Grid Alpha

HB_HOUSTON printed $142.55/MWh at 12:10 UTC on 2026-09-06, roughly 3.6x the 24-hour mean of $39.77/MWh. By the Monday morning snapshot at 06:03 CDT, the hub had faded to $37.31/MWh, essentially back on the mean. The spike lived and died inside a Sunday morning.

The timing matters. The 12:10 UTC print is 07:10 CDT, a Sunday morning hour, not a classic ramp window. The 24-hour low of $20.59/MWh printed less than three hours later at 15:00 UTC (10:00 CDT), which tells you the event was narrow: one or a handful of five-minute intervals inside a 260-bar series, not a sustained scarcity regime. What the tape does not tell you is whether the driver was scarcity pricing or congestion. The outage file shows at least 1,861 unplanned resource outage events over the prior seven days, including SCES, a 933 MW subbituminous coal unit on forced extension with 0 MW available since 2026-09-03. The SCED shadow-price feed logged at least 13,087 binding-constraint records over the same window, with constraint 6217__B (138 kV, limit 226) binding repeatedly. Both counts are sample-capped, so the true totals are higher. Neither dataset confirms coincidence with the 12:10 print, so treat the cause as unresolved.

That unresolved cause is the trade. If the spike was congestion-driven, the repeat risk lives in the same load pocket and the same constraint, and 6217__B's shadow price is the tell. If it was scarcity, the SCES outage and the broader forced-outage stack matter more than the constraint map. Either way, a 3.6x mean print on a Sunday morning with no corroborating news flow is the kind of event that either mean-reverts permanently or repeats at the worst time.

What I'd watch today: first, whether HBHOUSTON real-time prints above $60/MWh in any five-minute interval between 11:00 and 14:00 UTC (06:00-09:00 CDT), the same clock window as yesterday's event; a repeat there upgrades this from noise to pattern. Second, whether 6217_B shows a binding shadow price in the SCED feed during that window; if it binds and Houston separates from the ERCOT hub average by more than $20/MWh, the congestion read wins. Third, whether the 17:00-21:00 CDT evening ramp holds under $50/MWh; if Sunday's event was outage-driven, Monday's higher load is the harder test, and a quiet ramp kills the scarcity thesis.

> A 3.6x print that fades to the mean by the next open is either a one-off or a rehearsal; today's same-hour window decides which.

Not investment advice. For informational purposes only.

Yesterday's tape: PJM DOMINION HUB lmppeak >= 100 — verified (observed 118.26); PJM DOMINION HUB lmppeak >= 75 — not triggered (observed 72.26).

Grid Alpha — daily briefs · live dashboards, all nine markets · public call record

Dominion Hub printed $342.67/MWh at 22:35 ET Saturday night, roughly 7x the 24-hour mean of $47.12. By 23:45 the hub had round-tripped to its 24-hour low of $14.70, and as of 07:06 ET Sunday the real-time print sits at $24.19. The entire scarcity event lasted about an hour.

The spike landed in the dead of a Saturday night, when load is at its weekly floor and a $340 print implies something broke rather than something ramped. PJM's outage log shows 144 daily outage events over the last seven days, with sampled forced outage levels running 2,993 to 4,109 MW. The outage data is not timestamped tightly enough to pin the peak to a specific unit trip, and no binding constraint or reserve shortage flag accompanies the print, so the proximate cause stays open. What the shape of the series does say: 308 five-minute bars with a single isolated peak and an immediate collapse to below-mean levels is the signature of a transient, not a sustained scarcity regime. Whatever tripped, the system absorbed it within the hour.

The trade now is whether tonight's evening ramp repeats the signal. Sunday evening ramps are normally soft, but if forced outage levels stay in the 3,000 to 4,100 MW band into tonight, the same thin-supply setup that produced Saturday's print is still in the stack. Watch the day-ahead clearing price for tonight's HE19 to HE21 block when results post around 13:30 ET: if DA clears above $100 for those hours, the market is pricing a repeat; if it clears near the $47 mean, Saturday was noise. Second check: the real-time print at 18:00 ET, ahead of the ramp. If Dominion is already trading above $75 by then, the scarcity premium is rebuilding and the overnight DART spread widens.

