Kudos to MMs
- Thread starter nitro
- Start date
As I mentioned in previous posts, there is perhaps one more field to add to NFV that may give it exceptional accuracy. But what if that doesn't work? Another point I mentioned in a previous post is that sometimes a market becomes its own predictor. What do we do in that case?
A typical thing to do is to run an exponential MA
EMA(t) = alpha * Price(t) + (1 - alpha) * EMA(t - 1)
The effect of alpha is to focus the EMA on short or long term trends.
Using Reinforcement Learning on this EMA function, you alter it slightly so that price is not the market price, but NFV, and you then use a similar idea to what is in the current SAC article by John Ehlers, Error Correcting (EC):
http://www.traders.com/Documentation/FEEDbk_docs/2010/11/Ehlers.html
EC = alpha * (Price(t) + Gain * (Price(t) - EC(t - 1))) + (1 - alpha) * EC(t - 1)
where the Error Term is the difference between SPX and NFV. Note that if Gain = 0, it resolves to just the equation for a standard EMA, EC = EMA.
This is a cop out since it is curve fitting, but what else are you supposed to use when the market is self feeding? (either that, or my model is incomplete and the market only __appears__ self-feeding, which is always possible)
Eventually EC will be more or less = SPX as the Gain increases and becomes sufficiently large, which defeats the purpose of NFV.
A typical thing to do is to run an exponential MA
EMA(t) = alpha * Price(t) + (1 - alpha) * EMA(t - 1)
The effect of alpha is to focus the EMA on short or long term trends.
Using Reinforcement Learning on this EMA function, you alter it slightly so that price is not the market price, but NFV, and you then use a similar idea to what is in the current SAC article by John Ehlers, Error Correcting (EC):
http://www.traders.com/Documentation/FEEDbk_docs/2010/11/Ehlers.html
EC = alpha * (Price(t) + Gain * (Price(t) - EC(t - 1))) + (1 - alpha) * EC(t - 1)
where the Error Term is the difference between SPX and NFV. Note that if Gain = 0, it resolves to just the equation for a standard EMA, EC = EMA.
This is a cop out since it is curve fitting, but what else are you supposed to use when the market is self feeding? (either that, or my model is incomplete and the market only __appears__ self-feeding, which is always possible)
Eventually EC will be more or less = SPX as the Gain increases and becomes sufficiently large, which defeats the purpose of NFV.
NFV 1091.46. NFV(1) 1135.38. NFV(2) 1170.67. SPX 1193.48.
If you have not seen a short from me NFV(2), it is because NFV(2) has had a huge jump between yesterday and today and even though there is still edge short, the huge move has given me pause. On any indication NFV(2) starts to weaken, of course I am going to whack ES.
If you have not seen a short from me NFV(2), it is because NFV(2) has had a huge jump between yesterday and today and even though there is still edge short, the huge move has given me pause. On any indication NFV(2) starts to weaken, of course I am going to whack ES.