Probability of Touch vs Probability ITM: Why E Differ

Probability of touch and probability ITM dey measure two different things about the same option. Probability ITM dey look at the finish: where the underlying settle on expiration day. Probability of touch dey look at the journey: whether the underlying ever trade at the strike at any time before expiration. For out-of-the-money strike, touch dey roughly two times ITM, and that gap na one of the numbers traders dey under-rate most for short-premium trading.

Probability of touch versus probability ITM, side by side

Make we take stock wey dey trade at $100 and call wey get $105 strike, with one month left. Probability ITM dey ask one question for one particular moment: on expiration day, stock go close above $105? Probability of touch dey ask that same question for every print in between: stock trade at $105 even once? If price rise reach $106 for week two and later settle back at $99, na touch, but e no finish ITM.

Two things follow directly from these definitions. Touch no fit ever smaller than ITM, because every path wey finish above the strike must cross the strike on the way. Touch na the number wey control everything wey fit happen before expiration: alert, roll, stop order, or fear of assignment on short call. ITM na settlement itself, wey be the subject of wetin happen if option expire in the money.

Why probability of touch dey about two times probability ITM?

The short answer na mirror argument, wey probability theory dey call reflection principle. Imagine say underlying na random walk without drift: e dey move up and down with constant volatility, continuous monitoring dey happen, and e no get built-in direction.

Take any path wey touch the strike before expiration and later come back finish below am. From the first touch, turn every later move upside down. Up become down, and down become up. Under symmetric driftless model, that mirrored path get exactly the same chance as the original, and e finish as far above the strike as the original finish below am. Every touch-and-return path get one matching touch-and-stay path. Paths wey finish in the money na the touch-and-stay half, so touch dey come out around two times ITM.

Four assumptions support this result, but none of dem dey hold perfectly:

• Zero drift. Equity indexes don trend upward over long periods, and that dey move upside touch and upside finish rates away from the symmetric case.
• Constant volatility. Real volatility dey cluster, with calm periods next to violent ones. How implied volatility dey calculated explain how dem estimate the number from the start.
• Continuous monitoring. Touch only count if person dey watch. If price pierce the level between sessions, e no be touch on the regular-hours tape.
• One fixed strike. Rolls, adjustments, and early assignment on American-style contracts all dey change the level wey person dey monitor.

The doubling rule na rule of thumb wey get proof behind am: e dey close for near-the-money strikes, but e dey become less exact as strike move further out.

Wetin price history dey show about touch versus finish

You fit test the rule without any model. The panel below check every SPY session since 2011. E draw levels 1%, 2%, 3%, and 5% above that day close, then look forward 21 sessions, roughly one calendar month. E record two things: whether any session high during the month reach the level, which na touch, and whether the close at month-end land at or above am, which na the finish.

WITH
    bars AS (
        SELECT
            date,
            any(toFloat64(close)) AS c,
            max(toFloat64(high))  AS h
        FROM global_markets.stocks_daily_aggs
        WHERE ticker = 'SPY'
          AND date >= '2011-01-01'
          AND date <  '2026-07-01'
        GROUP BY date
    ),
    fwd AS (
        SELECT
            c AS spot,
            max(h) OVER (ORDER BY date ROWS BETWEEN 1 FOLLOWING AND 21 FOLLOWING)      AS fwd_max_high,
            leadInFrame(c, 21) OVER (ORDER BY date ROWS BETWEEN CURRENT ROW AND 21 FOLLOWING) AS close_fwd_21
        FROM bars
    ),
    trials AS (
        SELECT
            arrayJoin([1, 2, 3, 5])     AS pct_away,
            spot * (1 + pct_away / 100) AS level,
            fwd_max_high >= level       AS touched,
            close_fwd_21 >= level       AS finished_above
        FROM fwd
        WHERE close_fwd_21 > 0
    )
SELECT
    concat(toString(pct_away), '% above spot')   AS strike_distance,
    round(100 * avg(touched), 1)                 AS touch_pct,
    round(100 * avg(finished_above), 1)          AS finish_above_pct,
    round(avg(touched) / avg(finished_above), 2) AS touch_to_finish_ratio,
    count()                                      AS windows
FROM trials
GROUP BY pct_away
ORDER BY pct_away

Across 3874 monthly windows, level 1% above spot get touched 83.8% of the time and finish above am only 56.7% of the time, giving ratio of 1.48. At the far end of the curve, 5% above spot, touch reach 19.6%, compared with finish rate of 12.3%, giving ratio of 1.6. The ratio stay near two across the full curve. Na exactly where the mirror argument put am, based on fifteen years of tape and no model.

Delta good proxy for probability ITM?

Delta na the option sensitivity to $1 move in the underlying, and e also dey work as rough estimate of the chance say contract go finish in the money. 25-delta call go gain about $0.25 for every $1 move in the stock, and e get roughly 25% chance of expiring above the strike. The estimate no exact: delta dey calculated under the same driftless, constant-volatility assumptions, and true ITM probability for calls dey a little below delta. Our option delta explainer break down the mechanics.

Use delta as ITM estimate, and the doubling rule go give you touch estimate without extra work. The panel below group six months of SPY contract-days with 20 to 45 days to expiry according to delta size, then put the two figures side by side.

