Statisticians come in two varieties from which endless specialities/flavours are derived. At the practice's core are several sub-variants.
First we have the Frequentists, who carefully record events that have happened and count their occurrence again and again until they can determine what is likely to re-occur in the future based on the historical data they have collected.
Over the months I have documented and suggested endless versions of frequentist trade structures which we can use and accrue income from.
Secondly we have Bayesian statistical analysis which works very similar to the Frequentists' studies except it considers past events statistically and implies expected results despite not knowing the full facts.
We all do this when we complete a crossword puzzle... we use our knowledge to complete a series of questions to form a final pattern that, if successful, will finish the crossword even though we didn't know what the final crossword would "be". Bayesian mathematics works on the foundation principle that the probability of event B given that event A has occurred is as close to 100% as possible to be of worth.
Again, trades of this type have been posted.
These two types of statistical models have caused many a heated debate over which has the greater value in forums about algorithims, computer coding, or general matters such as credit scoring for mortgages and credit cards. or occasionally (now) infamous law cases (and subsequent proven miscarriages of justice).
Which brings me to the matter of subjectivity and control measures. If we are to use these modern-day techniques of assessment, then we need a means of measure that gives us confidence in its results. We need to determine what is truly ''a random event'' and what is a recurring event which we can confidently expect to occur and therefore trade from.
But should we ignore the random event? Should we say, as we have no meaningful long-term correlation of data, that the so called ''random event'' is a rogue movement in markets that we have to accept as ''the law of s*d
Frequentist and Bayesian analysts rely on one other core foundation rule and that concerns the size of the data set. Ten years of data is great as a large pool of results gives us consistency as a measure, but the flaw is that times change and markets move faster and respond to different world measures.
Ten years ago, telecoms were developing the mobile technology that permits mobile banking or simple online chat and these two areas have indeed changed the world's economic structures as well as the political landscapes of various countries (Egypt, for example). So we need to also look at smaller blocks of real-time history – say five or even three years.
The flaw in this, however, is that you have less data to support your view and we may be giving too much weighting to a market or equity or FX pair than is necessary, and thus we skew our results due to a lack of data.
The compromise is less conclusive data for more real-time trend analysis/overview but greater risk.
So for "fun", here are some very short-term quirks of markets that correlate well but have the really big caveat that they can only be viewed as coincidences because we just do not have the supporting (three-plus years) of data to support the theory;
What we do have, however, is an 84% or greater probability that, given the following has occurred, we can expect these trade setups to continue to have co-incidental confidence!
If the data drop below 84% probability I will let you know and then you can stand aside from the trade.
First, lets look at the USDCAD spot market...
Using the UK time reference here (you can correlate to your own time zones easily enough), look at the hourly close price at 1100 and compare to the price at 1200, one hour later. If the latter candle close is above the 1100 one, then immediately enter a long for plus 5 or if it is below, then immediately enter a short for the same amount; use the trading range of the 1100 candle as your stop.
Create your own charts with SaxoTraderGO click here to learn more
Source: Saxo Bank
Now let's look at silver; we can look at XAGUSD spot or the near-dated silver futures markets.
Silver has a cycle for buy and sell ''roll over'' periods that coincide with the lunar calendar. It has nothing to do with the moon but there is a coincidence in timing which we can trade:
- new moon = sell
- full Moon = buy
These are obviously only scalps only. Let me also say that I know absolutely nothing about astronomy, but this cycle also correlates well, coincidentally, with soy crush, wheat and corn.
Thirdly FGBL... near-dated bund futures. Due to space limitations I will post this as an update immediately after the article.
To be honest. I have a list of these ''oddities'' for making trades and will post them with the caveat that the US senator Donald Rumsfeld once described...
"We know what we do know.
We know what we don't know.
But the real problem is; that we don't know what we don't know."
It is that last line which has yet to be quantified so we run a greater risk as this must be viewed as a coincidental market move until confidently proven otherwise.