> A one-hour $343 print on a Saturday night is a warning shot, not a regime; tonight's DA ramp prices decide which.

Not investment advice. For informational purposes only.

Grid Alpha — daily briefs · live dashboards, all nine markets · public call record

CAISO's real-time market printed a $1,081.95/MWh maximum nodal spread in the 10:20 UTC interval on September 5, with the low end anchored by PODBISHOP1UNITS-APND at -$989.97/MWh. The high end sat at the GYSRVLLE6 nodes at $91.98/MWh. This is not a system scarcity print; it is a localized dislocation in the Eastern Sierra, and it happened at 03:20 local time on a Saturday.

The decomposition tells the story. Bishop's -$989.97 LMP breaks down to a -$1,030.28/MWh congestion component against a system energy component of $35.76 and a small loss adjustment. GYSRVLLE6N001/N003/N005 carried +$38.65 of congestion against energy of $35.79. The energy component was essentially flat across both ends of the spread, roughly $35.8/MWh everywhere, so the entire $1,082 gap is congestion. This is a transmission event, not a fuel or load event. The Bishop print was also not a single-node artifact: PODCONTRL1POOLE-APND and PODCONTRL1QF-APND posted identical -$989.97 LMPs with identical -$1,030.28 congestion, and PODBISHOP1_ALAMO-APND came in at -$988.45. A cluster of adjacent nodes moving in lockstep points to a common binding element, whether an outage, a derate, or a local generation event. The sources carry no Bishop-area outage or constraint record, so the physical driver is unconfirmed. CAISO also logged at least 1,668 curtailment events over the prior seven days, though sampled volumes were near zero and the count may be truncated by query limits.

The trading question is persistence. A -$1,030 congestion component at 04:00 PDT on a Saturday can be a single-interval artifact of the overnight dispatch stack, or it can be the early signature of a constraint that binds through the solar build and into the evening ramp. If PODBISHOP1UNITS-APND still shows a congestion component below -$500/MWh at the 15:00 UTC (08:00 PDT) snapshot, the event is structural for the day and the Bishop-to-Gysrvlle spread becomes a live congestion position rather than a curiosity. If the component retraces toward zero by then, treat the overnight print as noise and stand down. Second check: whether GYSRVLLE6 congestion holds above +$30/MWh into the 17:00-21:00 PDT ramp window, which would confirm the corridor is binding when it matters commercially. Watch the CAISO outage and constraint postings for any named Bishop-area element; a shadow price on a named constraint converts this from anomaly to tradeable setup.

The wider context is worth one line. Negative nodal pricing of this magnitude in a low-load overnight hour is the kind of print that shows up when renewable output or must-run generation meets a constrained pocket. With curtailment event counts already elevated over the trailing week, the Bishop cluster deserves a place on the watch list even if today's print fades.

> When the energy component is flat at $36 and the spread is $1,082, the wire is the market: watch whether Bishop's congestion survives sunrise.

Not investment advice. For informational purposes only.

Grid Alpha — daily briefs · live dashboards, all nine markets · public call record

MISO's widest real-time nodal spread hit $555.67/MWh in the 10:55 UTC interval, 05:55 CDT Thursday morning. NSP.CMMPA.JANE in Minnesota cleared $523.09/MWh against NIPS.CC.WHITN in northern Indiana at minus $32.58.

The system energy component at both ends was identical: $90.37/MWh, the same print at all ten of the top and bottom nodes in the snapshot. Every dollar of the split is therefore congestion and loss. JANE carried $440.87 of congestion and minus $8.15 of loss; WHITN carried minus $124.88 congestion and $1.93 loss. Three more Whiting-complex nodes, NIPS.WHITNGT1, WHITNGT2 and WHITNST, printed exactly minus $32.58 on exactly minus $124.88, which is the signature of one local element pinning a generation pocket rather than a zone-wide move. The sign flips inside that same zone: NIPS.BAILY.ARR cleared $289.57 on $195.74 of congestion.

