SellOptionScreen·BacktestDashboard★ My Watchlist

Every candidate the screener has emitted that has reached expiry — 7,699 trades from 2026-08-21 to 2026-10-23, wins and losses, nothing excluded. Outcomes are graded at expiry against the closing price.

By strategy

Winners, losers and drawdown

Profit factor is gross profit divided by gross loss — $7,184,174 taken in against $1,914,869 given back, so every dollar lost came with $3.75 earned. Anything above 1.00 made money; the ratio says how much room there was between the two, which a net total alone cannot.

Average win and average loss are the two numbers a win rate hides: $1,114 against $1,534 here, so a loss runs 1.4× a win. For a short option the credit is the most it can ever make while a move through the strike is not bounded that way, so the losing side is the longer one by construction — covered calls are the exception, because the shares appreciate up to the strike. At this ratio the book breaks even at a 58% win rate, against the 83.8% above. That gap, not the size of the wins, is where the result comes from, which is why a win rate slipping a few points matters more here than it would for a strategy with symmetric payoffs.

Max drawdown is the largest fall in cumulative P&L from a previous high: $9,957 off the high set at 2026-10-16, reached at 2026-10-23, 0.19% of that high. The curve it is measured on is the Cumulative P&L column below, so the figure is checkable rather than asserted.

Drawdown is measured per expiration date because that is the only ordering the data has — thousands of positions resolve on the same expiry with no sequence between them, so a trade-by-trade equity curve would be invented. A fall inside a single expiry is therefore invisible here.

Equity per $10,000 of principal

Selling one contract of every candidate tied up $142,905,182 of collateral, which is not a number anyone can picture. This is the same cumulative P&L line divided down to each $10,000 slice of it: $10,000 in, $10,369 out across ten expiration dates, which is the +3.69% at the top of this page in units you can hold.

It is a rescaling, not a $10,000 account. A real $10,000 could not hold 7,699 positions, or even the 1,749 that resolved on 2026-08-21 alone — one cash-secured put on a $100 stock is $10,000 of collateral by itself. The curve answers “what did each $10,000 of the collateral earn”, which is a question the data can answer, rather than “what would my $10,000 have done”, which needs position sizing this backtest deliberately does not model.

Not compounded, and that is the honest choice rather than the flattering one. Each expiry’s own return is not multiplied into the next, because five of the ten expiries resolved fewer than 30 trades — 2026-09-04 settled 26 — and treating a night like that as a full redeployment of the account would turn a thin sample into a growth rate. Positions in different expiries are also open at the same time, so there is no sequence of separate periods to compound in the first place.

The shape is worth reading: the line is a staircase, not a slope. 2026-09-18 and 2026-08-21 carry 87% of everything earned, because the monthly expiries are where the screener emits thousands of candidates and the weekly ones a handful. Each flat stretch is a week that resolved almost nothing, not a week that was sat out.

By expiry

Each row is one expiration date. Several scan dates feed one expiry, and a trade's outcome is fixed on the day it expires. Cumulative P&L is the running total down this column — the equity line the drawdown above is taken from.

Profitable and Not assigned are different questions and the gap between them is real: a put assigned just inside the strike can still make money. One GS contract closed 1% below a $1,050 strike against a $2,140 credit — assigned, and still +$1,068. “Profitable” is P&L above zero; “Not assigned” is the short leg finishing on the safe side of its strike.

Ten best underlyings

Ten worst underlyings

Grouped by underlying, which the per-contract tables below cannot show: APP’s worst single contract lost $12,393, but APP in total cost $248,049 across 30 contracts and 63 picks — 18% of every dollar this screener has lost, out of 342 tickers traded. “Total P&L” is everything that ticker made or cost; the single best or worst contract is kept as its own column so neither number has to stand for the other.

