Delayed optimal position

[oldlore]

I've been getting this "Delayed optimal position" for SGX on my daily Status reports for a week now and I haven't been able to figure out why. Contracts, daily prices, all seem to be up to date. There are some days without prices but that seems to be the case for many thinly traded instruments.

Thanks in advance.

============================
Delayed adjusted / FX prices
============================

             last_update
BB3M 2024-10-14 23:00:00


=============================================
          Delayed optimal positions         
=============================================

                                  last_update
dynamic_system/SGX 2024-10-28 20:40:48.486432

[tgibson11]

I have the same thing in my status report, it's not just you.

So far I've traced the problem to here

    @property
    def _keys_with_valid_data(self) -> list:
        valid_correlation_keys = self.covariance_matrix.assets_with_data()
        valid_optimal_weight_keys = self.weights_optimal.assets_with_data()
        valid_per_contract_keys = self.per_contract_value.assets_with_data()

None of those assets_with_data lists include SGX.

Beyond that,

I do know that the SGX contract is very thinly traded. It looks like there were no trades at all for the week starting October 28.

It's possible that after some time passes (maybe after the next roll), there will be enough data to generate optimal prices. But that is only a guess.

I'll dig into it some more if I have time, but I'm not sure I care enough to spend too much time figuring it out. Eventually, I'll probably just drop it from my system if it doesn't start generating optimal positions again.

[oldlore]

Glad to know it's not just me. Hopefully it'll correct itself as you said. Thanks for the quick response!

[tgibson11]

OK, I figured this out.

To generate optimal positions, dynamic optimization needs to know how the returns of each instrument are correlated.
It uses daily returns to do that.
To calculate a daily return, it needs the difference between the price on 2 consecutive days.
SGX has not had 2 consecutive days with a trade (and therefore a price) in some time.
The time series of daily returns for SGX is entirely NaN recently.

Seems obvious now, but it took a lot of digging in the code for me to understand what was going on.