Today's picks by confidence tier, what the recommendations have realized since the Schwab switch, and how the tiers compare with the v1 grade
Today's picks
Sign in to see the latest capture's picks by tier.
| Slice | Trades | Alerts/day | Backtested win | Backtested $/spread | v1 graded so far | v1 success | Realized win |
|---|---|---|---|---|---|---|---|
| All v1 | 219,257 | 192.5 | 86.4% | $88 | 67,432 | 28.6% | 86.7% |
| ≥ 80% | 159,542 | 172 | 89.8% | $109 | 45,381 | 24.5% | 85.9% |
| ≥ 90% | 73,429 | 138.6 | 96.1% | $172 | 25,735 | 5.3% | 79.4% |
| ≥ 93% strict | 54,825 | 125.2 | 96.8% | $193 | 19,636 | 4.6% | 78.3% |
Backtested win against realized win, by slice
About these numbers: each slice is the trades at or above that confidence; All v1 is every trade the formula produced. Backtested win and $/spread come from the research cells (confidence_cells, keep-50% close on the estimated price path), pooled by the confidence each cell yields; they are global, not per ticker. Trades, alerts/day and the v1 columns count the live table in the window: an alert is one ticker and expiration per capture day, divided by the days that slice had trades (confidence exists since 2026-10-09, so the tier rows start from that day); v1 graded and v1 success are the streak grade of trades that have expired and been graded, so they stay near zero until the first tiered trades expire. Realized win arrives with the v2 grader on Schwab bid/ask; the chart puts it beside the backtested rate for each slice, so a tier realizing under its research reads at a glance. The picks above and the Slack digest list the same trades; /system/alerts has the runs.
Since the Schwab switch (10/12/26)
strict list in the tiles · realized on stored quotes, after fees
Graded
23,822
15,674 still open
Won
78.3%
23,822 graded
$ per spread
$25
after fees
Closed at target
76%
3.3 days on average
Expired
20%
5% pinned
Worst spread
−$6,463
single graded spread
| Confidence | Recommendations | Alerts/day | Open | Graded | Win | $/spread | Closed at target | Pinned | Expired | Days to target | Worst |
|---|---|---|---|---|---|---|---|---|---|---|---|
| ≥ 93% strict | 39,496 | 143.2 | 15,674 | 23,822 | 78.3% | $25 | 76% | 5% | 20% | 3.3 | −$6,463 |
| ≥ 90% | 53,139 | 161.1 | 21,284 | 31,855 | 79.4% | $41 | 77% | 5% | 18% | 3.4 | −$6,463 |
| ≥ 80% | 122,223 | 202.7 | 49,517 | 72,706 | 85.9% | $103 | 83% | 5% | 12% | 4.1 | −$6,463 |
| All with a confidence | 167,279 | 215.9 | 66,866 | 100,413 | 86.7% | $134 | 83% | 5% | 12% | 3.7 | −$6,463 |
About these numbers: every recommendation the formula wrote since the switch to Schwab quotes, by the confidence it carried; each row includes the rows above it. The nightly v2 grade (/api/cron/grade-realized, 07:40 UTC Tue–Sat) closes a spread when the stored mid reached half the entry mid at any capture (close order filled at that price), buys it back at the 3:45 PM capture on expiry day when the stock sits within 1% of the short strike (pinned, at that capture's mid), or settles it at expiry against the stock's close, per contract after $1.30 a leg each way. Open is not graded yet. Win, $/spread, closed at target, pinned, expired and days to target are over graded spreads only; worst is the single worst graded spread. Alerts/day is one ticker and expiration per capture day. Rows the formula wrote with a negative credit (the mirrored pair of a real spread; the digest never posts them) are left out. The v1 streak grade over every trade since 2023 sits under “v1 history” below; /system/grading has the grader's runs.
v1 streak grade over every trade the formula produced; pick a confidence floor to see the same charts for that slice.
in relation to Days to Expiration
in relation to Days to Expiration
Total trades that are expected to succeed
Total trades that are uncertain
in relation to Days to Expiration
in relation to Days to Expiration
in relation to Days to Expiration
Realized insights
the realized (v2) grade, core names, one contract each
23,822 settled: 78.3% won, $25 a spread on average, 15,674 still open. Total $604,349. Under half settled on 15-30, 31-45, 46+ DTE: those read high until their expirations pass.
