# Corrected weekly Jev review

**$5 is the minimum opportunity to investigate, not the exit target.** The earlier table used the first close that crossed $5. That answered the threshold question and failed to show larger available profits. The corrected analysis below examines maximum later gains and a maximum-total-profit trade plan. The experimental live account has also been changed to let valid winners continue beyond $5.

Reviewed 2026-09-20 14:01 UTC to 2026-09-27 14:01 UTC: 10,080 Coinbase BTC-USD one-minute candles, zero gaps, plus 1,440 earlier warm-up candles. This is a fixed seven-day cut. Original trade answers are incomplete because the old system discarded them; this is disclosed throughout.

## Larger outcomes

- **Largest individual hindsight trade: $24.79 net**, LONG, 2026-09-20 14:19 UTC to 2026-09-21 20:26 UTC. Entry $80407.44, exit $87336.17. The first $5 crossing was at 2026-09-21 08:22 UTC, so stopping there would have left a much larger later gain unexplored.
- **Maximum-total-profit hindsight plan: $83.18 net across 12 non-overlapping trades**, under the fixed-stake cost assumptions below. This plan sometimes exits before an individual trade's ultimate peak so it can capture another subsequent swing.
- **5,090 qualifying entry/side candidates** have a best later net gain of at least $5 somewhere in the observed week. Each is paired with its maximum-profit later completed close, and the download also preserves its first $5 crossing. Candidates overlap; their gains cannot be added together.

These are hindsight upper bounds, not profits the bot could reliably predict. A live bot does not know a future high or low. The analysis considers completed close fills only; intrabar extremes or different stake/leverage choices would produce different bounds. No future candles are used in causal policy decisions or Jev snapshot requests.

## Assumptions and calculation

Every research trade reuses a fixed $100 stake at 3× leverage; profits are added without compounding. Entry and exit each pay 0.10% fees on notional and 0.05% slippage. Funding is approximated at 0.01% of entry notional per eight hours; production funding uses current notional. Both LONG and SHORT are evaluated. Exits must occur before the seven-day data cut; later outcomes are censored, not treated as zero.

The original count used a 24-hour holding limit and first $5 exits: 3,363 overlapping entry candidates and twelve earliest-finish episodes. The new full-window candidate count is 5,090. A forward dynamic programme evaluates every eligible entry/exit pair and chooses the non-overlapping plan with greatest total net profit. A separately implemented reverse dynamic programme agrees to numerical precision. Trades can close and open another side at the same completed close, with each order's costs charged.

The largest individual trade is not simply added to all other winners: it spans competing opportunities. Likewise, the larger later outcomes from the same original entries overlap and cannot be summed. “Maximum” here means maximum under these explicit completed-close, fixed-stake, observed-window assumptions.

## Best total-profit hindsight plan

| Side | Entry UTC | Exit UTC | Net gain | Hours | 1m RSI | 5m RSI | Worst intervening net if exited |
| --- | --- | --- | --- | --- | --- | --- | --- |
| LONG | 2026-09-20 14:19 UTC | 2026-09-21 09:38 UTC | $16.46 | 19.3 | 31.4 | 44.6 | $-0.85 |
| LONG | 2026-09-21 10:51 UTC | 2026-09-21 15:09 UTC | $6.85 | 4.3 | 30.0 | 56.2 | $-0.73 |
| LONG | 2026-09-21 16:39 UTC | 2026-09-21 20:26 UTC | $5.03 | 3.8 | 35.2 | 44.1 | $-0.68 |
| SHORT | 2026-09-21 20:26 UTC | 2026-09-22 05:32 UTC | $6.78 | 9.1 | 73.6 | 73.2 | $-0.87 |
| LONG | 2026-09-22 05:32 UTC | 2026-09-23 04:33 UTC | $6.56 | 23.0 | 32.2 | 32.9 | $-0.76 |
| SHORT | 2026-09-23 04:33 UTC | 2026-09-23 12:19 UTC | $5.87 | 7.8 | 67.1 | 65.3 | $-0.89 |
| SHORT | 2026-09-23 14:01 UTC | 2026-09-23 16:02 UTC | $7.09 | 2.0 | 67.5 | 60.6 | $-0.52 |
| SHORT | 2026-09-23 22:08 UTC | 2026-09-24 09:34 UTC | $5.16 | 11.4 | 74.3 | 47.0 | $-0.86 |
| LONG | 2026-09-24 09:34 UTC | 2026-09-24 14:29 UTC | $6.02 | 4.9 | 17.7 | 25.2 | $-0.45 |
| LONG | 2026-09-24 15:02 UTC | 2026-09-25 11:14 UTC | $5.25 | 20.2 | 27.8 | 40.7 | $-0.45 |
| SHORT | 2026-09-25 11:14 UTC | 2026-09-25 14:05 UTC | $6.31 | 2.9 | 79.3 | 75.8 | $-0.86 |
| LONG | 2026-09-25 14:05 UTC | 2026-09-27 13:33 UTC | $5.79 | 47.5 | 23.4 | 25.8 | $-0.58 |

