Why a 0.20-delta credit spread pays only about 14% of its width
A trading brief we tested required two things at once: sell an option at about 0.20 delta, and collect at least 25% of the spread width as credit. On weekly Nifty options those two conditions almost never hold together. Here is the arithmetic, and what we measured.
The brief and the problem
The brief described a bull put (or bear call) credit spread: sell the put near 0.20 delta, buy a lower strike to cap the loss, and "ensure net credit is at least 25% of the spread width". We pre-registered the rules, as we do for every study (see how we test trading ideas), and before looking at any profit and loss we counted how many of the entry signals could actually satisfy the rule.
Over the first 700 sessions of our sample (December 2020 onward) the entry rules produced 292 signals. The 25% credit rule rejected 291 of them. Run exactly as written, the strategy produced one trade.
Why delta decides the credit
An option's delta is a rough guide to the market's probability that it finishes in the money. A credit spread has a short strike and a long strike further out. Its credit is the difference of the two premiums, and as a fraction of the strike gap it is approximately the average delta of the options between the two strikes.
- Short leg 0.20 delta, long leg about 0.14 delta: average about 0.17, so the credit is roughly 14 to 17% of the width.
- To collect 25% of the width, the short strike needs to be around 0.30 delta or more, which is a different trade: a higher probability of being tested, in exchange for more credit.
The two numbers in the brief ask for something the option chain does not offer. This is not a quirk of one broker or one year; it follows from how option prices work, and our measurements agree with the arithmetic.
What we measured
For each of the 291 signals that reached the pricing step we priced a 0.20-delta short leg and a long leg at several widths, using the minute-level weekly Nifty option data (observed prices where the strike was inside the data window, a standard Black-76 model with a fitted volatility skew elsewhere). The credit divided by the strike gap:
| Spread width (points) | Mean credit / width | Median | 25th to 75th percentile | Maximum |
|---|---|---|---|---|
| 100 | 13% | 14% | 11% to 16% | 23% |
| 150 | 11% | 12% | 9% to 14% | 20% |
| 200 | 9% | 10% | 7% to 12% | 19% |
| 250 | 8% | 8% | 6% to 11% | 17% |
Two things stand out. The credit fraction falls as the spread widens, because the far wing adds little premium but adds a lot of risk. And no signal in the sample reached 25% at any width; the maximum was 23%. The 0.20-delta short strike sat a median of 167 points from the index (25th to 75th percentile: 123 to 229 points).
What a 14% credit means for the win rate
A credit spread has a capped gain (the credit, c) and a capped loss (the width minus the credit). If the trade wins with probability p, expected profit before costs is zero when
p = 1 − c / width
| Credit as % of width | Break-even win rate (before costs) | Loss on a full stop-out, in multiples of the credit |
|---|---|---|
| 14% | 86% | 6.1 times |
| 20% | 80% | 4.0 times |
| 25% | 75% | 3.0 times |
| 33% | 67% | 2.0 times |
At a 14% credit the strategy must win about six trades in seven just to stand still before brokerage, taxes and slippage. That is the real meaning of "selling 0.20 delta": a high win rate is not an edge, it is the price of admission.
What happened when we relaxed the rule
We recorded a dated amendment lowering the credit floor from 25% to 10% of the width (before any profit and loss was seen), and ran the full test: 137 trades in the development period and 28 in validation, on a capital base of ₹10 lakh, risking 0.5% per trade.
- Average gross result: about +₹14 per trade in development, −₹182 in validation.
- Charges: about ₹98 per trade (four orders at ₹20 flat brokerage plus statutory charges).
- Net: −₹17 a day in development (t-statistic −1.4) and −₹25 in validation. At twice the costs, −₹36 a day.
- Roughly half the trades ended on the structural stop; about one in eight reached the profit target.
The full audit, including a random-trigger comparison and a version without the gamma filter, is on the Strategy Review page. The short version: the gross edge was a few rupees a trade, and costs were several times larger.
What to take from this
- Check the credit-to-width ratio before the entry rule. If a system needs a credit well above the average delta of the strikes it sells, it will rarely trade, or the numbers in its description are inconsistent.
- Compute the break-even win rate. 1 − credit/width is a one-line check that most spread tutorials leave out.
- Count the charges per trade, not per month. A ₹98 charge against a ₹14 gross edge cannot be fixed by sizing.
- Model-priced legs are a weakness. Every spread here used at least one leg outside the observed strike window; real quotes could be better or worse.
Limits of this study
The sample is 165 trades over about five years and is dominated by bear call spreads (97% credit spreads). Prices for out-of-window strikes are modelled. Slippage of 0.5% per fill is an assumption. A 0.20-delta rule on a different underlying or on monthly expiries would give different ratios, though the average-delta logic would still apply. This is research into how a rule behaves, not a recommendation to trade any structure.
Sources and method
- Minute-level weekly Nifty option data from December 2020, as described on our Strategy Review.
- Charges per the exchange and broker schedules; see what the April 2026 STT rise changed.
- Dates and counts in this article come from the study's pre-registration and results files; the sealed final 20% of sessions was not opened.