That $195.74 is worth more attention than the headline spread. Three Minnesota and Iowa nodes, NSP.OWATONSIG and NSP.OWATONA7 at $278.87 and ALTW.MMPA.OGWF at $276.36, all share a congestion component of $195.97. A near-identical shadow appearing at Baily in Indiana and at Owatonna in Minnesota suggests a second binding element with broad sensitivity, distinct from whatever is holding Whiting down. The snapshot does not name today's constraint. The one NIPSCO element on record this week, NIP13X01BARTONLKBNO2_XFMR, bound in five consecutive five-minute intervals from 12:50 to 13:10 UTC on Aug. 27 with breakpoints of 100 and 102; that is precedent for transformer-level binding in the NIP area, not evidence about this morning. Background congestion is dense either way: at least 5,693 real-time binding-constraint events in the trailing seven days, alongside 84 logged generation outage events and an Aug. 28 00:00 CDT snapshot showing 12,762 MW forced, 11,664 MW derated and 7,985 MW planned.

Positioning around a live split assumes it survives more than one interval, and this is one interval, 10:55 UTC, with no prior-interval history and no day-ahead print supplied for either node. Three checks make it tractable. If the congestion component at NIPS.CC.WHITN is still below minus $100/MWh at 14:00 UTC today, the constraint is holding through the morning load build rather than clearing as a single-interval artifact. If the Owatonna congestion component falls below $50/MWh by 16:00 UTC while JANE stays above $300/MWh, the Minnesota story is local to JANE and the two shadows are unrelated. And if NIPS.BAILY.ARR and NIPS.CC.WHITN are both still on opposite sides of zero at 18:00 UTC, the binding element sits inside the NIPSCO footprint, not on the interface into it. On the flow direction implied above, the data here shows sign and magnitude at nodes, not direction across a path; the snapshot does not name the constraint, so treat the north-to-Indiana framing as inference until MISO's binding-constraint feed prints a name for the 10:55 UTC interval.

> A $555.67 spread sitting on a flat $90.37 energy component is not a scarcity print: it is a map of which element is binding and which side of it you are standing on.

Not investment advice. For informational purposes only.

Grid Alpha — daily briefs · live dashboards, all nine markets · public call record

The maximum real-time nodal spread across CAISO hit $265.22/MWh in the 10:55 UTC snapshot, which is 03:55 PDT on a Wednesday. Overnight. The system energy component sat at $20.22/MWh across every node on the tape, so none of that dispersion is scarcity: it is transmission, and it is local.

The high print was PODELCAJN6UNITA1-APND at $199.98/MWh, of which $163.76/MWh was congestion, $20.22/MWh energy and minus $0.11/MWh loss. Roughly 30 miles east, PODCRSTWD6KUMYAY-APND cleared at minus $65.24/MWh on minus $98.36/MWh of congestion and minus $3.21/MWh of loss. Both load aggregation points on the tape, CLAPDAC-APND and CLAPBUNDLD-APND, printed identically at $68.00/MWh with $31.87/MWh of congestion, so the LAP is picking up a positive shadow price while a generator inside the same footprint is paying to inject. One caveat on the data: several nodes in the snapshot carry a 10:05 UTC stamp rather than 10:55, so the headline spread is not strictly a same-interval measurement. No load, net-load, outage or derate data accompanies the print, so the binding constraint behind El Cajon and Kumeyaay is unidentified.

