Dude! That is so wrong. I would have replied in detail, but too many Mojito's already. Do some googling, will ya?
Construct continuous series any way you choose. You might pick a splice date, say 10 sessions before expiration. Or 5 sessions. Choose whatever you want. Test it out. There may not be a single correct way to do some things.
Exploit the fact that you (mizhael) are a hedge fund manager with a multi million dollar account at an FCM, Newedge, which only accepts institutional business. There are no penny ante retail scum newbies @ Newedge. Diversify. Create continuous futures contracts four different ways. Backtest your suite of trading strategies, having them trade all four of these continuous contracts. (If you have 7 strategies, backtest with 28 strategy+continuous_contract combos, traded simultaneously.) Divide your trading capital into four portions, and trade each portion of your capital on a different rollover methodology. Some possible candidates that you might allocate 1/4 of your capital to, could include Rollover N days before expiration Rollover on the J-th occurrence of (volume in new month exceeds volume in old month) Rollover on the K-th occurrence of (open interest crossover) Rollover on calendar date D, as Pinnacle Data does in their commercial database Enter a limit order to trade the rollover spread, T ticks better than the delta between the contracts at the close of the previous bar. (Daily bar? 2-hour bar? etc? Your choice). If unfilled, keep trying to get a T-tick better fill, for as many as Z bars. If unfilled after Z bars, eat shit and rollover at market or as best you can. Don't rollover at all. Instead, exit the position in the old month contract, X days before expiry (X=0?? 1?? your choice) and treat the new month contract as a new tradeable instrument. DO NOT automatically enter a position in the new month contract, unless and until the trading strategy signals a brand new & fresh trade-entry signal on the new month contract's data Start thinking like a professional risk manager.
You are thinking that futures contracts are like stocks. When you buy a futures contract at $100 you don't spend $100, and likewise when you sell a contract for $100, you don't get $100. I would take the advice of a previous poster and spend a large chunk of time on google learning the basics about futures. Too much here to go into in a forum.
Yep, I fully agree with you that futures need no upfront cash payment to trade. When you enter into a long/short contract, you don't have cashflow at that time. However, for rolling it is slightly different, and you asked about "roll PNL", right? How would you roll? You either buy or sell a spread. When you buy or sell a spread, do you pay or get nil? Maybe your definition of roll PNL is different? ----------------------------------------- Or if you meant the [overall] PNL on the roll day? then that's (the position as of roll_date -1 )* [(price as of roll_date) - (price as of roll_date - 1)] in this case, the "roll PNL" doesn't come into picture, due to the specialty of futures trading... So please note that "roll PNL" and "[overall] PNL on the roll day" are two different things...
Thanks a lot MGJ. And thanks for your warm encouragement. I am an assistant/clerk at this moment... yeah becoming a fund manager is my dream and I am working very hard towards that... And thanks very much for your very enlightening and information guidelines. I will back test these nice ideas. However, there is one more thing: you talked about "when" to jump, and provided lots of ideas. When I try to extrapolate your thoughts onto "where" to jump, of course "diversification" is the key. However, there is a problem - yeah, you can "diversify" as much as you want in backtest. And it's a combinatorial exercise. For example, you might find that jumping on the very last tradable day and jumping to a very far-out contract/curve a very profitable strategy in backtest. (Indeed, I often found that the most profitable results come from very illiquid and thinly traded assets). One might think that's a illiquidity premium - the reward you receive when you bear the illiquidity risk. But I think that's more of spurious backtest and illusion. So my question to consult with your expertise and experience is: how would you filter these "illusions" out in your combinatorial and exhaustive backtest? For "when" to jump, it might be easier: you might have hard-rules such as "don't trade the very last day of trading", or some soft rules such as "don't trade the very last day if you have large positions", etc. But for "where" to jump, it 's harder; how would you do that? In addition to the potential liquidity problem in this combinatorial exercise, what other potential traps might arise, based on your past experiences? Thank you and happy holidays everyone!
The roll means you are merely "exchanging" one contract for another so that you maintain exposure even though the contract you are currently in is going off the board...no pnl (other than transaction costs) involved. In effect you close out the pnl (since inception) on the old contract and open up a brand new pnl stream using the new contract. If your models are showing any sort of persistent pnl effects due to roll you are doing something wrong. You are a clerk right now, right? When do your bosses/traders roll their positions relative to expirations? Why don't you use that as a guide? In fact, I would think your current place of employment would be the best source of information about all this sort of stuff. Ask more questions at work!