the prices they needed to make their business profitable) and the Non-Commercials are done buying (because price has started to move sideways and their trend following systems are no longer firing buy orders). So at the bottom of the move, you would have seen the Commercials relative buying in the 75-80% range, the Non-Commercials in the 20-25% area and the Nonreportables in the 20-25% region. Then at the top of the move, you'd see the Commercials relative buying below 20-25%, the Non-Commercials would be pretty heavy in the 75-80% and the Nonreportables would be heavy 75-80%. So at that point, he would look for price to begin trending down. Obviously this is in the perfect world--but that's his general idea.
So I'm looking to get my hands on the COT data within NT so that I can build this relative COT indicator and then see if what he's saying is BS or not (ie, via backtesting). My first idea is to use these relative areas as the starting and ending points of trend bias--and everything in between as noise. So longs only once the Commercials are 75-80% until the Commercials his 20-25%, then move to short bias.
In theory I really like this COT data because it is pretty objective...however the 3 year look back period and trigger levels are obviously places where our onboard biases can muck these up.
I like the "fundamental" reasons that these ideas are based upon, but backtesting will indicate value (or not).
Thoughts? Have you guys found a free way to get the COT data into NT?