Kalman payoff interpretation

[aliakhavan89]

I have some basic questions about Kalman setup. I've generated a synthetic data, using the same model, and I'm trying to use this payoff:

Code: Select all

*CK
model|data/variance

Which contributes to:

Code: Select all

- (model-data)^2/variance/2 - LN(variance)/2

I think I should use EVM1 as the variance even though I have only 1 series (right?), but I temporarily defined my payoff function as:

Code: Select all

*CK
PN1|PN1 Data/1

Because I thought I might have a more controlled experiment. This is also the .prm setup:

Code: Select all

PN1Stock/sp1_var/init_sp1_var

Now, if you change 'esp1 std' with filtering enabled, you would see PN1 could fit PN1 Data. But I'm confused about the payoff values I'm getting. Especially, if I manually set esp1 std = 0.01, I get a relatively high payoff without a good fit apparently. At the same time, with esp1 std = 10, I get a very good fit but with a very bad payoff value. Could you please help me understand what's happening?

Screenshot 2025-06-25 at 6.47.17 PM.png (884.01 KiB) Viewed 52 times

Screenshot 2025-06-25 at 6.48.51 PM.png (884.72 KiB) Viewed 52 times

Note: I'm not doing any calibration yet. Also, here's the model:

Kalman_v04_Share.zip

(1.96 MiB) Downloaded 10 times