Premium Radar answers one question: is today a good day to be selling options premium at all? It is a read on market conditions, not a verdict on any particular contract.
On each scan it pulls the VIX and SPY from Yahoo Finance, then pulls full option chains for a fixed basket of ten liquid names from the CBOE. It scores the resulting picture from 0 to 100 across five pillars, and separately lists the cash secured put nearest 30 delta on the nearest monthly expiry for each ticker, sorted by annualised yield.
The basket is SPY, QQQ, AAPL, MSFT, NVDA, AMD, AMZN, GOOGL, META and TSLA. It is defined as BASKET at the top of premium_scan.py and can be changed there.
A high score means the environment favours selling premium. It says nothing about whether any specific contract in the candidate table is a good trade, and the engine applies no position sizing, portfolio rules, or earnings filter. See section 6.
Each pillar is worth 20 points. There are no penalties and no bonuses: five pillars at twenty points is exactly one hundred, so the score is a true scale rather than a clamped total.
| Pillar | What it measures | Points | |
|---|---|---|---|
| 1 | VIX level | Whether there is premium worth selling at all. Peaks in the low-to-mid twenties and falls away at both extremes. | 0 to 20 |
| 2 | VIX trend | The VIX against its own 10-day average. Falling volatility is IV crush working for you; a spike is not. | 0 to 20 |
| 3 | IV percentile | Where each name's at-the-money IV sits within its own past year of realised volatility, averaged across the basket. | 0 to 20 |
| 4 | SPY trend | Where the index sits against its 50 and 200-day moving averages, adjusted by the 5-day change. | 0 to 20 |
| 5 | Premium yield | What the ~30-delta puts actually pay, annualised and averaged across the basket. | 0 to 20 |
Each pillar is a piecewise-linear curve through a handful of control points, interpolated between them. These are the control points as they run today, straight out of premium_scan.py.
VIX → points: 10 → 2 · 13 → 5 · 15 → 8 · 18 → 13 · 22 → 18 · 28 → 20 · 35 → 14 · 45 → 7 · 80 → 3
The curve deliberately turns back down above 28. A very high VIX means rich premium, but it is also the environment in which short-premium positions get run over, so it is not scored as an unambiguous positive.
Percent above/below the 10-day average → points: −30% → 20 · −15% → 18 · −5% → 14 · 0% → 11 · +5% → 8 · +15% → 4 · +30% → 1
Average percentile → points: 0 → 0 · 20 → 4 · 40 → 9 · 50 → 12 · 60 → 15 · 75 → 18 · 90 → 20
Base: above both moving averages → 15 · above the 200 only → 10 · above the 50 only → 8 · below both → 3.
Then the 5-day change adds: −8% → −3 · −3% → −1.5 · 0% → 0 · +2% → +2 · +5% → +4. The result is clamped to 0–20.
Average annualised yield → points: 0% → 0 · 8% → 5 · 12% → 9 · 16% → 13 · 20% → 16 · 30% → 20, flat above that.
| Score | Band | Reading |
|---|---|---|
| 80–100 | Prime premium selling | Rich premium, vol falling, index healthy. |
| 60–79 | Favorable conditions | More going for it than against it. |
| 40–59 | Mixed conditions | Some pillars firing, others not. |
| 0–39 | Poor conditions | Thin premium, or a tape that punishes short puts. |
For each ticker the engine takes every put with 14 or more days to expiry, picks the expiry closest to 30 days out, and within that expiry selects the contract whose delta is nearest −0.30 among those with a live bid.
Yield is the mid price divided by the strike; the annualised figure scales that by 365 ÷ days to expiry. The table is sorted by annualised yield, high to low.
The candidate list is not risk-adjusted or filtered for quality. The highest-yielding row is usually the most volatile stock in the basket, which is exactly why it pays the most. Yield ranking is a starting point for a look, not a ranking of which trade is best.
Two public sources, neither of which needs an API key:
Yahoo Finance chart API for the VIX, SPY and each ticker's daily price history. CBOE delayed quotes for the option chains, which carry implied volatility, delta, bid/ask and open interest per contract.