Away from the nodal tape, the fuel-side numbers are the ones with duration. EIA puts US LNG exports at 17.4 Bcf/d in the first half of 2026, 23% above the same period in 2025, attributed to higher liquefaction capacity, with the STEO carrying 17.3 Bcf/d in 2H26 and 18.7 Bcf/d in 1H27. On the equipment side, Ansaldo Energia is exploring a return to the US generation market after more than 30 years away, which Power Magazine attributes to a widening gas turbine supply crunch. Neither moves a San Diego constraint tonight; both shape what marginal gas costs and what new steel is available to relieve one.

What I would watch is persistence, because a one-interval congestion print is noise and a repeating one is a constraint. If PODELCAJN6UNITA1-APND still shows a congestion component above $163.76/MWh at the 12:55 UTC interval, the binding element is structural rather than a five-minute dispatch artifact, and the El Cajon versus Kumeyaay pair becomes worth pricing against DA. If instead the maximum CAISO nodal spread falls below $68.00/MWh by 16:00 UTC today, treat the overnight print as a single-interval outlier and stand down. The second check is directional: if PODCRSTWD6KUMYAY-APND congestion is still more negative than minus $98.36/MWh at 12:55 UTC while the energy component holds near $20.22/MWh, the export path out of that pocket is the constraint, not aggregate supply. On the auto-topic failure that produced this template, the honest read is that the anomaly table referenced in the dek is not in the source material: there are no z-scores, no ranked setups, and no PJM, ERCOT, MISO, ISO-NE or NYISO data on this tape at all. The CAISO nodal split is the only anomaly the feed actually delivered, and it deserves a same-day recheck rather than a narrative.

> A $265.22/MWh spread on a flat $20.22/MWh energy component is not a market signal about supply: it is a map of one transmission constraint, and it is worth exactly as much as its persistence.

Not investment advice. For informational purposes only.

Grid Alpha — daily briefs · live dashboards, all nine markets · public call record

PJM's real-time market printed a maximum nodal LMP spread of $329.62/MWh in the 10:50 UTC interval Tuesday, 06:50 Eastern. That is a pre-dawn print, hours before the morning ramp, and it is entirely a congestion story: the system energy component sat at $39.60/MWh across every node in the snapshot.

The high side was concentrated in two 138 KV transformer nodes. COMERC 138 KV TR1 cleared at $301.09/MWh on $263.01/MWh of congestion, and REPUBSM 138 KV TR1 printed $295.50/MWh with $257.39/MWh of congestion. The low side was just as extreme: all four ASYLUM 23 KV buses (LIBRTY10, LIBRTY20, LOAD1, LOAD2) cleared at negative $28.53/MWh on congestion of negative $67.46/MWh, with SCOTCHHO 230 KV close behind. When the energy component is $39.60/MWh and two nodes are paying nearly eight times that while a cluster of 23 KV buses goes negative, the grid is telling you a constraint is binding hard between those pockets. What is driving it — an outage, a load pocket, a facility rating — is not in the data we have, and the honest read is that the cause is unconfirmed.

The structural backdrop adds a second thread. Per RTO Insider, PJM, American Municipal Power and the Independent Market Monitor each presented proposals to the Markets and Reliability Committee on August 30 to rework the reserve market, aimed at poor performance among some resources and growing forecast uncertainty from wind and solar growth. Three competing proposals at MRC means the reserve product design is now actively in play, with details and timelines still thin. A morning where congestion, not energy, does all the work is exactly the kind of tape that gives reserve-reform arguments oxygen: scarcity and deliverability are showing up in the congestion component before they show up in energy.

For the tape itself, the watch items are mechanical. First, persistence: if COMERC's congestion component still exceeds $200/MWh at the 13:00 UTC (09:00 ET) snapshot on September 2, this is a sustained constraint rather than an overnight artifact, and DART spreads on the COMERC-to-ASYLUM path become the trade to size. Second, breadth: if the negative prints at ASYLUM 23 KV are still below negative $20/MWh at that same 09:00 ET check, the low side is structural too, not a single-interval fluke. Third, the policy track: if the MRC posts a vote date or proposal comparison matrix for the three reserve-market designs before the next scheduled session, expect forward ancillary and capacity-adjacent products to start pricing design risk. If the spread collapses back under $50/MWh by the evening peak on September 1, treat the whole print as a transient and move on.