Worst intervening net is measured from completed closes and includes costs if the position were exited then; it is not intrabar drawdown. Entry indicators below are independently recomputed research features, which may differ slightly from the saved production indicator implementation. The original production values survive in saved trade lessons and replay responses.

- **LONG, 2026-09-20 14:19 UTC, $16.46 net:** 1m RSI 31.4, 5m RSI 44.6, 5m EMA direction aligned, prior 1h return -0.10%, prior 4h return 0.11%, prior 24h range 2.30%, minute volume 0.97× prior-day average, ATR 0.046% of price. This can support reviewing continuation, but did not establish the future exit. Exact Jev answers at most of these entries no longer survive, so a specific internal reason for missing them cannot be established.
- **LONG, 2026-09-21 10:51 UTC, $6.85 net:** 1m RSI 30.0, 5m RSI 56.2, 5m EMA direction aligned, prior 1h return -0.68%, prior 4h return 3.03%, prior 24h range 6.25%, minute volume 0.37× prior-day average, ATR 0.075% of price. This can support reviewing continuation, but did not establish the future exit. Exact Jev answers at most of these entries no longer survive, so a specific internal reason for missing them cannot be established.
- **LONG, 2026-09-21 16:39 UTC, $5.03 net:** 1m RSI 35.2, 5m RSI 44.1, 5m EMA direction opposed, prior 1h return -0.40%, prior 4h return 0.44%, prior 24h range 7.18%, minute volume 1.56× prior-day average, ATR 0.101% of price. This required a turn before lagging trend confirmation; a trend-only entry gate could miss it. Exact Jev answers at most of these entries no longer survive, so a specific internal reason for missing them cannot be established.
- **SHORT, 2026-09-21 20:26 UTC, $6.78 net:** 1m RSI 73.6, 5m RSI 73.2, 5m EMA direction opposed, prior 1h return 1.24%, prior 4h return 1.76%, prior 24h range 8.01%, minute volume 9.46× prior-day average, ATR 0.130% of price. This required a turn before lagging trend confirmation; a trend-only entry gate could miss it. Exact Jev answers at most of these entries no longer survive, so a specific internal reason for missing them cannot be established.
- **LONG, 2026-09-22 05:32 UTC, $6.56 net:** 1m RSI 32.2, 5m RSI 32.9, 5m EMA direction opposed, prior 1h return -0.48%, prior 4h return -0.77%, prior 24h range 7.39%, minute volume 0.89× prior-day average, ATR 0.074% of price. This required a turn before lagging trend confirmation; a trend-only entry gate could miss it. Exact Jev answers at most of these entries no longer survive, so a specific internal reason for missing them cannot be established.
- **SHORT, 2026-09-23 04:33 UTC, $5.87 net:** 1m RSI 67.1, 5m RSI 65.3, 5m EMA direction opposed, prior 1h return 0.64%, prior 4h return 0.81%, prior 24h range 2.57%, minute volume 1.90× prior-day average, ATR 0.094% of price. This required a turn before lagging trend confirmation; a trend-only entry gate could miss it. Exact Jev answers at most of these entries no longer survive, so a specific internal reason for missing them cannot be established.
- **SHORT, 2026-09-23 14:01 UTC, $7.09 net:** 1m RSI 67.5, 5m RSI 60.6, 5m EMA direction opposed, prior 1h return 0.54%, prior 4h return 0.07%, prior 24h range 2.43%, minute volume 1.90× prior-day average, ATR 0.104% of price. This required a turn before lagging trend confirmation; a trend-only entry gate could miss it. Exact Jev answers at most of these entries no longer survive, so a specific internal reason for missing them cannot be established.
- **SHORT, 2026-09-23 22:08 UTC, $5.16 net:** 1m RSI 74.3, 5m RSI 47.0, 5m EMA direction opposed, prior 1h return 0.43%, prior 4h return 0.42%, prior 24h range 4.51%, minute volume 0.61× prior-day average, ATR 0.057% of price. This required a turn before lagging trend confirmation; a trend-only entry gate could miss it. Exact Jev answers at most of these entries no longer survive, so a specific internal reason for missing them cannot be established.
- **LONG, 2026-09-24 09:34 UTC, $6.02 net:** 1m RSI 17.7, 5m RSI 25.2, 5m EMA direction opposed, prior 1h return -0.99%, prior 4h return -1.37%, prior 24h range 3.62%, minute volume 1.42× prior-day average, ATR 0.105% of price. This required a turn before lagging trend confirmation; a trend-only entry gate could miss it. Exact Jev answers at most of these entries no longer survive, so a specific internal reason for missing them cannot be established.
- **LONG, 2026-09-24 15:02 UTC, $5.25 net:** 1m RSI 27.8, 5m RSI 40.7, 5m EMA direction aligned, prior 1h return -0.86%, prior 4h return -0.05%, prior 24h range 2.69%, minute volume 1.23× prior-day average, ATR 0.168% of price. This can support reviewing continuation, but did not establish the future exit. Exact Jev answers at most of these entries no longer survive, so a specific internal reason for missing them cannot be established.
- **SHORT, 2026-09-25 11:14 UTC, $6.31 net:** 1m RSI 79.3, 5m RSI 75.8, 5m EMA direction opposed, prior 1h return 0.89%, prior 4h return 1.70%, prior 24h range 2.23%, minute volume 1.85× prior-day average, ATR 0.078% of price. This required a turn before lagging trend confirmation; a trend-only entry gate could miss it. Exact Jev answers at most of these entries no longer survive, so a specific internal reason for missing them cannot be established.
- **LONG, 2026-09-25 14:05 UTC, $5.79 net:** 1m RSI 23.4, 5m RSI 25.8, 5m EMA direction opposed, prior 1h return -1.58%, prior 4h return -1.44%, prior 24h range 2.60%, minute volume 12.98× prior-day average, ATR 0.186% of price. This required a turn before lagging trend confirmation; a trend-only entry gate could miss it. Exact Jev answers at most of these entries no longer survive, so a specific internal reason for missing them cannot be established.

