[RT] Nasdaq-100 $1.1 Million Equity Model Validation

["BL" <blee7@xxxxxxxxxxxxxx>]

Here are just a few items to examine when developing neural net trading
models, such as the attached $1.1 million hypothetical equity Nasdaq-100
Futures model (222.gif).  Are there a sufficient number of trades to
minimize the chance of the neural model being the result of a pure chance
happening (Yes, total model trades = 264)?  Has the model continued its
profitability with true, holdback data (Yes, +$74,700 since 4-1-2001 data
holdback date)? Is the model signal optimized (No, greater confidence in the
results if it is not optimized)? Is the equity curve increasing in all 4
periods, train/test/validate/holdback (Yes)?  Is the test/validate/holdback
data sufficient as compared to the training data (This is subjective but you
do want some substantial test/validate/holdback data as compared to the
training data.  This model uses 50% / 25% / 25% train/test/validate data
split, respectively).  Is the total equity dependent on one or two big
winning trades (No)?  Is it relatively easy to produce additional, highly
profitable models using the same inputs/target variable(Yes)?  If the
model's neural signal is optimized, how different would the maximum
profitability be compared to the non-optimized model (optimization does not
improve this model - it is at its maximum profitability level with no post
optimization)?  Is this optimized profit level dependent upon fragile
criteria, ie, does the optimized neural signal possess a small sweet-spot
(This model has a huge profitable sweet spot when using all data exclusive
of training data- see 111.gif)? Are any of the model inputs optimized to
better predict the target variable (No)? Will passage of these tests result
in a tradable model (They don't ensure it, but they definitely lend greater
confidence in the model's predictability)? - Brian

To unsubscribe from this group, send an email to:
realtraders-unsubscribe@xxxxxxxxxxxxxxx

 

Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/

Attachment: Description: "111.gif"

Attachment: Description: "222.gif"