An options backtest can show a clean entry at the midpoint, a clean exit at the midpoint, and a tidy profit that no order could have captured. The midpoint is arithmetic, not a promise: when a call is quoted $1.00 bid and $1.20 ask, $1.10 is a mark between two resting prices. It may be useful for valuing a position. It does not tell you that a buyer or seller would trade with you there.
I’ve seen this slip into research because midpoint data looks less noisy than bid and ask, and because it makes strategy comparisons easy. The trouble begins when a valuation convention quietly becomes an execution model. Three mistakes account for a lot of the damage.
Mistake 1: Treat every midpoint as a fill
Suppose a backtest buys a call at a $1.10 midpoint and sells later at $1.18. It reports an eight-cent gain. But if both quotes were ten cents wide, crossing the spread would mean paying the $1.20 ask to enter and receiving the $1.08 bid to exit: a twelve-cent loss before fees.
The wreckage often appears as a strategy that trades frequently, earns small gains per trade, and looks robust across many contracts. Its edge may be no larger than the spread it erased. A midpoint can still serve as a reference mark, especially for estimating portfolio value, but execution needs a price side and a stated fill assumption.
| Use | Reasonable price reference | What it does not establish |
|---|---|---|
| Marking an open position | Midpoint, with a recorded quote-quality rule | That the position could be closed at midpoint |
| Marketable buy | Ask, plus modeled impact where relevant | That the displayed size was enough for your order |
| Marketable sell | Bid, less modeled impact where relevant | That a stale quote was still available |
Mistake 2: Trust a quote with no life in it
An option quote can sit unchanged while the underlying moves, the market pauses, or the displayed market disappears. A backtest that samples the latest bid and ask at each decision may keep reusing that old midpoint as if a trader could act on it. The underlying’s fresh price does not refresh the option quote.
The wreckage is a suspiciously calm option equity curve: few slippage spikes, fills at prices that look precise to the cent, and trades in contracts whose quotes barely changed all day. Track quote age and reject or downgrade observations beyond a threshold suited to the venue and strategy. Also check that bid and ask are positive, ordered, and accompanied by plausible displayed size. A crossed or zero bid market is a data condition to handle, not a bargain to exploit.
On a quiet afternoon, I once watched an option chain look frozen while the underlying kept twitching. It was a useful reminder that a neat table is not the same thing as a live market.
Mistake 3: Treat displayed size as yours
Even a fresh quote with a sensible spread only shows what was displayed when the feed captured it. If the ask shows one contract and the backtest buys ten, assigning all ten the same ask assumes depth that the quote never showed. If the strategy posts a limit at midpoint, assigning it a fill because the underlying later touched that level assumes queue priority and a trade through the order.
The wreckage tends to grow with position size. A one-lot test seems plausible, while portfolio results scale almost linearly to 50 contracts, with no added spread cost or missed fills. Keep fill quantity within observed size when possible; otherwise apply a transparent impact assumption and run sensitivity cases. For passive orders, require evidence of a trade through the limit or model queue and partial fills conservatively. A bar high or low alone cannot tell you whether your order was first in line.
What should an options backtest do with the midpoint?
Separate marking from trading. Record the bid, ask, quote timestamp, and size used for each decision. Mark inventory with a declared valuation rule, then simulate buys and sells against executable sides or a documented limit-order model. Charge commissions and exchange fees separately; a realistic price does not include every cost.
Then compare results under a few explicit assumptions: midpoint marks, bid/ask execution, and a wider-spread or lower-fill case. If the thesis only survives midpoint fills, the research has identified a market-access question, not an established trading edge. That distinction is especially important in options, where wide spreads and thin size can make the most attractive backtest trades the hardest ones to execute.
← All posts
