Kudos to MMs
- Thread starter nitro
- Start date
I will be doing a study to see if large divergences (LDs) from NFV happen at regular intervals. For example, say that LDs are considerably more likely to occur near quadruple witching (QW). That would mean that my model is missing a real causal effect.
If we did find such an effect, we could do two things, we could add a term to the model to account for "distance" to QW in the same way that Einstein adds a cosmological term to his General Relativity equations to account for a static universe, or I think a better way would be to leave NFV alone, but have the coordinate system that is the add points, change to be more conservative as we approach a QW. So maybe the chain would be D*chain, where D was a measure to proximity to QW, e.g. 14->24->35->etc maps to 14 * D->24 * D->35 * D-> etc. At more than two weeks distance, D would be 1, so that out normal chain would be intact, but as QW gets closer, D would start to be > 1, so we would need more edge to initiate or add a position.
This would affect profitability and Sharpe/Sortino ratios both, so it would be a worthwhile study. My fear is that the casual effect is not as simple as D to QW, but maybe a non-linear term like (D to QW / Amount of bond auctions), or something complicated like this. Feed a bunch of variables like this to a Neural Net, and it would find the most likely relations though.
If we did find such an effect, we could do two things, we could add a term to the model to account for "distance" to QW in the same way that Einstein adds a cosmological term to his General Relativity equations to account for a static universe, or I think a better way would be to leave NFV alone, but have the coordinate system that is the add points, change to be more conservative as we approach a QW. So maybe the chain would be D*chain, where D was a measure to proximity to QW, e.g. 14->24->35->etc maps to 14 * D->24 * D->35 * D-> etc. At more than two weeks distance, D would be 1, so that out normal chain would be intact, but as QW gets closer, D would start to be > 1, so we would need more edge to initiate or add a position.
This would affect profitability and Sharpe/Sortino ratios both, so it would be a worthwhile study. My fear is that the casual effect is not as simple as D to QW, but maybe a non-linear term like (D to QW / Amount of bond auctions), or something complicated like this. Feed a bunch of variables like this to a Neural Net, and it would find the most likely relations though.
NFV 1067.62. SPX 1105.03.
There won't be anything to do, as we already have our third unit short on. We are short (approximately) from 1089, 1104 and 1107. So we are losing 16 handles on 1 unit on the "14", losing 1 handle on 1 unit on the "24", and winning 2 handles on 1 unit on the "35".
There won't be anything to do, as we already have our third unit short on. We are short (approximately) from 1089, 1104 and 1107. So we are losing 16 handles on 1 unit on the "14", losing 1 handle on 1 unit on the "24", and winning 2 handles on 1 unit on the "35".