> When energy is $39.60 and the spread is $329.62, congestion is the market — and PJM's reserve redesign just got its Exhibit A.

Not investment advice. For informational purposes only.

Grid Alpha — daily briefs · live dashboards, all nine markets · public call record

PJM's real-time nodal spread hit $261.64/MWh in the 10:45 UTC interval Monday morning, 06:45 local, before the ramp even started. That is a pre-dawn print, not a peak-hour artifact, and it is entirely a congestion story.

The decomposition is clean. The system energy component sat uniform at $30.09/MWh across every listed node, so none of the spread is fuel or scarcity. At the top of the book, MACNEW_T 115 KV cleared $181.37/MWh on $151.26/MWh of congestion. At the bottom, LENOX 115 KV printed negative $80.27/MWh on negative $109.98/MWh of congestion. Same energy price, $261 apart. The constraint map, not the supply stack, is doing all the work this morning.

The negative side of the book is where the structure shows. Three 961 LEE 34.5 KV wind-named nodes, LEEDKBS, LEEDKWF1, and LEEDKWF2, all priced identically at negative $27.40/MWh. Uniform pricing across sibling wind nodes at that hour points to curtailment-adjacent conditions: overnight wind output into a load trough, with the local constraint trapping the megawatts. Whether those prints persist into the morning ramp or evaporate as load picks up is the open question, and the fact pack does not answer it.

Context around the tape: PJM, American Municipal Power, and the Independent Market Monitor each put reserve market redesign proposals in front of the Markets and Reliability Committee, aimed at poor resource performance and forecast uncertainty from growing wind and solar. A morning where wind nodes clear at negative $27 while constrained load pockets pay $181 is the exact stress pattern those proposals are trying to price. Separately, Delaware enacted strict clean power requirements for data centers, adding a compliance layer to the region's load-growth math.

One honesty note: the auto-topic pipeline failed overnight, so this is the default rundown, and the [MULTI] tag oversells it. The fact pack contains no ERCOT, MISO, CAISO, NYISO, ISO-NE, or SPP data. PJM is the whole tape today, and the anomaly table the angle references never made it into the source material.

What I'd watch: whether the $261 spread is a single-interval print or holds through the ramp. If congestion into the MACNEW_T corridor persists past sunrise, DART spreads in that pocket widen against hub, and the LENOX and 961 LEE negatives become a basis trade rather than a curiosity. If it collapses by 08:00 local, it was an overnight transmission artifact and the reserve-market conversation is the more durable signal.

> When energy is $30.09 everywhere and nodes are $261 apart, the grid is the trade, not the fuel.

Not investment advice. For informational purposes only.

Grid Alpha — daily briefs · live dashboards, all nine markets · public call record

PJM Western Hub real-time cleared $819.66/MWh at 23:10 UTC Saturday, 19:10 local, roughly 26 times the $31.21 24-hour mean across the same snapshot. By the Sunday 07:04 pull the hub was marking $22.53, below its own mean. Whatever bid the evening ramp did not survive into the overnight.

The mechanism is unconfirmed, and that matters more than the headline number. No binding-constraint record, reserve adder, load print, or gas mark accompanies the 23:10 interval, so there is no attributable driver: not a transmission limit, not a scarcity trigger, not a unit trip. The PJM outage feed shows at least 144 daily-outage anomaly events over the trailing seven days, but the returned sample is five rows, every one stamped 2026-08-24 00:00 EDT, with forced outages spanning 2,837 to 3,849 MW and maintenance spanning 2,606 to 6,274 MW against 0 MW planned in every row. That is a timestamp-collapse artifact, not a fleet condition, and it says nothing about 8/29. The snapshot itself has internal problems worth naming: 310 five-minute bars against a declared 11:05 to 23:10 UTC window that should hold about 145, and a $9.56/MWh low stamped 08:25 UTC Sunday, after the window closes. Treat the peak and the mean as real prints and the window label as unreliable.

