Rollover and your statistics

Four times a year the contract you trade stops being the contract you trade. Your journal keeps calling it NQ, which is what you want for almost every statistic and wrong for exactly one.

What rollover actually is

A futures contract has an expiry. The one everybody trades is the front month — the nearest expiry with real volume — and roughly a week before it expires, liquidity moves to the next one. For the index futures that is quarterly: March, June, September, December, written H, M, U and Z.

So a trader who says "I trade NQ" has, over two years, traded NQH4, NQM4, NQU4, NQZ4, NQH5 and so on. Eight distinct instruments, each with its own price history, each of which stopped existing.

What our journal does with that

Our importer strips the month and year from the ticker and keeps the root. NQH5 and NQM5 both become NQ, and everything grouped by symbol treats them as one instrument.

This is the right default, and it is worth being explicit about why:

  • Point value is per root, so the contracts convert to currency identically. Nothing about the money changes.
  • Your edge is per root, not per contract. You do not have a March strategy and a June strategy. Splitting them would give you eight samples of thirty trades instead of one of two hundred and forty, and everything that needs sample size — expectancy, profit factor, SQN — would become unreadable.
  • You would not recognise the alternative. A symbol list showing eleven NQ variants is not a useful view of an account.

The statistic it invalidates

Merging contracts is safe for anything computed from differences and unsafe for anything computed from levels.

At rollover the front-month price changes discontinuously. The new contract carries a different amount of time value and financing, so it typically prices some distance away from the old one — sometimes tens of points on an index. Nobody could have traded that gap; it is not a move, it is a change of instrument.

Anything that reads prices as a continuous series therefore inherits a jump that never happened:

  • Average entry price across a period spanning a rollover is an average of two different scales.
  • Price-level analysis — "I do badly above 20,000" — mixes levels that were not comparable.
  • Any chart of your fills over time on futures shows steps at each rollover that look like gaps in your trading and are not.

What stays correct is everything built on per-trade results: P&L, win rate, payoff ratio, R-multiple, holding time, session breakdown. Each of those uses entry and exit within one trade, and no trade spans a rollover, so the discontinuity never enters the arithmetic.

That split — differences safe, levels unsafe — is the whole rule, and it is worth carrying because it applies to any tool you use on futures, not only to ours.

Two practical consequences

Your holding time may look shorter than it is. Traders who hold through an expiry do not hold through it — they close and reopen in the next contract. The journal correctly sees two trades, because there genuinely were two. Average holding time on a swing futures account therefore reports the leg, not the position, and the difference grows the longer you hold.

A rolled position is not a losing trade and a winning one. Closing the March contract at a loss and reopening June is one economic decision recorded as two trades, and one of them can be red while the position was never in trouble. If your futures win rate looks worse than your account balance suggests, count how many of the losses land within a day of a quarterly expiry.

Why we do not publish rollover dates

The obvious companion to this page would be a table of the exact roll dates per contract. We do not publish one.

Roll timing is not a fixed rule — it is where volume actually moves, which varies by contract and by quarter, and the exchange's official last trading day is not the same thing as the day liquidity left. A table of dates would be right most of the time and silently wrong occasionally, which is the failure mode this whole section is built to avoid. Your own fills tell you when you rolled, and that is the date that matters for your statistics.

More in Market hours and time

  • Broker server timeWhy MetaTrader timestamps carry no timezone, what brokers actually use, and how to find your own offset before trusting any analysis by hour.
  • Sessions and overlapsWhen the major sessions run, why the overlap matters more than the session, and how daylight saving moves everything twice a year.
  • Futures point valuesThe point value of the common CME, CBOT, NYMEX and COMEX contracts, how a ticker like NQH5 decodes, and what a journal does when it meets a contract it does not know.
  • What "1" means in the size columnThe size column in a broker export means lots, coins, shares or contracts depending on the platform, and nothing in the file says which.
  • News windows and your instrumentHow to work out which currencies your instrument is exposed to, why the window matters more than the forecast, and where a firm rule replaces your judgement.