Symbol Overlap

Symbol overlap (concordance S, 0–1): how much two reports hold a position in the SAME instrument at the SAME time — any direction — from their trade intervals. 1 = always in the same markets together; 0 = never co-exposed. Catches crowding that return correlation can miss.

Computed from: Trades list
Scope: Across reports
Range: 0 – 1
Direction: Lower is better

Symbol overlap answers a question that return correlation can't: do these two strategies actually hold positions in the same instrument at the same time? It's measured from the trade list — when each report had any position open on which symbol — not from the equity curve. Two strategies can show near-zero return correlation yet both be in EURUSD at the same moment; that hidden crowding is exactly what blows up "diversified" portfolios when one market gaps.

How it's calculated

S  =  Σ (time both hold a position in the same symbol)
    ─────────────────────────────────────────────────
    Σ over each shared symbol: max(Tᴀ, Tʙ)

symbol: the unit of analysis — all of a report’s positions in one instrument, long and short merged into a single "in the market" presence
Tᴀ, Tʙ: each report's time-in-market on that symbol, within the pair's common active window
S: similarity ∈ [0,1]; NaN if the pair never traded a common symbol at all

In this product: NOT a correlation — there is no −1 pole; it's a time-overlap coefficient (Jaccard-style, max-normalized) on each trade's [open, close] interval. It works on the raw intervals rather than binning trades onto a grid, because grid-binning would fake a number that depends on the bin width (the Epps effect). Max-normalized (divided by the busier trader's time): a small position nested in a large one is a partial match — A in EURUSD 100 days, B for 5 → 0.05, not 1. Lot size is ignored — presence only. Each report's two sides on a symbol are merged into one presence first, so a long and a short held at once count as in the market once, not twice. The common active window is the stretch when both reports were trading at all — a report's history before the other existed is excluded. A symbol (or, in the variant, a symbol×side) that only one report trades still enters the denominator: it can only dilute S, never raise it — which is why a non-shared position drags the number down rather than making it undefined.

The shaded regions are when both reports hold a position in the same symbol at once. S is the share of time spent co-exposed, normalized by the busier trader. High S = crowded; the equity curves might still look only loosely correlated.

What it tells you

Value	Reading	Notes
0.7 – 1.0	Crowded	The two are in the same markets almost whenever either is — little diversification, concentrated single-market risk.
0.3 – 0.7	Partial	Meaningful shared exposure on some instruments.
0.0 – 0.3	Distinct	They rarely hold the same instrument at once — genuine diversification of exposure.
NaN	No shared market	They never traded a common symbol — nothing to overlap at the instrument level (still check for shared currency legs).

The bands are rough defaults — judge S relative to your strategy mix. A book of majors-only trend followers runs structurally high; a basket of unrelated instruments runs structurally low, so a 0.3 that is alarming for one is normal for the other.

Pitfalls

Pitfalls & caveats

It ignores direction — by design. Two reports both in EURUSD overlap here even if one is long and the other short. That's the point of this view (shared market exposure), but it can overstate crowding when the positions actually offset. To separate same-direction crowding from hedges, switch to Directional overlap.
It's normalized by the busier trader. S divides by the larger book's time-in-market, so it reads crowding from that book's point of view. A small strategy whose entire EURUSD risk sits inside a much larger EURUSD book shows a low S (e.g. 0.05) even though 100% of the small book's risk is shared. Before sizing the smaller strategy as a satellite, check the asymmetry — the number understates its crowding.
Size doesn't count either. S ignores lot size (presence only). Two reports both in EURUSD overlap even if one risks 10× the other — it flags the shared market, not the magnitude of the shared risk. The number that actually matters is overlap × relative size; read S next to trade leverage.
Different symbols ≠ different risk. Reports trading entirely different symbols return NaN — nothing to overlap at the instrument level. But on a forex book that is not the same as uncorrelated: long EURUSD and short USDCHF are both a short-USD bet through a shared currency leg, yet share no symbol. Treat NaN as "no instrument-level crowding," then still check for shared base/quote currency or factor exposure before calling it true diversification.

Symbol overlap vs return correlation

They are two halves of "are these strategies really different," and you need both:

	Return correlation	Symbol overlap
Reads	the equity curve (daily returns)	the trade list (open positions)
Range	−1 to +1 (signed)	0 to 1 (unsigned)
Catches	co-movement of results	co-holding of the same instrument
Blind spot	same-market crowding that nets out in returns	the direction of the shared exposure

The dangerous case is low return correlation + high symbol overlap: the equity curves look independent in calm markets, but both books are in the same pair — so the portfolio carries concentrated single-market gap risk (a gap = price jumping to a new level with no trading in between, where a shared position hurts most). Whether that risk doubles up (both on the same side) or partly offsets (one long, one short) is exactly what Directional overlap tells you next. Symbol overlap is the metric that sees the shared exposure coming; directional overlap says which way it cuts.

Related metrics

Compare withReturn Correlation