CBOE quotes are 15 minutes delayed. Outside market hours they are simply the last thing that traded, so a weekend scan describes Friday's close. Results are cached for three hours; the Rescan button forces a fresh pull.
A true IV rank needs a year of stored implied-volatility history, and no free data source provides it. So pillar 3 measures something adjacent: where today's at-the-money IV sits within the distribution of that stock's own realised volatility over the past year.
The two are not the same. Implied vol usually trades above realised vol, so this proxy reads lower than a true IV rank would in normal conditions, and the reading is most useful as a relative comparison between names on the same day.
Every scan appends the day's IVs to cache/premium_history.json. Once roughly a year of daily scans has accumulated, a genuine IV rank becomes computable and this pillar can be switched over.
Ten large, liquid names. A market-wide reading drawn from ten mega-caps is a reading of those ten mega-caps, and it will miss dispersion elsewhere.
The score describes market conditions and option pricing. It reads prices, volatility, liquidity and dates. It does not read financial statements, revenue, debt, margins or valuation, and has no opinion on whether a business is good or a stock is cheap.
Specific limits worth naming, because they are the difference between this and a trading system:
The engine does not know when any company reports. A candidate whose expiry straddles an earnings announcement is listed exactly like any other, and will often top the table precisely because earnings are inflating its premium. Check the calendar yourself before acting on any row.
No portfolio rules. There is no position sizing, no diversification cap, no duplicate check and no record of what you already hold. It scores the day and lists contracts; everything after that is yours.
No view on direction. The engine does not forecast whether a stock will rise or fall. It assesses whether the environment for selling premium is favourable, which is a narrower question.
No automatic track record. The scan itself records nothing but the daily IV snapshot in section 5. Trades are logged by hand in Trade Ideas, and any row in the candidate table can be sent there with its score attached.
That is a journal you keep, not forward-tracking. A hand-kept log records what you chose to enter, so its results describe your decisions as much as the engine's — the engine publishes no prediction of its own that could later be graded.
Assignment is the real risk. With a cash secured put, being assigned means buying the shares. If you would not want to own the company, a high score on its options does not change that.
Premium Radar scores the day. Today's Opportunities scores each contract, using the fifteen signals below. It runs over a watchlist of higher-volatility names, scoring every put and call between 10 and 50 days out whose delta sits between 0.15 and 0.42 — roughly 750–800 contracts per scan — and shows the best two dozen.
| Signal | What it measures | Points | |
|---|---|---|---|
| 1 | Premium % of strike | How much the contract pays relative to the strike price. The single heaviest signal. | up to 40 |
| 2 | Days to expiry | How long until the contract expires. Preferred windows differ for puts and calls. | up to 30 |
| 3 | Capital efficiency | Lower strike prices tie up less collateral for the same structure. | up to 17 |
| 4 | IV rank | How expensive options on this stock are versus their own recent history. | up to 10 |
| 5 | Volume | How actively the contract traded. Thin contracts are harder to exit. | up to 5 |
| 6 | Delta alignment | Whether the contract sits in the intended risk band. | up to 3 |
| 7 | Support levels | Whether the strike sits near a price the stock has held before. | up to +3 |
| 8 | Proven ticker | A bonus for tickers with an established record in this system. | up to +8 |
| 9 | Market conditions | A broad-market adjustment, applied differently to puts and calls. | −10 to +10 |
| 10 | Ticker intelligence | An adjustment from accumulated per-ticker history. | −10 to +8 |
| 11 | Bid/ask spread | Tight spreads are rewarded; wide ones are penalised. | −8 to +3 |
| 12 | Open interest | How many contracts are outstanding. A liquidity check. | −8 to +3 |
| 13 | Covered call momentum | A penalty only. A stock in a strong uptrend is a poor covered call candidate, because the shares get called away just as they run. | 0 to −15 |
| 14 | Earnings proximity | A penalty near earnings, and a hard exclusion if earnings fall inside the contract's life. Source: Nasdaq calendar. | 0 to −15 |
| 15 | Open position penalty | A penalty only. Discourages stacking more of a ticker already held. | 0 to −25 |
Unlike the five pillars, these weights do not sum to 100. If every positive signal fired at once the total would be 140, so the sum is clamped into the 0–100 range. A score of 100 means the total hit the cap, not that every signal was perfect, and several different combinations reach it. Each card shows its own raw total next to the clamped score, so you can always see which case you are looking at.
