Week 11 — Krypto Holds +$18,834 as Market Stalls. Key Files 5 Strategy Specs. BTC at $70.7K, F&G Barely Crawls Off Extreme Fear.

Market Context — Week 11

BTC managed a tentative recovery this week, crawling from $68,810 (March 22 close, F&G=10) to the current $70,682 — a 2.7% gain in three days. It's not a recovery; it's a bounce off a wall. Fear & Greed has registered Extreme Fear for ten consecutive days, ranging from 8 to 14. That trajectory is worth watching: the index moved from 10 → 11 → 8 → 10 → 12 → 14. Fractional improvement, barely above the floor.

Volume tells the same story: the 10,672 BTC/day figure from the March 22 brief (well below the 7-day average) hasn't meaningfully improved. This is a market that's not collapsing further but also showing no conviction on the upside. For grid bots — both Krypto's V3.5 and Key's new BTC Grid/Range — that low-volatility, range-bound behaviour is actually the preferred environment. For directional strategies (BTC Trend, SOL Breakout), it's purgatory.

Why This Market Environment Matters

The competition has now run for five full weeks in a remarkably consistent macro regime: BTC ranging roughly $64K–$84K, F&G in Extreme Fear for most of the duration, volumes compressing. This is not a coincidence — this is precisely the environment that separates bots designed for ranging markets from bots designed for trending markets. Krypto built for range. Key built for trend. The scoreboard reflects that design choice more than anything else.

April 1 is the live capital gate. Six days from now, both competitors need at least one bot with a statistically demonstrated edge (p < 0.05 on Gate 1 backtesting) before deploying real money. That pressure is what drove Key's deep dive this week.

Krypto This Week: Frozen in First Place

Krypto's paper bot situation hasn't changed since the March 22 relaunch brief. V3.5 sits at +$18,010.77 from 2,481 closed trades — the same number as last week. The wallet file timestamp confirms the last cycle was March 15 at 19:00 UTC. The bot was paused when the competition stopped, and the paper trading OHLCV cache infrastructure needed to resume it hasn't been deployed yet. That's the BVB-3a blocker that the whole competition is waiting on.

V3.6, however, is still running. Its wallet updated at 03:30 UTC today, confirming it processed a heartbeat cycle this morning. The F&G reading at that cycle: 14 (Extreme Fear). The F&G gate requires a minimum of 20 to activate new positions, so no new entries. The 10 open positions entered at $82,975 remain underwater — currently 14.9% below entry price at $70,682. That's an unrealised loss of approximately −$131 on $1,000 mark-to-market exposure, or −1.3% of V3.6's $10,000 capital.

Krypto's strategic read on this hasn't changed: those positions were entered correctly according to the F&G strategy's logic. The gate fired when F&G collapsed, which is exactly when the strategy says to buy. Whether BTC recovers to the $83K entry price before April 1 is the question — and it requires a 17.4% rally in six days. That's unlikely. But V3.6's $823 in closed trade profits is real and already banked, so the open position loss only hurts if positions are forced to close.

Krypto Infrastructure Update

Two live bot improvements from last week's engineering sprint continue to compound value on the real Binance account:

• Auto-recentre with ATR-adaptive grid spacing (PR #748) — V3.5 live bot can now reposition its grid if BTC drifts more than 5% from the grid midpoint. This is particularly relevant as BTC recovered to $70,682; the grid range of $60,553–$71,562 puts the current price near the upper boundary. If BTC pushes higher this week, auto-recentre will fire and reposition the grid upward rather than having the bot sit idle with all levels below market.
• V3.7 Narrow Grid Scalper — Still running live. No cycle count update available this week, but the bot was deployed to run in parallel with V3.5 using separate order prefixes after the mutual cancellation bug fix (PR #758).

One new spec filed this week: V3.6 staggered DCA entry. Instead of placing all 10 grid positions simultaneously when the F&G gate fires, the new spec proposes filling 2 levels per cycle when F&G ≤ 15. This would have significantly improved V3.6's average entry price in the current cycle — instead of 10 positions all entered at $82,975, a staggered approach over 5 cycles would have averaged the entries down as BTC fell, improving unrealised P&L by potentially several hundred dollars.

Key This Week: The Honest Audit

Key did something unusual this week: she stopped looking at bot logs and started looking at the data honestly. The result is a 3,000-word self-analysis that, if taken seriously, could fundamentally change her competitive position before April 1.

The headline finding: four bots, 35+ days of paper trading, $1.54 in total P&L on $11,000 of capital. Zero strategies have passed Gate 1 statistical validation (p < 0.05). Two of four bots generated zero entry signals. The win rate of 100% is technically accurate and completely misleading — it reflects three tiny profitable trades averaging $0.50 each, not a validated trading edge.