## The same original entries, with their larger later outcomes

| Entry | Side | First $5 exit net | Best later net | Best later exit |
| --- | --- | --- | --- | --- |
| 2026-09-20 14:20 UTC | LONG | $5.04 | $24.74 | 2026-09-21 20:26 UTC |
| 2026-09-21 08:23 UTC | LONG | $5.03 | $18.14 | 2026-09-21 20:26 UTC |
| 2026-09-21 08:45 UTC | LONG | $5.02 | $12.75 | 2026-09-21 20:26 UTC |
| 2026-09-21 13:38 UTC | LONG | $5.01 | $7.59 | 2026-09-21 20:26 UTC |
| 2026-09-21 20:29 UTC | SHORT | $5.37 | $13.80 | 2026-09-24 09:34 UTC |
| 2026-09-22 05:32 UTC | LONG | $5.12 | $6.56 | 2026-09-23 04:33 UTC |
| 2026-09-23 04:56 UTC | SHORT | $5.01 | $13.78 | 2026-09-24 09:34 UTC |
| 2026-09-23 14:01 UTC | SHORT | $5.14 | $9.43 | 2026-09-24 09:34 UTC |
| 2026-09-23 15:16 UTC | SHORT | $5.09 | $5.33 | 2026-09-24 09:34 UTC |
| 2026-09-24 09:34 UTC | LONG | $5.10 | $7.30 | 2026-09-25 11:14 UTC |
| 2026-09-24 15:02 UTC | LONG | $5.00 | $5.25 | 2026-09-25 11:14 UTC |
| 2026-09-25 11:15 UTC | SHORT | $5.14 | $6.27 | 2026-09-25 14:05 UTC |

## What changed in Jev

The original autonomous account, formerly labelled Zac, is now **Jev autonomous**. Its existing trial is retained as the baseline. The new **Jev improved** tab runs a separate experimental prompt and policy, version **net-swing-v2**, with an isolated $100 paper account. It is not a new underlying foundation model.

