May 11-17, 2026: Seven-Three Week, +5.96R Net Across a Model Upgrade

Ten trades, seven winners, three losers, +5.96R net on the TP1 baseline. That is the scorecard for May 11-17, 2026 across the canonical-instrument set, and it sits in the upper band of the published expectancy. Cumulative equity opened at $100,000, climbed to $102,912.28 on Tuesday's opening NAS100 short, gave back $2,000 on Tuesday's US30 stop, recovered through Wednesday's mixed session, and closed the window at $111,923.39 after Friday's three-winner run. Through May 18, 2026, the system has banked +20.43R YTD across 99 trades from Jan 12 inception. The simulated $100,000 account at 2 percent risk per trade sits at $140,856.93 on a static basis and $147,308.83 on the compounded line. The six winners with case studies are published in parallel: the May 12 NAS100 short ran TP3 for +1.46R, the May 12 NAS100 short on a pullback to resistance ran TP2 for +0.70R, the May 13 US30 short ran TP1 for +0.83R, the May 13 GBPUSD short ran TP1 for +1.14R, the May 15 USDJPY long ran TP1 for +0.20R, and the May 15 GBPUSD short ran TP1 for +1.89R. The seventh winner, the May 15 EURUSD short at +2.00R, is the week's top trade and is covered here. The three losses are aggregated below. Last week's recap sits at the May 4 recap; April's monthly recap covers the longer window.

Act 1: Tuesday opens deep, then takes the week's first stop

Tue May 12 carried three trades. At 14:19 UTC a NAS100 short triggered on a VWAP rejection and ran the full ladder to TP3 for +1.46R, the deepest single TP1-baseline contribution of the window. At 15:42 UTC a US30 short on a pullback into the opening-range high cleared threshold at C+ grade, never reached TP1, and stopped for -1R. Two minutes later, at 15:44 UTC, a second NAS100 short on a pullback to a resistance zone ran to TP2 for +0.70R. Tuesday closed at $102,309.27 cumulative, +1.15R net on a two-winner, one-loser session.

Act 2: Wednesday mixes a four-trade session, GBPUSD debuts green

Wed May 13 produced four trades and the widest dispersion of the week. At 14:15 UTC a US30 short on a sell-the-rip at trend resistance with daily-pivot confluence ran TP1 for +0.83R. At 14:33 UTC a USDJPY long on a NY AM VWAP pullback continuation stopped early for -0.25R, a partial stop rather than a full -1R print. At 14:43 UTC a NAS100 short on a VWAP and Fib rejection stopped for -1R. Then at 15:54 UTC GBPUSD made its canonical debut. A short on a sell-the-rally into VWAP resistance ran TP1 for +1.14R. Wednesday closed at $103,742.44 cumulative, +0.72R net after two winners and two stops.

Act 3: Friday switches to Opus 4.7 and runs three straight winners

Fri May 15 produced three trades, all on the upgraded Claude Opus 4.7 master automations. At 14:15 UTC a USDJPY long on a pullback ran TP1 for +0.20R. At 14:18 UTC a EURUSD short on a NY AM session pullback ran for +2.00R, the largest single contribution of the week. At 14:22 UTC a second GBPUSD short on a VWAP pullback ran TP1 for +1.89R at B grade, the highest-graded entry of the window. The window closed at $111,923.39, +5.96R net. Three trades, three winners, +4.09R booked in a single session on the new model.

Pattern of the week

The week's recurring pattern was rejection and pullback shorts into VWAP and structural resistance, seven of which cleared TP1 or deeper, three of which stopped before reaching it. Eight of the ten trades were shorts; the two longs were USDJPY pullback continuations and split one win, one partial loss. The short side carried the week: NAS100, US30, GBPUSD, and EURUSD all sold rallies into resistance and the majority cleared at least TP1. The C+ grade described the entry card on most trades, the outcome varied with what the tape did after the trigger.

Inside the ten canonical trades, GBPUSD carried the cleanest green at +3.03R net across two debut trades, EURUSD banked +2.00R on its single trade, NAS100 was net positive at +1.16R across three trades, US30 went 1-for-2 for -0.17R, and USDJPY went 1-for-2 for -0.05R. The aggregate R-multiple shape was 1.46 - 1 + 0.70 + 0.83 - 0.25 - 1 + 1.14 + 0.20 + 2.00 + 1.89. That is a seven-three split where the winners averaged well above 1R and the three losers cost only -2.25R combined because one stopped partial.

How a seven-three week compounds to +5.96R

Seven winners contributed +1.46 + 0.70 + 0.83 + 1.14 + 0.20 + 2.00 + 1.89 = +8.22R at the TP1 baseline; three losers contributed -1 - 0.25 - 1 = -2.25R. The net at +5.96R is the arithmetic of a 70 percent win rate with an average winner of +1.17R and an average loser of -0.75R. The expectancy math runs strongly positive when the win rate clears 60 percent and the winners hold above 1R; this week did both. The partial -0.25R stop on Wednesday's USDJPY long is the one figure worth noting. The position took a structural exit before the full stop printed, which trimmed the loss column by 0.75R relative to a clean -1R.

