Strategy Analysis

What Is a Pullback to Session Support?

A pullback to session support is a bread-and-butter intraday setup. After price establishes a directional move and prints a session high or low, you wait for a retracement back to a structural level — a prior swing point, a VWAP band, a Fibonacci cluster — and buy the dip (or sell the rally). The logic is that the level where institutional orders previously accumulated should attract buyers again on a retest.

The setup works when the broader trend is intact and the pullback is a healthy pause, not the beginning of a reversal. It fails when the structural level that looked solid was actually the last line of defense, and the retracement turns into a full breakdown. The challenge is that both scenarios look identical at the moment of entry. The only defense is a well-placed stop loss and the discipline to accept -1R when the level does not hold.

EURUSD on April 14 was a textbook example of the setup failing at the execution layer — not because the analysis was wrong in aggregate, but because the entry trigger conditions were never cleanly satisfied, and the model entered anyway.

How Claude Identified This Trade

The analysis was thorough. Claude's pre-trade breakdown ran 10,875 characters and identified a conditional long setup at the 1.17935-1.17965 session support zone. The macro picture supported EURUSD longs: DXY below its 5-day EMA and falling for three consecutive sessions, 10-year yields dropping to 4.274% (bullish for EUR), VIX declining at 18.26 in a normal regime. Three of the four macro inputs — DXY, yields, VIX — pointed the same direction.

The Trend Agent read EURUSD as NEUTRAL at 58% confidence in a TRANSITIONING regime, with key support at 1.17929 and resistance at 1.18083. The Macro Agent was lean_bear at only 42% confidence. The agents diverged — Macro nominally bearish, Trend nominally neutral — but Claude's analysis correctly identified that the divergence was largely timeframe-driven: the Macro Agent's bearish lean reflected the longer-run ECB/Fed rate differential narrative, not intraday flow dynamics.

The confluence scoring landed at 6 of 8, rated Medium-High (7/10). Factor (a) failed — Macro Agent was lean_bear at sub-60 confidence. Factor (f) was conditional — it required price to actually reach the entry zone. The other six factors checked out: Trend Agent aligned at 67% confidence, DXY confirmed, yields supported, 60-minute EMA stack was bullish, 15-minute RSI at 62.6 was healthy, and the calendar was clear.

On paper, this was a legitimate setup. The problem was in what happened next.

The Decision Journey — Five Evaluations, No Clean Trigger

Claude evaluated this trade 5 times over 6 minutes. The confidence scoring system reveals exactly where conviction never quite materialized.

Evaluation 1 (42% — WAIT): Price at 1.17964, sitting at the very top of the entry zone (97% position). The last three 5-minute candles were bearish (1.18023, 1.18021, 1.17990), price still falling. The 5-minute RSI at 43.6 had already broken below the 45-50 zone that the original analysis required for a valid bounce trigger. No buying conviction. Volume low at 23 versus 36 SMA.

Evaluation 2 (48% — WAIT): Price at 1.17947, now inside the zone. But the forming candle was bearish (open 1.17984, close 1.17969), 5-minute RSI at 40.5 — further below the trigger threshold, not bouncing off it. MACD histogram strengthening bearish on both the 5-minute and 15-minute timeframes. The pullback had not found a floor.

Evaluation 3 (52% — WAIT): Price at 1.17952, zone position 57%. Still no bullish confirmation candle. RSI at 40.0, MACD strongly negative. The AI's own reasoning noted "consecutive lower closes on declining volume" and acknowledged that entering would be "chasing into selling pressure."

Evaluation 4 (52% — WAIT): Price bounced to 1.18015 — well above the entry zone at 267% position. The 5-minute RSI had dropped to 41.3 on the prior candle, which broke below the required 45-50 bounce zone. The candle that touched the zone closed at 1.17975, above 1.17960 but not a clear engulfing or hammer pattern. Claude noted the realtime price appeared "stale or inconsistent with the most recent candle data," adding uncertainty.

Evaluation 5 (45% — WAIT): Price at 1.17995, zone position 200%, above the entry zone. The dip to 1.17954 had come and gone without producing the required entry signal. RSI broke the 45-50 floor. MACD was bearish on 5-minute. The setup's conditions were not met.

Five evaluations. Five "WAIT" decisions. Confidence peaked at 52% — never reaching the threshold that would justify entry. And yet the trade was executed. Entry at 1.17965.

This is the critical detail. The evaluation system said "wait" every single time. The entry happened despite that, because the automated execution layer operates on a separate trigger mechanism from the evaluation agent. The analysis identified a valid zone, the price entered that zone, and the system executed within the parameters. But the higher-level evaluation agent — the one designed to filter out entries where trigger conditions are marginal — never gave a green light above 52%.

Stop Loss and the Absence of Targets

Stop loss at 1.17885 — placed below the Trend Agent invalidation level of 1.17929, with a 4.5-pip buffer for slippage. This was structurally sound: if price broke below 1.17929, the bullish thesis was dead.

TP1 at 1.18045-1.18067 — prior swing highs on the 5-minute chart. This was the nearest structural resistance, approximately 1R-1.25R from mid-entry.

TP2 at 1.18110 — today's high, approximately 1.9R.

TP3 at 1.18200 — a round number extension target, approximately 3R.

None of them were reached. Highest TP hit: 0. Price entered at 1.17965, struggled upward briefly, and then reversed. The stop at 1.17885 was hit. Duration: 20 minutes. Result: -1.0R (SL).

What Went Wrong — An Honest Assessment

The analysis was not wrong. The macro alignment was genuine — DXY was weak, yields were falling, VIX was declining. The session support zone at 1.17935-1.17965 was a real structural level with multiple confluence points.

But three things conspired against the trade:

1. The entry trigger was never satisfied. The original analysis explicitly stated the required trigger: "Price pulls back into zone AND prints a 5min bullish engulfing or hammer candle with close above 1.17960, OR 5min RSI bounces off 45-50 zone without breaking below." Neither condition was met. The RSI broke below 45 to 40.0. No bullish confirmation candle appeared. The setup was conditional, and the condition was not fulfilled.

2. The Trend Agent's TRANSITIONING regime was a caution flag. A NEUTRAL direction at 58% confidence in a TRANSITIONING regime is materially different from the BULLISH 67% TRENDING regime that supported yesterday's successful EURUSD trade. Transitioning regimes are where reversals live. The system scored this as a 6/8 confluence trade, but the quality of those confluences was lower than the number suggested.

3. The 60-minute RSI was overbought at 77.8. The analysis noted this as a risk factor. Price had already moved 47 pips above the open by the time the entry zone was identified. The pullback was happening from an extended position, not from a healthy consolidation. Extended pullbacks are more likely to break through support than to bounce.

The stop loss worked exactly as designed. It capped the damage at -1.0R (SL). There was no slippage disaster, no gap through the stop, no system failure. The trade lost because the market did not cooperate, and the entry lacked the confirmation that would have filtered it out.

This is what a controlled loss looks like. The system risked 1R and lost 1R. The season record moves to 4W-1L. The benchmark publishes it with the same depth as the wins — because that is the entire point of running a transparent experiment.