The thesis that risk sits in the ramp block rather than the round-the-clock strip is plausible but not yet sized, and the missing piece is duration. If $819.66 was a single five-minute interval, the hourly integrated price for HE20 is a fraction of it and the block barely moves; if six or more consecutive bars cleared triple digits, the hour prints meaningfully and the evening-block premium is a real, recurring exposure. What I would pull first: the consecutive-bar count above $100/MWh, then the day-ahead Western Hub clear for HE20 to size the DART on the interval that actually spiked. Absent that, there is no DA/RT number to trade against, only a peak on a chart.

Second confirmation to run is geographic. If AEP-Dayton, Dominion, and the RTO aggregate printed alongside Western Hub, the event is a system-wide ramp-hour tightness and the on-peak strip carries it; if Western Hub moved alone, it is congestion, and the expression is a basis question rather than an energy-price question. Those two tests, duration and breadth, separate a structural evening-scarcity story from a five-minute data artifact, and nothing in the current fact set distinguishes them. No forward or on-peak block quotes for Western Hub are available either, so any claim that the evening block is mispriced going forward has no reference price behind it.

Worth noting what did not happen: no news feed reported the spike. Coverage over the 48-hour window ran to uranium output and electric cranes. A 26x hub print that draws zero trade-press attention is usually short, local, or both.

> A $819.66 print with no constraint record, no duration, and no neighboring-hub confirmation is a data point, not yet a position.

Not investment advice. For informational purposes only.

Grid Alpha — daily briefs · live dashboards, all nine markets · public call record

Every dollar of SPP's $466.21/MWh maximum nodal spread at 10:50 UTC is congestion. The energy component printed $0.93/MWh at every node in both the top-five and bottom-five lists, so OKGE.VOLT.0209 at $258.08 (congestion $257.09) and KCPL.VOLT.0271 at -$208.12 (congestion -$209.02) are not a supply story. They are a transmission story with a system marginal cost of essentially zero underneath.

The two ends of the spread have different shapes, and that matters more than the headline number. The high side is a pocket, not a point: OKGE.VOLT.0209, 0208 and 0195 all cleared $258.08 on an identical $257.09 congestion component, with OGE.TWELVE.ONE and OGE.TWELVE.TWO one cent behind at $258.07 on the same congestion. Five buses sitting on the same shadow price is a group of load trapped behind a binding element, not a metering artifact. The low side is the opposite: KCPL.VOLT.0271 stands alone, with the second-lowest node in the market, OKGE.VOLT.0116, at -$56.06, roughly $152/MWh above it. The rest of the negative tail is shallow and scattered: EDE_SWMP at -$55.57, SPS.SPINSPUR at -$55.06, and the wind bus WAUE.BEPM.TRIPLEHWIND at -$46.31. Note that OKGE.VOLT.0116 is negative while three of its zone-mates set the market high, so the Oklahoma zone is split against itself.

Two caveats belong on the same page as the number. First, the print is 05:50 CDT on a Saturday, the deepest minimum-load hours of the week, when a single element out of service can produce a spread this wide on very little flow. Second, the interval does not name the binding constraint or publish a shadow price, and there is no prior-interval or day-ahead comparison in the data, so this is one five-minute photograph rather than a trend.

What I would watch is persistence and structure. If the OKGE cluster holds a common congestion component across consecutive intervals, the binding element is structural and the DA/RT basis on the Oklahoma-to-Kansas City path is the thing to reconcile; if 0209, 0208 and 0195 decouple from each other in the next few intervals, it resolved as a switching event and the spread is untradable noise. The stated afternoon setup is the open question, not the finding: nothing in this snapshot establishes that a pre-dawn Saturday constraint survives into the load pickup. Separately, the outage and curtailment anomaly counts (at least 1,173 and 1,943 events over the trailing seven days) are sampled and dated 2026-08-22, and the sampled coal-on-outage readings of 1,404, 1,458.7, 3,854, 3,154 and 1,018.7 MW have no established causal link to this interval. SPP's three energy emergency alerts this summer and the interim tariff fix stakeholders endorsed on 2026-08-28 sit in the Western BAA, a different corner of the footprint from this path.