WITH graded AS (
    SELECT
        toUInt16(round(100 * abs(delta) / 5) * 5) AS delta_bucket,
        abs(delta)                                AS abs_delta
    FROM global_markets.options_greeks
    WHERE underlying_symbol = 'SPY'
      AND date >= '2026-01-01'
      AND date <  '2026-07-01'
      AND iv_converged = 1
      AND volume > 0
      AND days_to_expiry BETWEEN 20 AND 45
      AND abs(delta) BETWEEN 0.05 AND 0.45
)
SELECT
    concat(toString(delta_bucket), ' delta') AS short_strike,
    round(100 * avg(abs_delta), 1)           AS approx_itm_pct,
    round(200 * avg(abs_delta), 1)           AS approx_touch_pct,
    count()                                  AS contract_days
FROM graded
GROUP BY delta_bucket
ORDER BY delta_bucket

For 25 delta bucket, average delta across 5211 contract-days na 25%. The doubling rule put the chance of underlying visiting that strike at any time during the contract life at 49.9%. Strike wey only finish in the money about one time in four dey nearly like coin flip to get touched.

Why stop at the short strike dey turn winners to losers

Na here the difference stop to be academic. Trader wey sell 25-delta call and place stop order at the strike don create one position with two different exit tests. Option payoff dey settle according to the finish, wey be roughly one-in-four event. Stop dey settle according to the touch, wey be roughly one-in-two event.

With those numbers, stop go trigger about half the time, and most trades wey e close na the ones wey touch and later come back. Those trades for expire worthless and allow trader keep the full premium. Stop dey turn plenty winners into realized losses, but e no do much to control the tail risk wey dem set am for, because gap through the strike go fill below the stop price anyway. If you size position based on touch rate instead of delta, you dey price the exit wey trader go actually face. Sizing against the expected move from implied volatility na the same idea from the volatility side.

A path wey touch and later come back

July 2024 give one clear fixed example. Since the period don pass, the numbers no go change. The panel start from the first session of July, draw level 2% above that close, and track SPY through the first week of August. The spy_high series na the one wey do the touching; the spy_close series na the one wey settle the finish.

WITH
    bars AS (
        SELECT
            date,
            any(toFloat64(close)) AS c,
            max(toFloat64(high))  AS h
        FROM global_markets.stocks_daily_aggs
        WHERE ticker = 'SPY'
          AND date BETWEEN '2024-07-01' AND '2024-08-08'
        GROUP BY date
    ),
    anchor AS (
        SELECT argMin(c, date) * 1.02 AS lvl
        FROM bars
    )
SELECT
    toString(b.date)                  AS session_date,
    round(b.c, 2)                     AS spy_close,
    round(b.h, 2)                     AS spy_high,
    round(a.lvl, 2)                   AS strike_level,
    round(100 * (b.c / a.lvl - 1), 2) AS close_vs_level_pct,
    round(100 * (max(b.h) OVER (ORDER BY b.date ROWS BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW) / a.lvl - 1), 2) AS run_max_vs_level_pct
FROM bars AS b
CROSS JOIN anchor AS a
ORDER BY b.date

SPY open the window at $545.34, so the level stand at $556.25. The running high finish the period 1.6% above that level, meaning price reach the level during the window. The final close print -4.6% against the same level. Short call struck there for touch partway through the period, but e still for be out of the money at the end. Stop at the strike for close the position somewhere between the two.

Where the doubling rule dey break down

Two cases matter pass. Far out-of-the-money strikes fit push the ratio above two as ITM probability dey approach zero. By that point, both numbers don small enough that the difference rarely change any decision. Strong trend in the underlying fit break the symmetry wey mirror argument depend on, because drift make up-path and down-path unequal from the beginning. Apart from those cases, monitoring frequency matter for short horizons. Price wey gap through a strike overnight no trade at that level, so touch measured from the tape and touch measured from continuous model fit give different results on the same day.

FAQ

Wetin be probability of touch for options trading?

Probability of touch na the chance say underlying go trade at a particular strike at any time before option expire, instead of only on expiration day. Na the relevant number for anything wey fit trigger during the option life, including stop orders, alerts, rolls, and early assignment risk.

Probability of touch always dey two times probability ITM?

No. Doubling na approximation wey work under zero drift, constant volatility, and continuous price monitoring. E dey close for near-the-money strikes, but e dey move further away as strike go farther out or underlying start to trend.

Delta dey tell you probability of touch?

Delta dey approximate probability ITM, not touch. If you double delta, you get workable touch estimate. Na why people often describe 25-delta strike as roughly 50% chance of getting touched at some point before expiration.

Why stop-loss orders at the short strike dey hurt short option positions?

Stop at the strike dey settle when price touch am, but option payoff dey settle according to the finish. Since touch dey happen about two times as often, stop go close many positions wey for expire worthless, turning some profitable trades into realized losses.

Every panel here carry the exact SQL wey produce am, so you fit trace the touch and finish counts row by row. To run the same touch-versus-finish test for another ticker or another horizon, ask the question in plain English on the Strasmore terminal.