1. **Profit objective:** maximize expected net swing profit; consider opportunities above $5. There is no fixed $5 or $5.25 take-profit order in the improved account. A valid winner can stay open while the thesis remains supported.
2. **Profit protection:** reaching $5 net arms a trail. The floor is the greater of $3 and peak net minus the greater of $2 or 25% of peak net. For example, a $10 peak gives a $7.50 floor; a $20 peak gives a $15 floor. The floor is assessed at completed closes and can be exceeded adversely during a gap. This is an experimental give-back rule, not proof of an optimal exit. Jev can close sooner when the thesis fails.
3. **Larger context:** the prompt compares required price movement with ATR, full-day range, trend, actual price confirmation, observed support/resistance and reversal evidence. $5 no longer becomes a profit ceiling. The horizon is up to 72 hours, subject to the seven-day account's end.
4. **Entry checks:** ordinary entries need a model choice score of at least 0.65 and confidence at least 0.50, with aligned completed five-minute EMA/MACD and one-minute price movement. An observed recovery continuation or peak rollover can permit an earlier turn before trend confirmation, but needs stronger scores: at least 0.75 and 0.60. RSI extremes alone cannot trigger a reversal or ban a continuation. Model scores are not calibrated probabilities of making money.
5. **Loss and turnover controls:** a $3 net loss budget checked at completed closes, 30-minute entry cooldown after a close, no immediate reversal, and five-minute reviews. The review uses the next available completed candle if a scheduled boundary is missed. A 10% adverse intrabar emergency stop remains active. Stops and trails cannot guarantee their exact dollar floors.
6. **Evidence preservation:** the durable journal saves actual supplied model requests, answers, scores, execution vetoes, indicators and applied actions for all three bot accounts and the manual pot. The revised prompt also asks for a concise observable evidence category. This is an additional structured answer, not recovered hidden reasoning; it may disagree with the selected action, and the journal preserves both.

## What the tests actually establish

The earlier rule tests closed at $5, matching the mistaken exit objective. They are retained in the JSON as fixedFiveTests. The corrected tests below use the same four causal entry policies with the new trailing exit and 72-hour maximum holding time. They use past-only data. Each research trade uses a fresh fixed $100 stake, so results are additive P/L rather than compounded portfolio returns. Drawdown shown is realised trade-to-trade drawdown, not unrealised exposure.

The final three days had already been inspected in the earlier analysis. They are therefore a time-separated evaluation segment, **not an untouched holdout** for this revision. The exits were changed to match the user's corrected objective, not selected to achieve a desired percentage.

| Policy | Period | Trades | Net P/L | ≥$5 closes | Captured benchmark episodes | Capture | Realised drawdown |
| --- | --- | --- | --- | --- | --- | --- | --- |
| breakout | first four days | 9 | $-7.89 | 1 | 1/8 | 12.5% | $25.20 |
| breakout | final three days (previously inspected) | 7 | $-17.26 | 0 | 0/3 | 0.0% | $19.52 |
| pullback | first four days | 5 | $-7.91 | 0 | 0/8 | 0.0% | $12.34 |
| pullback | final three days (previously inspected) | 4 | $-12.76 | 0 | 0/3 | 0.0% | $12.76 |
| momentum | first four days | 5 | $15.69 | 1 | 1/8 | 12.5% | $6.19 |
| momentum | final three days (previously inspected) | 3 | $-9.68 | 0 | 0/3 | 0.0% | $9.68 |
| combined | first four days | 8 | $-4.74 | 1 | 1/8 | 12.5% | $22.05 |
| combined | final three days (previously inspected) | 4 | $-12.94 | 0 | 0/3 | 0.0% | $12.94 |

Capture means a completed tested trade with at least $5 realised net, in the same direction and overlapping a benchmark episode. Each episode counts once. This generous overlap measure does not prove prediction of its exact entry. Eight maximum-profit episodes lie wholly in the first four days, three in the last three days, and one crosses the split and is excluded from the period denominator. None of these rule proxies demonstrates 75% capture; the strongest early result still loses in the later segment.