A $466 spread with a $0.93 energy component tells you where the wires are tight and nothing about whether the market is short. The trade lives in whether that congestion component repeats.

> When energy prints $0.93 at both ends, the spread is not a price signal about supply: it is a map of one binding element, and it is worth exactly as long as that element stays out.

Not investment advice. For informational purposes only.

Grid Alpha — daily briefs · live dashboards, all nine markets · public call record

CAISO's widest real-time nodal gap printed at $345.77/MWh: PODELCAJN6UNITA1-APND at $172.13 against AEC3001_N001 at -$173.64. Both legs are almost entirely congestion. Neither leg is what it looks like at first glance.

Start with the components. El Cajon's $172.13 breaks into $94.53 congestion, $73.42 energy, $3.91 loss: the constraint term is larger than the system energy price, so this node is not expensive because CAISO is short, it is expensive because power cannot get there. AEC300 runs the same trick in reverse and harder, with congestion at -$244.09 against a positive $66.64 energy component, meaning the node is paying $244 to inject. The tell sits one line below: GWFTRCY17B1, GWFTRCY27B1, GWFTRCY37B1 and GWFTRACY1B1 all cleared between -$143.00 and -$142.76 on an identical -$212.43 congestion component. Four separate Tracy buses moving in lockstep on the same shadow price is one binding constraint in the NP15 area, and AEC300 is almost certainly sitting behind the same wall, 32 dollars deeper in.

Now the clocks. El Cajon printed at 11:00Z; AEC300 and the whole Tracy cluster printed at 10:20Z, forty minutes earlier, on a $66.64 energy component versus $73.42. That is a stitched spread, not a same-interval one, and nothing in the snapshot confirms both legs ever cleared simultaneously. The clean same-interval comparison is El Cajon at $172.13 against CLAPDAC-APND and CLAPBUNDLD-APND at $107.33, both carrying $31.12 congestion at 11:00Z. The SP15 on-peak hub print of $132.25 is stale by six hours, dated 04:55Z, and should not be used as a reference leg for anything.

The timestamps also gut the obvious story. 11:00Z is 04:00 PDT on a Friday and 10:20Z is 03:20 PDT: this is deep overnight, so a -$244 congestion component cannot be solar oversupply looking for a home. The curtailment feed offers no help either. It reports at least 1,668 CAISO events over the trailing seven days, but all five sampled records show 0 MW curtailed, with MWh values down at 4.575e-05. A feed showing effectively zero curtailment alongside a -$244/MWh congestion print means the explanation is a transmission constraint or an outage that the anomaly data is not carrying, and no constraint name, shadow price or flow-versus-limit appears anywhere in the source.

What I would watch is whether the Tracy quartet reprints on the same -$212.43 congestion in the next few intervals. If it does, the constraint is persistent and northern, and the tradable question is the NP15 export path, not the El Cajon leg at all. If El Cajon holds $94.53 congestion while Tracy congestion decays, these are two unrelated constraints that a max-spread screen has bolted together, and the SDG&E-area premium is the smaller, more local story. Absent day-ahead prints at either node, DA/RT basis and any CRR value here are uncomputable, and a single snapshot cannot carry the claim that this structure survives the session.

> A $346 headline spread built from two intervals forty minutes apart is a screen artifact until both legs clear on the same clock; the four Tracy nodes sharing one shadow price are the only thing in this print that is actually a constraint.

Not investment advice. For informational purposes only.

Grid Alpha — daily briefs · live dashboards, all nine markets · public call record