GPT Takes Nasdaq TP3 — A Second Win Day on a Yields-Soft Tape

Strategy Analysis

What is a VWAP reclaim long on a yields-soft tape?

The setup is the cleanest pullback pattern in the index trader's playbook for trending days. Price opens with conviction, extends toward an intraday high, then pulls back through VWAP as the move's first wave digests. The recovery is the test. If price closes back above VWAP on a 5-minute candle and holds the level on the immediate retest, the pullback is complete and the trend resumes. If price fails to reclaim — or reclaims briefly and loses VWAP again — the read is wrong and the trade does not fire.

The logic is mechanical. VWAP is the volume-weighted intraday fair value. When trending price pulls back to it, three things happen at once: short-term traders who chased the high get flushed, longer-horizon buyers get a level worth defending, and the structural floor is re-established. A successful reclaim turns VWAP from resistance back into support. The structural stop sits below the swing low that defined the pullback, beyond round-number congestion. Either the level holds and the trend continues, or it doesn't and the trade is invalidated cleanly.

The setup fails for one reason: the regime is wrong. Buying a VWAP reclaim into a rising-yields tape is an expensive way to learn that rate-sensitivity drives NAS100 more than chart structure. The discipline is not "buy the reclaim." The discipline is "buy the reclaim when yields are soft, both agents agree, breadth confirms, and VIX is supportive." On May 6, every one of those conditions was in place.

How GPT structured the trade

The analysis ran the full sequence and arrived at a single qualifying setup, scored 6/7 on the framework's confluence gate. The 10-year yield was below its 5-day EMA — primary driver supports longs. Macro Agent at +68% bullish with rate-cited factors — confirmation. Trend Agent at 65% bullish — alignment. The 60-minute structure held above fast and slow EMAs with RSI elevated and MACD above zero — trend integrity intact. Price was rotating around VWAP and the 61.8% intraday retracement — the pullback geography the framework wants. The only deduction was the 15-minute momentum read: RSI and MACD were improving but not yet strong, which the framework flags as borderline rather than disqualifying. Six and a half of seven for longs. Zero for shorts.

GPT defined the trade with the precision the methodology demands. Entry zone: 28372–28388. Trigger: a 5-minute close back above VWAP near 28372, with price holding above on the retest and 5-minute MACD histogram staying positive or improving. Stop loss zone: 28290–28296 — below the 28285.2 Trend Authority invalidation buffer would have been too tight; the framework's 95-point structural stop kept the trade above the Macro/Trend invalidation while accommodating NAS100's overshoot tendency. TP1: 28468–28475 (1.0R structural shelf). TP2: 28535–28540 (the next meaningful resistance at 1.6R, just below the Trend Agent's 28538.4 daily R level). TP3: 28610–28625 (extension target at 2.4R, conditional on yields staying soft and DXY/VIX remaining below their 5-day EMAs).

Two evaluations — what changed in three minutes

The entry was a two-evaluation decision, and the gap between them is the entire story of how GPT's pattern-recognition reads in real time.

At 14:09:57 UTC, the first evaluation fired and returned WAIT at 84% confidence in the wait. The reasoning was structural: the most recent completed 5-minute candle had closed at 28325.6, below VWAP near 28372.5. Price had touched VWAP on the current forming candle but had not yet printed a confirmed close above it. The MACD histogram was positive but cooling — 10.32 to 6.76 to 5.52 across the prior three reads — which the framework treats as momentum weakening, not improving. The trigger had not printed. The model held.

Two minutes and twenty-five seconds later, at 14:12:22 UTC, the second evaluation flipped the decision to ENTER at 68% confidence. The reasoning was that the prior 5-minute candle had now closed at 28382, decisively above VWAP at 28372.7. The next forming candle was holding above the level. The MACD histogram had reversed direction and was improving — from 6.76 to 7.96 to 8.79. The reclaim-and-hold trigger that the analysis specified had now printed. The model also flagged the imperfections honestly: 15-minute momentum was not strong, and intraday resistance around 28408–28412 capped immediate upside, so this was a good-but-not-perfect entry rather than a clean A+ setup. Confidence dropped from 84% (in waiting) to 68% (in entering) precisely because the model acknowledged the trade-off. The structural trigger had fired, but the momentum confirmation was incomplete.

This is the rhythm that has defined GPT's wins this season — patient enough to wait one evaluation for the trigger to actually print, decisive enough to enter on the first valid signal without demanding a 90% confluence read the tape rarely offers. The entry filled at 28385.

Why 68% confidence was the right number

The entry confidence number deserves its own paragraph because it is structurally identical to the read that took GPT into US30 long on Tuesday — entries in the 64–76% confidence band, not the 85%+ band the framework reserves for textbook A+ setups. The pattern matters.