### Actual revised Jev requests

Four purposively selected historical snapshots use only completed candles available at each timestamp, a flat $100 account and no future trade lessons. They are real requests to Jev, not rule simulations. Their small, selected sample does not measure a reliable capture rate. The first two are known hindsight opportunities; the third is the actual losing short entry; the fourth is a late-week range with a censored later outcome.

| Case | Status | Jev proposed | Applied by entry check | Confidence | Stated evidence | Execution reason |
| --- | --- | --- | --- | --- | --- | --- |
| saved-window continuation | complete | WAIT | WAIT | 0.45 | INSUFFICIENT_EDGE | WAIT |
| countertrend reversal opportunity | complete | WAIT | WAIT | 0.52 | INSUFFICIENT_EDGE | WAIT |
| actual losing short entry | complete | OPEN_SHORT | WAIT | 0.34 | INSUFFICIENT_EDGE | Entry probability below 65% or confidence below 50%; these are model scores, not validated win probabilities. |
| late-week range | complete | WAIT | WAIT | 0.37 | INSUFFICIENT_EDGE | WAIT |

The initial four net-target prompt requests are retained separately in the JSON. They all chose WAIT. The revised requests must not be described as successful opportunity capture merely because they avoided a loss. The sampled short proposal is low-confidence and rejected by the entry check; that verifies a guard, not a profitable overall strategy. **The requested 75% capture rate has not been achieved or validated.**

## Losses and actual record coverage

| Owner at time | Trial | Lifetime closes | Wins | Snapshot equity | Saved actions / lessons |
| --- | --- | --- | --- | --- | --- |
| Dan | touch-vs-autonomous-1789905804 | 4 | 0 | $95.62 | 141 / 0 |
| Dan | touch-vs-autonomous-1789914264 | 0 | 0 | $99.48 | 2 / 0 |
| Dan | touch-vs-autonomous-1789914384 | 6 | 0 | $92.39 | 500 / 1 |
| Dan | touch-vs-autonomous-1789980684 | 1 | 0 | $99.87 | 210 / 1 |
| Dan | dan-jev-managed-1789995127 | 36 | 15 | $86.06 | 500 / 20 |
| Zach | touch-vs-autonomous-1789905384 | 6 | 0 | $94.12 | 148 / 0 |
| Zach | touch-vs-autonomous-1789914264 | 0 | 0 | $100.00 | 2 / 0 |
| Zach | touch-vs-autonomous-1789914384 | 12 | 0 | $86.95 | 500 / 2 |
| Zach | touch-vs-autonomous-1789980684 | 6 | 2 | $100.58 | 500 / 6 |

Recovered 39 individual trade records or lessons, including 11 paired entries with saved model answers. The old system kept only 500 actions and 20 lessons, and stripped original answers from current trials. Full price history does not imply full decision history. Missing choices or private thought processes cannot be reconstructed; inference is labelled. Trial reset balances are not added as if one continuous account. Lesson closed_at can reflect later record creation after a stalled scheduler, rather than exact original execution time.

Current baseline Jev's original six closes produced two wins, four losses and +$0.58 net. Its largest saved win was +$5.77 at 1m RSI 75.5 / 5m RSI 95.9, which argues against a simple overbought ban. One short lost $1.87 after entering at 1m RSI 18.3 / 5m RSI 36; rebound risk is a plausible failure mechanism. Another long lost $0.83 with little market movement: fees and slippage overwhelmed the move. The previous main Sunday trial had twelve closes with no wins and finished around $86.95; saved entries often had marginal choice scores and low confidence, illustrating churn.

Some archived reversals were rule overrides, so they must not be blamed on discretionary Jev choices. Manual actions are likewise separate. A recorded indicator diagnosis is an association and hypothesis, not proof of causation or a guarantee that a filter prevents future losses.