What the wins add and the losses take, per settled spread
dollars per spread by days to expiration at entry; the marker is the net. Break-even is the win rate a band must hold for its wins to cover its losses; a red win rate is under it. A faded band is early: under half settled
| Band | Settled | Won | Break-even | Avg win | Avg loss | Net per spread |
|---|---|---|---|---|---|---|
| 0-5 DTE | 10,452 (93%) | 73.8% | 72.4% | $231 | −$605 | $12 |
| 6-14 DTE | 9,465 (70%) | 78.2% | 81.2% | $162 | −$697 | −$25 |
| 15-30 DTEearly | 2,626 (36%) | 85.9% | 64.2% | $228 | −$408 | $138 |
| 31-45 DTEearly | 561 (30%) | 100.0% | — | $264 | — | $264 |
| 46+ DTEearly | 718 (13%) | 100.0% | — | $290 | — | $290 |
Did the confidence predict the outcome?
one dot per confidence bin and DTE band with 30+ settled, sized by count. On the dashed line the score was right; a red dot won five or more points less than it said
Realized dollars by entry week, one contract each
each bar is the net on the spreads entered that week, green up and red down; the line is the running total
Which expiration days cost the most
dollars lost on the spreads that expired that day, the worst ten; a long bar is losses landing together
What a dollar of margin earned per day held
net dollars per $100 of margin per calendar day, by days to expiration at entry: how well each band pays for the capital it ties up
About these numbers: the realized (v2) grade of every recommendation on the core names at the chosen confidence floor, in-session captures included, one contract each, after $1.30 a leg each way. "Since the switch" is every row written since 10/12/26 with a confidence, graded on stored quotes by the 07:40 UTC grade-realized cron; "All history" adds every earlier row the backfill lane has graded on the formula's own price basis (it reads a little high the way the formula's estimate does), filling in as the lane runs. The rows before the switch were never scored at the time; the confidence-backfill cron scores them with today's cells, so a historical tier is in-sample (the cells were fit on those rows) and a row whose shape has no cell at the floor shows only under "Every row". The first chart splits each settled spread's mean into what the wins add (average win × win rate) and what the losses take (average loss × loss rate); the table under it gives each band's win rate against its break-even, the win rate at which those payoffs cover each other (average loss over average loss plus average win: a band that wins $90 and loses $350 needs to win 79.5% of the time), and how many settled. A band under half settled is the early target hits, so it reads high until its expirations pass. The calibration chart puts the mean confidence of a bin against what it won. The weekly chart is realized dollars by entry week, each week's net as a bar and the running total as a line; the tooltip carries the margin still open from that week's entries. The expiration chart sums the losses by the day the spreads expired. Return per margin-day is net dollars over margin × days held, a same-day close counting as one day, per $100 of margin. Cached ten minutes.
Observation names since 10/16/26
captured, scored and graded; never alerted or taken
1 names, 175 strict recommendations, 146 graded at 43.1% won against the core's 86.3% over the same window (37 names, 16,809 graded): thin: 146 of 200 strict trades graded.
| Ticker | Since | Recs | Strict | Strict graded | Won | Predicted |
|---|---|---|---|---|---|---|
| DIA | 10/16/26 | 620 | 175 | 146 | 43.1% | 94.5% |
About these numbers: observation names run through the same captures, formula, confidence cells and nightly realized grade as the core (close order at half the entry mid on stored quotes, else expiry, after fees), but the digest, the day's sizing and the model accounts skip them. Won is the realized strict win rate; Predicted is the mean confidence of those graded trades. The set is thin until 200 strict trades are graded and holds when its win rate sits within 2 points of the core's over the same window. The Friday scorecard carries the same line.