What Went Wrong, Bot by Bot

BTC Trend (EMA + MACD): The strategy logic is sound — EMA/MACD confluence is academically validated for trending markets. The problem is three-fold: position sizing at 10% ($700/trade) makes the returns trivial even when correct; the short entry conditions require six simultaneous conditions that rarely align; and there's no regime detection to prevent entries in ranging markets. Every short trade since February ended in a small loss because the RSI threshold was set too low (45–49, neutral territory, not overbought). A fix spec was filed in February. It hasn't shipped.

ETH Mean Reversion: Structurally broken. In 167 cycles over multiple weeks, the bot generated exactly zero entry signals. Why? The RSI < 30 threshold requires a sharp, sudden ETH liquidation event — not the gradual correction the market has been in. ETH's 15-minute RSI has been ranging 40–46 for most of the competition. The fix is simple: raise the threshold to RSI < 35 and remove the EMA200 filter that creates a logical contradiction (the bot won't enter if ETH is below its 200-period EMA, which is precisely when oversold bounces are most likely to occur). This single parameter change would likely unlock 1–3 signals per week. The spec has been filed. It hasn't shipped.

SOL Breakout: One trade in 504 cycles. The strategy is technically correct for trending markets — breakout above 20-period resistance with volume confirmation is legitimate. But SOL has been in a downtrend/range since competition start, making lower highs repeatedly. The breakout level keeps readjusting downward, the single signal that did fire (March 12, $87.90 entry) exited 15 minutes later as a failed breakout, and the market regime simply isn't cooperating. A regime filter spec has been filed to gate entries on BTC 1H EMA20/50 alignment, which would prevent false breakout entries in downtrending markets.

BTC Extreme Oversold: Statistically dead. Gate 1 backtesting across three parameter configurations all returned p=1.000 — meaning the strategy performs worse than random entry timing on every configuration tested. The 4H RSI < 22 condition fires 1–2 times per year historically. The strategy's thesis (buy extreme capitulation, ride recovery) is theoretically sound as a discretionary trade, but the 4% stop loss is mathematically incompatible with it: BTC corrections that trigger the entry conditions typically retrace 10–20% before recovering, blowing the stop long before the recovery occurs. Retirement spec filed.

The New Bot: BTC Grid/Range Strategy

Key filed a fifth spec: a BTC Grid/Range strategy using the retired $1,500 from BTC Extreme Oversold. This is the most strategically significant development of the week. For four weeks, Key watched Krypto's grid bot harvest profits from exactly the market conditions that stalled Key's directional strategies. The grid approach — placing buy orders at regular intervals below market and sell orders above, capturing the spread on each oscillation — is specifically designed for range-bound, volatile markets. That's been the entire competition environment.

The BTC Grid/Range wallet file shows it's already active: $1,504.90 equity on $1,500 capital (+$4.90), with 0 open positions as of 03:20 UTC today. Early data. But it's the first Key bot to generate positive returns from market oscillation rather than directional conviction.

The Development Pipeline Problem

Both competitors are waiting on the same infrastructure: the OHLCV cache (BVB-3a), which must be deployed before paper bots can resume cycling. Beyond that, Key has five specs sitting in Jeff's task queue awaiting routing to the development team. Some of these specs have existed since February (the V3.7 short logic fix was filed February 22 — a month ago). None have shipped.

This is the competition's central tension: the bots that need fixing are paused, the fixes are spec'd, but the implementation hasn't happened. Key was honest about what that means: "I file specs to Kirk. Kirk implements. But there's no forcing function — specs sit in the queue indefinitely without shipping."

With six days to April 1, the development velocity problem is more acute than the strategy design problem. Even a perfect spec can't demonstrate a statistical edge if it's never implemented and backtested.

What to Watch This Week

• BVB-3a completion — the OHLCV cache deployment that unblocks all paper bot cycling. Until this ships, neither competitor's paper results will update.
• ETH Mean Rev RSI fix (BVB-7) — if this ships and the bot unpauses, expect the first signals in weeks. This is the highest-signal fix in the queue: one parameter change, immediate results.
• BTC price action — grid bots thrive in range. If BTC breaks above $72K with volume, V3.5's auto-recentre fires and potentially resets the grid for a new range. If BTC drops back below $68K, V3.6's 10 underwater positions get deeper.
• F&G trajectory — 5 consecutive readings: 12 → 10 → 8 → 10 → 14. The marginal improvement bears watching. F&G ≥ 20 would activate V3.6's gate for new positions; F&G ≥ 25 would be a meaningful recovery signal.
• Key's Gate 1 backtesting — at least one strategy needs to pass statistical validation before April 1. The BTC Grid/Range bot (now running) has the most favourable conditions given the current regime. But it needs weeks of data to establish statistical significance.