The gap between Contracts and Times picked is the part worth reading. Re-emitting the same stock night after night is a different bet from picking many once, and the screener does it in both directions: MRVL was re-emitted 110 times across 64 contracts for +$154,900, APP 63 times across 30 for −$248,049. A stock dropping 34% is a high-IV-percentile stock, so the gate that finds premium also finds falling knives, then keeps finding them until expiry. This is a concentration and sizing problem, not evidence the signal is broken — the same 342 tickers still return the figure at the top of this page.

Best contracts by strategy

Three per strategy rather than one global list, because a global list ranked in dollars can never reach the defined-risk strategies: every spread and condor resolved so far gained at most $3,470 and lost at most $3,013, while the single-leg lists run from −$12,393 to +$45,995. Four of the seven strategies would have no row on this page at all.

Within a strategy the ranking is still in dollars, so the biggest positions surface: a $1,500 strike ties up ~$150,000 and moves in units a $60 strike never will. These are the largest outcomes, not the sharpest trades — return on capital by strategy is in the table at the top.

Grouped by contract, so a contract the screener re-emitted across scan dates is one row with a “Times picked” count rather than several rows crowding the list. On a put row picked once, “Best P&L” and “Credit” are the same number — a short put cannot earn a dollar more than the premium it collected, however far the stock rises. (Where a contract was picked several times the Credit column averages those scans, so the two differ by that averaging, not by any extra gain.)

Only covered calls beat their credit. The shares appreciate up to the strike, so the ceiling is the credit plus that move — which is why the top of this table is $45,995 against a $7,107 premium while every other structure tops out at what it sold for. It is also the only strategy near both extremes: the same exposure to the shares cuts each way.

Worst contracts by strategy

Two strategies fail upward. Every iron condor and bear call spread row above lost because the stock went up through the call wing — the condor’s three worst are all MSFT closing near $513 against a $345 put strike, one gap sold at three adjacent strikes. Read only the put rows and you would conclude this screener has one way to lose. It has two.

The separation between structures is clean: no spread or condor row is worse than any single-leg row, $3,013 against $12,393. And losses have no ceiling the way gains do — the worst cash-secured put took in $3,173 and gave back $12,393, close to four times its credit, which is why a high win rate on its own says very little. The exception worth noting is GS: a bull put spread that lost $2,839, near its maximum, on a 5% move, because the width was narrow enough for a small drop to clear both legs. Defined risk bounds the loss; it does not require a large move to reach it.

What this does and does not show

• Ten expiration dates over nine weeks. This is not a multi-year validation, and it covers a single market regime. Treat it as evidence that the pipeline grades itself honestly, not as a forecast.
• It assumes one contract of every candidate. That is the only assumption that makes the total meaningful, and it required the capital shown above. No position sizing, no selection, no account of any realistic size.
• Credit is indicative, not a fill. This key has no options quote entitlement, so premium is the contract's end-of-day close. Real fills, especially multi-leg, would be worse.
• Outcomes are at expiry only. No early close, no rolling, no management of a losing position — all of which a real seller would do.
• Returns are not annualized. A month is too short a base to extrapolate from, and annualizing it would flatter the result.
• The per-$10,000 curve is a rescaling, not a simulation. It divides the collateral the whole program required into $10,000 slices and shows what one slice earned. It is not compounded and it is not what a $10,000 account would have done, which would depend entirely on which positions it could afford.
• Drawdown is the curve's, not an account's. It is the peak-to-trough fall of one contract of every candidate, with one point per expiration date (ten so far) and no ordering inside a date. An account that sized positions, or held through assignment instead of booking it, would have a different and almost certainly deeper one.
• Covered calls are scored as stock plus a short call, so the row carries what the shares did: credit + (min(close, strike) − entry) × 100. That caps the gain at the strike and lets a falling stock produce a loss even when the call expires worthless. Entry is the price recorded when the scan priced the row.
• Assignment is modelled, management is not. A put assigned at expiry is booked as a loss at that price; in practice you would hold the shares and could sell calls against them. The same applies in reverse to a called-away covered call. Every row is the outcome of leaving one position alone until expiry.