- **Zach, LONG, 2026-09-20 14:17 UTC, $-1.13** (touch-vs-autonomous-1789905384). Price moved against the position or costs exceeded the move. Missing original entry evidence prevents a reliable individual causal explanation. Coverage: Only exit record retained.
- **Zach, SHORT, 2026-09-20 15:32 UTC, $-1.56** (touch-vs-autonomous-1789914384). Price moved against the position or costs exceeded the move. Missing original entry evidence prevents a reliable individual causal explanation. Coverage: Only exit record retained.
- **Zach, LONG, 2026-09-20 15:59 UTC, $-0.42** (touch-vs-autonomous-1789914384). Rule override, not an autonomous Jev choice. After reaching the lower RSI area, a rising one-minute RSI crossed up through 40 on a green completed candle Coverage: Saved model answer and indicators.
- **Zach, SHORT, 2026-09-20 16:10 UTC, $-1.41** (touch-vs-autonomous-1789914384). Rule override, not an autonomous Jev choice. After reaching the upper RSI area, a falling one-minute RSI crossed down through 60 on a red completed candle Coverage: Saved model answer and indicators.
- **Zach, SHORT, 2026-09-20 16:14 UTC, $-0.97** (touch-vs-autonomous-1789914384). Saved choice had low confidence (0.3) and marginal advantage over WAIT/HOLD. Repeated round trips paid costs without a durable move. Coverage: Saved model answer and indicators.
- **Zach, SHORT, 2026-09-20 16:17 UTC, $-0.75** (touch-vs-autonomous-1789914384). Saved choice had low confidence (0.35) and marginal advantage over WAIT/HOLD. Repeated round trips paid costs without a durable move. Coverage: Saved model answer and indicators.
- **Zach, SHORT, 2026-09-20 16:25 UTC, $-0.61** (touch-vs-autonomous-1789914384). Saved choice had low confidence (0.27) and marginal advantage over WAIT/HOLD. Repeated round trips paid costs without a durable move. Coverage: Saved model answer and indicators.
- **Zach, SHORT, 2026-09-20 16:32 UTC, $-1.07** (touch-vs-autonomous-1789914384). Saved choice had low confidence (0.22) and marginal advantage over WAIT/HOLD. Repeated round trips paid costs without a durable move. Coverage: Saved model answer and indicators.
- **Zach, SHORT, 2026-09-20 16:36 UTC, $-0.93** (touch-vs-autonomous-1789914384). Saved choice had low confidence (0.31) and marginal advantage over WAIT/HOLD. Repeated round trips paid costs without a durable move. Coverage: Saved model answer and indicators.
- **Zach, SHORT, 2026-09-20 16:40 UTC, $-0.78** (touch-vs-autonomous-1789914384). Saved choice had low confidence (0.29) and marginal advantage over WAIT/HOLD. Repeated round trips paid costs without a durable move. Coverage: Saved model answer and indicators.
- **Zach, LONG, 2026-09-20 17:53 UTC, $-1.65** (touch-vs-autonomous-1789914384). Price moved against the position or costs exceeded the move. Missing original entry evidence prevents a reliable individual causal explanation. Coverage: Saved model answer and indicators.
- **Zach, SHORT, 2026-09-20 19:35 UTC, $-1.40** (touch-vs-autonomous-1789914384). Recorded post-mortem: short opened on a green one-minute candle; entry volume was weak; price edge was too small relative to two orders of fees and slippage. This is a diagnostic hypothesis, not proof of causation. Coverage: Saved model answer and indicators.
- **Zach, LONG, 2026-09-21 04:18 UTC, $-1.50** (touch-vs-autonomous-1789914384). Saved choice had low confidence (0.28) and marginal advantage over WAIT/HOLD. Repeated round trips paid costs without a durable move. Coverage: Saved model answer and indicators.
- **Dan, LONG, 2026-09-21 07:48 UTC, $-1.36** (touch-vs-autonomous-1789914384). Recorded post-mortem: entry volume was weak; price edge was too small relative to two orders of fees and slippage. This is a diagnostic hypothesis, not proof of causation. Coverage: Saved trade lesson; original action timestamp/answer may be unavailable.
- **Dan, LONG, 2026-09-21 10:51 UTC, $-0.13** (touch-vs-autonomous-1789980684). Recorded post-mortem: price edge was too small relative to two orders of fees and slippage. This is a diagnostic hypothesis, not proof of causation. Coverage: Saved trade lesson; original action timestamp/answer may be unavailable.
- **Zach, LONG, 2026-09-21 16:41 UTC, $5.77** (touch-vs-autonomous-1789980684). Captured a meaningful cost-adjusted swing. High RSI did not prevent this win; use trend and price confirmation rather than an RSI ceiling. Coverage: Saved trade lesson; original action timestamp/answer may be unavailable.
- **Zach, LONG, 2026-09-21 20:50 UTC, $2.19** (touch-vs-autonomous-1789980684). Profitable but below $5. Compare holding time, costs and early exits; a larger target would not guarantee that this winner stayed profitable. Coverage: Saved trade lesson; original action timestamp/answer may be unavailable.
- **Zach, LONG, 2026-09-21 21:12 UTC, $-2.36** (touch-vs-autonomous-1789980684). Recorded post-mortem: the anticipated move failed despite aligned entry evidence; require stronger confirmation or a better price. This is a diagnostic hypothesis, not proof of causation. Coverage: Saved trade lesson; original action timestamp/answer may be unavailable.
- **Zach, LONG, 2026-09-22 16:14 UTC, $-2.33** (touch-vs-autonomous-1789980684). Recorded post-mortem: the anticipated move failed despite aligned entry evidence; require stronger confirmation or a better price. This is a diagnostic hypothesis, not proof of causation. Coverage: Saved trade lesson; original action timestamp/answer may be unavailable.
- **Zach, LONG, 2026-09-23 05:22 UTC, $-0.83** (touch-vs-autonomous-1789980684). Recorded post-mortem: price edge was too small relative to two orders of fees and slippage. This is a diagnostic hypothesis, not proof of causation. Coverage: Saved trade lesson; original action timestamp/answer may be unavailable.
- **Dan, LONG, 2026-09-24 08:39 UTC, $-0.94** (dan-jev-managed-1789995127). Recorded post-mortem: long opened against an aligned five-minute downtrend; price edge was too small relative to two orders of fees and slippage. This is a diagnostic hypothesis, not proof of causation. Coverage: Saved trade lesson; original action timestamp/answer may be unavailable.
