Everything in markets, with only a few exceptions, are rising in value over time and therefore trended to time. The simplest method I found is to use link relative or first difference detrending, before calculating a correlation between assets.
I also updated it to include more assets, to use the latest Cryptocap indices like BTC .D/ETH.D/ TOTAL2 and the top 100 crypto index CIX100 . I improved the colour schemes, too
Included in the source code are some other ideas like the FRED:M1 and FRED:M2 (only on ), the Yuan/Yen/ EUR/USD , etc. There's lots of scope for correlating unrelated markets.
I'll keep updating it as I use it to find truly correlated assets. Some kind of signal line of known correlations, to subtract from the baseline fuzz of market activity.
Any improvements are most welcome; I'm a novice at best at statistics and build on others' work.
- changed the short title to be very short, for those of us using it with 2/4/8 chart layouts
- "Same as symbol" timeframe support
- min-max normalisation
- mean normalisation
- z-score normalisation (aka standardisation: mean over stdev)
- the differentials will choose first-difference then link relative if both are enabled
- the normalisations are mutually exclusive and the first one enabled will be the one used
- you can use 1 differential (normal, first difference, or link relative) and 1 normalisation (min-max, mean, or standardise) together
- different plot styles and transparency defaults for maximum visibility
Example use case was that I saw a drop in correlation between USDCNH and CNYUSD which provided a great signal to trade.
I recommend you exercise caution and try to understand the reason for using the different algorithms. Each differential and each normalisation has its own purpose and this is only a crude tool.
As always, this isn't financial advice, do your own research, etc etc.