Yesterday I noticed a
conceptual issue thanks to the NG rollover ("information transfer"). It actually costed us about 25K in missed profits, but anyway it was very useful to settle once for all this conceptual matter. Let see what is this about.
Assume we have layer, say NG or whatever instrument, and that the layer looks like this:
Now it becomes critical what we use to recreate the position. Lately I took the habit to inject
players to recreate the position. Now, I can clearly see this is
wrong from a conceptual standpoint. The problem is that if we inject buy a player to recreate the position, which is actually representative of orders made at a much higher price, it can be closed almost immediately because it can easily go in profit at the first retracement, so we are left only with the sell orders used to close the position, and this is actually equivalent to
destroy the entire
game on that layer by placing a stop at a very low price.
Yesterday, I wasn't following the game, but when I returned in the evening I just realized how wrong was that, so I re injected out of game orders to correct the situation (we had already lost over 25K of missed move at that point, since NG was shooting up).
Anyway, never matter, it is important that we have finally found the
proper way to do the rollover:
1- Close order (both
player or
out-of-game orders are fine, or
virtual orders if the position was liquidated by the broker)
2- Re-open order: an
out-of-game order
must be used
So keep that in mind when rolling over.
When closing, instead, it seems to me that is indifferent to either use a player or an out-of-game order. I would leave that to the fund manager preference. At the moment, I think it's preferable to use a player (as it would automatically recover the
contango at the first chance, with an overload of the position).
(Clearly, for instruments in backwardation, similar considerations apply.)
(as I also include spread and commissions to close.)