- **Dan, LONG, 2026-09-24 09:43 UTC, $-2.63** (dan-jev-managed-1789995127). Recorded post-mortem: long opened against an aligned five-minute downtrend; long opened on a red one-minute candle; entry volume was weak. This is a diagnostic hypothesis, not proof of causation. Coverage: Saved trade lesson; original action timestamp/answer may be unavailable.
- **Dan, LONG, 2026-09-24 12:39 UTC, $-0.91** (dan-jev-managed-1789995127). Recorded post-mortem: long opened on a red one-minute candle; price edge was too small relative to two orders of fees and slippage. This is a diagnostic hypothesis, not proof of causation. Coverage: Saved trade lesson; original action timestamp/answer may be unavailable.
- **Dan, SHORT, 2026-09-24 13:38 UTC, $-2.83** (dan-jev-managed-1789995127). Recorded post-mortem: the anticipated move failed despite aligned entry evidence; require stronger confirmation or a better price. This is a diagnostic hypothesis, not proof of causation. Coverage: Saved trade lesson; original action timestamp/answer may be unavailable.
- **Dan, LONG, 2026-09-24 14:37 UTC, $0.61** (dan-jev-managed-1789995127). Profitable but below $5. Compare holding time, costs and early exits; a larger target would not guarantee that this winner stayed profitable. Coverage: Saved trade lesson; original action timestamp/answer may be unavailable.
- **Dan, LONG, 2026-09-24 15:49 UTC, $1.08** (dan-jev-managed-1789995127). Profitable but below $5. Compare holding time, costs and early exits; a larger target would not guarantee that this winner stayed profitable. Coverage: Saved trade lesson; original action timestamp/answer may be unavailable.
- **Dan, SHORT, 2026-09-24 16:16 UTC, $-0.17** (dan-jev-managed-1789995127). Recorded post-mortem: short opened against an aligned five-minute uptrend; short opened on a green one-minute candle; entry volume was weak; price edge was too small relative to two orders of fees and slippage. This is a diagnostic hypothesis, not proof of causation. Coverage: Saved trade lesson; original action timestamp/answer may be unavailable.
- **Dan, LONG, 2026-09-24 16:19 UTC, $0.77** (dan-jev-managed-1789995127). Profitable but below $5. Compare holding time, costs and early exits; a larger target would not guarantee that this winner stayed profitable. Coverage: Saved trade lesson; original action timestamp/answer may be unavailable.
- **Dan, SHORT, 2026-09-24 17:24 UTC, $0.26** (dan-jev-managed-1789995127). Profitable but below $5. Compare holding time, costs and early exits; a larger target would not guarantee that this winner stayed profitable. Coverage: Saved trade lesson; original action timestamp/answer may be unavailable.
- **Dan, SHORT, 2026-09-25 12:21 UTC, $-0.14** (dan-jev-managed-1789995127). Recorded post-mortem: entry volume was weak; price edge was too small relative to two orders of fees and slippage. This is a diagnostic hypothesis, not proof of causation. Coverage: Saved trade lesson; original action timestamp/answer may be unavailable.
- **Dan, LONG, 2026-09-25 13:02 UTC, $-0.96** (dan-jev-managed-1789995127). Recorded post-mortem: long opened against an aligned five-minute downtrend; long opened on a red one-minute candle; price edge was too small relative to two orders of fees and slippage. This is a diagnostic hypothesis, not proof of causation. Coverage: Saved trade lesson; original action timestamp/answer may be unavailable.
- **Dan, LONG, 2026-09-25 14:41 UTC, $0.94** (dan-jev-managed-1789995127). Profitable but below $5. Compare holding time, costs and early exits; a larger target would not guarantee that this winner stayed profitable. Coverage: Saved trade lesson; original action timestamp/answer may be unavailable.
- **Dan, SHORT, 2026-09-25 14:58 UTC, $-1.36** (dan-jev-managed-1789995127). Recorded post-mortem: price edge was too small relative to two orders of fees and slippage. This is a diagnostic hypothesis, not proof of causation. Coverage: Saved trade lesson; original action timestamp/answer may be unavailable.
- **Dan, LONG, 2026-09-25 15:21 UTC, $-1.25** (dan-jev-managed-1789995127). Recorded post-mortem: price edge was too small relative to two orders of fees and slippage. This is a diagnostic hypothesis, not proof of causation. Coverage: Saved trade lesson; original action timestamp/answer may be unavailable.
- **Dan, SHORT, 2026-09-25 15:56 UTC, $0.15** (dan-jev-managed-1789995127). Profitable but below $5. Compare holding time, costs and early exits; a larger target would not guarantee that this winner stayed profitable. Coverage: Saved trade lesson; original action timestamp/answer may be unavailable.
- **Dan, LONG, 2026-09-26 16:33 UTC, $-1.12** (dan-jev-managed-1789995127). Recorded post-mortem: long opened on a red one-minute candle; price edge was too small relative to two orders of fees and slippage. This is a diagnostic hypothesis, not proof of causation. Coverage: Saved trade lesson; original action timestamp/answer may be unavailable.
- **Zach, SHORT, 2026-09-26 21:15 UTC, $-1.87** (touch-vs-autonomous-1789980684). Recorded post-mortem: the anticipated move failed despite aligned entry evidence; require stronger confirmation or a better price. This is a diagnostic hypothesis, not proof of causation. Coverage: Saved trade lesson; original model answer unavailable.
- **Dan, LONG, 2026-09-27 03:55 UTC, $0.14** (dan-jev-managed-1789995127). Profitable but below $5. Compare holding time, costs and early exits; a larger target would not guarantee that this winner stayed profitable. Coverage: Saved trade lesson; original action timestamp/answer may be unavailable.
- **Dan, LONG, 2026-09-27 10:30 UTC, $0.46** (dan-jev-managed-1789995127). Profitable but below $5. Compare holding time, costs and early exits; a larger target would not guarantee that this winner stayed profitable. Coverage: Saved trade lesson; original action timestamp/answer may be unavailable.

