Technical issues of using COT data Pipruit: Well, here is another problem. I’ve slightly experimented with COT data, and found out that we can’t estimate absolute extreme for positions. Am I right? Commander in Pips: Absolutely. This data, as you’ve understand correctly, is not normalized. In other words it means that those extremes that was 3-5 years ago will be just flat in nowadays. We have to adjust them in time, to get correct information. This happens because the market became bigger, since money supply is growing geometrically, the market value also grows. Pipruit: So, how we can apply it then? Commander in Pips: Well, there are two different ways that could be useful. 1. Create a non-normalized indicator; 2. Create an index. Anyway it will demand some sort of manual work. Let’s start from first point. Here you will have to re-read how we deal with Detrended Oscillator – just remember how to estimate critical areas for overbought and oversold here. Pipruit: Oh, I think I’ve got it! We need some numbers of COT data first. With analogy to DOSC, I suppose for 120-130 previous weeks. Then we have to choose average most high and most low among them, and use them as extreme points! Commander in Pips: That’s right, but you need to experiment with the number of periods – probably you will need more periods, since highs and bottoms on sentiment appears more rarely than on market. Probably you will need data for 3-5 previous years. But I want to make your work a bit easier, so that you will not have to deal with both commercial and large traders’ data, you may use their difference: COT Indicator = Net Large positions – Net Commercial In fact, the shorter time frame you use for taking highs and lows of COT data – the more signals you will get, but the more often these signals will fail, since they are not solid extremes. If you will take larger time for data calculation, as we’ve said 3-5 years, then you will get fewer signals but they will be more reliable.