This engine needs data the conditions score does not. All three are now live, and all are free:
Earnings dates come from the Nasdaq earnings calendar, swept day by day across a 60-day window and cached for 24 hours (about 1,350 symbols). A contract whose expiry falls on or after a report is excluded outright — not penalised, removed. A report landing within 7 days after expiry costs 8 points, within 14 days costs 4.
Per-contract volume comes from the CBOE chain alongside open interest, so the liquidity signals read the day's actual trading rather than just resting interest.
Position history comes from your own journal in Trade Ideas. The "proven ticker", "ticker intelligence" and "open position penalty" signals read it directly: tickers you have closed profitably score higher, tickers that have lost you money score lower, and a ticker you already hold is penalised for concentration. With an empty journal those three signals contribute zero, which is the correct starting state rather than a guess.
The same engine also scores PMCC diagonals — buy a deep-ITM call as a share substitute, sell a short-dated call against it. They are shown under their own filter on Today's board and in a separate section per ticker in the Ticker Browser.
The structure is scored as a whole, not as its short leg. Capital is the net debit rather than strike × 100; premium quality is the credit measured against that debit; liquidity signals read the worse of the two legs; and the momentum penalty runs at half weight, because a rising stock also lifts the long call instead of purely calling away shares.
A PMCC is only listed if net debit < strike width. Pay more than the width and the spread cannot be worth what you paid even when the stock runs straight through the short strike — you lose at maximum profit. Clearing that bar narrowly is also penalised, up to −18 points, because clearing it by $2 is not the same as clearing it by $1,125.
In practice the textbook recipe — a two-year 0.80-delta LEAPS against a 30-delta short — usually fails the width rule outright on higher-volatility names, because a long-dated call on a high-IV stock is mostly time premium. Each row shows the long leg's extrinsic percentage for that reason: it is what the short leg has to keep earning back.
The PMCP is the same structure mirrored: buy a deep-ITM put as a substitute for 100 short shares, sell a short-dated put against it. It runs through the identical scoring engine and the identical width rule, with the strike relationship inverted — the long put sits above the short put.
This is the one structure on the site that is not a bullish or neutral premium sale. Net delta is roughly −0.50: it profits when the stock falls. It is the mirror of a PMCC, not a cheaper cash-secured put — a CSP wants the stock to hold up, a PMCP wants it to drop, and they do not belong in the same rotation. Every PMCP row carries a Bearish badge for that reason.
Two signals are deliberately treated differently from the PMCC:
Capital efficiency is measured against Reg T initial margin on 100 short shares (50% of market value) rather than the cost of buying them. On high-IV names the long put is expensive enough that the debit can exceed that margin — the structure then ties up more capital than simply shorting, and the row says so rather than reporting a negative saving.
IV rank is the same proxy as section 5 — implied vol against the name's own realised-vol range, not a true one-year IV rank. It is worth 10 points of 140.
There is no tracked portfolio. The engine ranks and shows; it does not add anything to a paper book or grade itself later. The only portfolio rule it applies is a diversification cap of three contracts per ticker per scan, and the count it declined is printed at the top of the board. Nothing is logged unless you press "Log this trade".
Because nothing is auto-logged, the win rate shown on that page is your record from the journal, not the engine's. It measures the trades you chose to take.
The five pillars and their curves were chosen by judgment, not fitted to data. That is deliberate: an engine tuned on its own history will look good on that history and prove nothing. But it means the weights are a starting hypothesis rather than a result.
The rule is that weights change on evidence and a stated reason, not on hunches, and that changes are recorded here with the date. The daily IV history is the first piece of evidence being accumulated.
premium_scan.py as of 5 September 2026.