## Manual separation and accidental trade

The accidental manual LONG at 27 September 14:23 UTC and its manual close at 14:25 UTC were voided at the user's request. The original Jev balance is restored to **$100.58125416**, fees to **$3.75744713**, funding to **$0.03920440**, and trade count to **six with two wins**. The $0.76153261 net loss was refunded, including the manual fees and funding. No autonomous trade was removed. The trial ID is unchanged. Original evidence remains archived and marked voided, with a visible correction record. This accidental trade occurred after the reviewed seven-day data cut.

Bot accounts reject manual commands, including from an old cached dashboard. **My paper pot** is a separate $100 manual-only paper account; Jev cannot open, close or reverse it. It uses 3× leverage and the same fee, slippage and funding assumptions, with a 10% adverse emergency stop. Public access remains as requested. Dan's existing held position is preserved. The three old research navigation tabs are removed; historical research data remains.

## Verification and limits

65 tests across 14 files pass, including manual-account isolation, rejected bot commands, winner continuation beyond $5, trailing profit exits, fee accounting, stale-data protection and missed review boundaries. Worker type checking passes. The complete candle set passes OHLCV, chronology and gap checks. The maximum-profit plan is independently verified by reverse dynamic programming.

These changes correct the objective and make future behaviour reviewable. They do not demonstrate a reliably profitable model or the requested 75% capture rate. A later assessment needs genuinely new market data, actual completed trades, net P/L, costs, realised and unrealised drawdown, and missed-opportunity review.