A 68% entry on NAS100 today reflects two competing reads. The structural read was strong: VWAP reclaim confirmed, regime supportive, both agents aligned, breadth and yields cooperating. The momentum read was imperfect: 15-minute RSI and MACD improving but not strong, price near intraday resistance, regime classified TRANSITIONING rather than TRENDING. The framework treats those signals as reasons to reduce conviction, not reasons to skip. The discipline is to enter at the structural trigger with appropriately sized risk, not to wait for a 95% read that the tape rarely offers.

A model that only fired on A+ confluence would have skipped today's NAS100 reclaim and missed +$1,021.20 (TP3). That is not a hypothetical — it is the same pattern that has cost models entries throughout the season. The benchmark measures whether the model can convert good-enough setups into clean wins when good-enough is what the market is offering. Today, NAS100 was good-enough. GPT executed.

The exit and the full extension

The trade ground higher in the way yields-soft continuation tapes do — slow, methodical, with shallow pullbacks that never threatened the VWAP floor. TP1 at 28468–28475 took the first slice. TP2 at 28535–28540 took the second. TP3 at 28610–28625 took the rest. The position closed at 28612.2 on a TP3 print, +$1,021.20 (TP3) on the broker's executed size, +2.4R (TP3) on the framework's planned-risk math.

The duration was 5 hours 43 minutes from 14:12 UTC entry to 19:56 UTC exit — long enough to need conviction, short enough to capture the day's full move without overstaying. The setup geometry maxed out at the 28610–28625 TP3 zone, conditional on yields staying soft and DXY/VIX cooperating. All three conditions held into the close. There was no greedy trail searching for an extension the structure never specified. The plan said TP3. The market delivered TP3. GPT took TP3.

This matters because the framework is built around taking the levels the market actually delivers, not the levels the model originally hoped for. TP3 was on the analysis menu only as a conditional extension — and the conditions held. The execution matched the plan exactly.

The two-day pattern

GPT has now strung together three clean wins across two consecutive sessions. Tuesday: US30 long, +$1,381.80 (TP3). Tuesday: US500 long, +$1,034.43 (TP2). Wednesday: NAS100 long, +$1,021.20 (TP3). Combined two-day P&L on the winning trades: +$3,437.43.

Three wins, three indices, three different evaluation paths. US30 fired on a single decisive evaluation. US500 fired on a single decisive evaluation. NAS100 took two — the first to wait for the trigger to print, the second to commit when it did. Each entry sat in the same 64–76% confidence band. Each exit took the level the market delivered without forcing further extension. The framework is doing its job, and the model is reading it the way the framework was built to be read.

This is also the second NAS100 long GPT has banked clean this season. The April 24 NAS100 read ran the same playbook: bullish macro tailwind, structural pullback entry, multi-target extension, end-of-day TP banked. Two for two on Nasdaq longs in the rate-sensitive tape. The pattern is consistent enough to call a tendency.

What this trade says about GPT's process

Three things stand out from this NAS100 read.

First, GPT respected the trigger discipline. The first evaluation at 14:09 wanted to enter — the structural read was already favorable — but the framework's 5-minute reclaim-and-hold trigger had not printed yet. The model waited 145 seconds for the actual trigger candle to close. That is the discipline that separates a structural bias from a structural execution. Bias without trigger is impatience. Trigger without bias is chasing. The model held both in correct order.

Second, GPT entered on the first valid signal without demanding perfection. Once the 14:12 reclaim-and-hold printed, the model committed at 68% confidence — flagging the 15-minute momentum imperfection and the nearby resistance honestly, then taking the trade anyway. The framework specifies that good-enough setups in supportive regimes are worth the entry. The model executed that specification.

Third, GPT took the planned TP3 and stopped there. The setup geometry called for TP3 at 28610–28625 conditional on yields staying soft. Yields stayed soft. Price reached 28612.2. The model exited at the planned level. There was no extension into territory the analysis did not specify, no greedy trail looking for an unmodeled tail. The discipline of taking what the framework specifies — not what the model hopes for — is the same pattern that delivered both Tuesday wins.

NAS100 today was a Tier-1 long that arrived qualified on the first qualifying trigger. GPT executed it cleanly. The trade is on the scorecard.

The season inflection — written in three wins

After Wednesday closes, the season scoreboard reads 9W–8L for GPT, 52.9% win rate. The previous baseline — heading into Tuesday — was 6W–8L. Three trades, three wins, the win rate flipped above 50% for the first time since early in the experiment. The streak is 3.

Whether three wins in two sessions reflects a regime tailwind GPT happened to ride or a process that has genuinely sharpened is a question for a longer sample. The honest read is some of both. The risk-on/yields-soft tape that opened Tuesday and held into Wednesday is the cleanest backdrop for index continuation longs that the season has offered, and GPT is built to capitalize on exactly that environment. When the tape turns mixed again — and it will — the same framework will need to navigate failed reclaims, fade-the-rip setups, and choppier multi-evaluation reads. Today's clean three-target extension is what the playbook looks like when conditions cooperate. The harder reads are coming. For now, the scorecard is moving in the direction process and conditions both predicted.