Decision highlights

The Tuesday decision to hold the opening NAS100 short through the full TP1-TP2-TP3 ladder is the cleanest runner read of the week. The 14:19 UTC short cleared confluence on a VWAP rejection and the architecture did not flatten at TP1 or TP2 as the move extended. The position ran to TP3 for +1.46R, the deepest single contribution of the window, on the same trail-after-TP1 logic that governs every winner.

The Wednesday partial stop on the 14:33 UTC USDJPY long is the most instructive loss decision of the window. The position took a structural exit at -0.25R rather than running to the full -1R stop, which trimmed the loss column by 0.75R against a clean stop-out. The architecture's structural-exit logic did exactly what it is written to do. It read the invalidation before the hard stop printed and closed the position early.

The Friday decision to keep firing three sequential entries on the freshly upgraded Opus 4.7 automation is the discipline read of the week's green side. Between 14:15 and 14:22 UTC the system cleared confluence on USDJPY, EURUSD, and GBPUSD, and the Risk Agent did not pull sizing on the second or third entry as exposure stacked. Each trade was sized at the standard 2 percent of equity at trigger; the model upgrade did not change the sizing rule or the confluence floor.

Loss worth learning from

What the system saw that was right

Each of the three losers cleared the published confluence threshold at trigger. The Tuesday US30 short ran a pullback into the opening-range high that has banked positive R in other windows; the Wednesday USDJPY long cleared on a NY AM VWAP pullback continuation; the Wednesday NAS100 short triggered on a VWAP and Fib rejection. Every grade described an entry card that scored above threshold, and the Macro Agent did not veto regime on any of the three. The USDJPY long's partial -0.25R exit shows the structural-exit logic reading invalidation correctly before the hard stop.

What the system got wrong

The two full -1R stops share the regime-shift sensitivity that recent recaps have documented. The Tuesday US30 short held the opening-range retest at entry, then the local tape repriced inside the trade lifecycle and the runner never appeared. The Wednesday NAS100 short stopped cleanly off the entry on a VWAP and Fib rejection that did not hold. Both losses came inside the mid-week Opus 4.6 stretch rather than the Friday Opus 4.7 session, though three trades is too thin a sample to read that as a model effect rather than tape variance. The recap surfaces the pattern, not the trade-by-trade autopsy.

From the desk

Through May 18, 2026, the cumulative ledger reads +20.43R YTD across 99 trades from Jan 12 inception. The same $100,000 account at 2 percent risk per trade sits at $140,856.93 on the static line and $147,308.83 on the compounded line. The compounded number runs $6,451.90 ahead of the static, the visible footprint of disciplined sizing through a positive-expectancy edge across 99 trades. The +$47,308.83 compounded return reflects a season where the average winner has stayed above 1R and the average loser has held near -1R.

The honest reading is that this week did exactly what the published expectancy says a 70 percent win rate week should do at our risk-per-trade. Ten trades, seven winners, three losers, +5.96R net. A 70 percent win rate with an average winner of +1.17R and an average loser of -0.75R produces a strongly positive R-multiple, and that is what landed. The week also carried a mid-week model upgrade from Claude Opus 4.6 to Opus 4.7, and the architecture ran the same confluence floor and the same fixed-R policy across both versions.

The architecture point is that the sizing did not change to reflect the run or the upgrade. Friday's three sequential entries on the freshly deployed Opus 4.7 automation were sized at the same 2 percent risk the Risk Agent ran on Tuesday's Opus 4.6 winners. A discretionary trader closing the morning green on a new model would have second-guessed the sizing or stepped aside on the third entry. The system did not. The fixed-R policy and the published confluence threshold ran unchanged through the week and across the model boundary. The compounded versus static spread at +$6,451.90 is the visible footprint of that discipline over the longer window.

The TP1-baseline reading of +5.96R undercounts the broker fills on the winners that ran past TP1. Subscribers running scale-out at TP1 and beyond booked closer to the headline TP2 and TP3 figures on Tuesday's two NAS100 shorts. The full -1R stops printed identically across baselines; the Wednesday USDJPY long's partial -0.25R exit reflects the structural-exit logic, not a baseline difference. Cumulative figures exclude XAUUSD, retired from the active master automation set on April 24, 2026. From the SkyAnalyst Team.

What we're tuning

The two full-stop losses are filed under the regime-shift category the recent recaps have documented. The volume-aggregation fix already in testing is the relevant intervention; we are not opening a new instrumentation track from this week. The one signal worth flagging is the GBPUSD debut, two trades, both shorts into VWAP resistance, both clearing TP1 on the Claude automation. We are reviewing whether the new pair's confluence floor is calibrated correctly relative to the established instruments, and that review was opened when GBPUSD entered the canonical set.

The mid-week model upgrade from Opus 4.6 to Opus 4.7 is the second item on the bench. Friday's three-for-three run on the new model is too small a sample to read as a calibration shift, so we are holding the confluence floor and the fixed-R policy unchanged across both model versions and watching the next several windows for dispersion. No threshold change has been pushed on the basis of a single three-trade session.