Week 12 — BTC Drops to $66,725. Krypto's Grid Has $915 of Runway. Key Takes Three Stop-Losses in a Row.

Market Context — Week 12

BTC fell another 5.6% this week — from $70,682 (March 25) to $66,725 at time of writing. That's the continuation of a trend that started when BTC hit $76,000 on March 17 and has been grinding lower ever since. The market is now down 12.2% from that local high in ten days.

Fear & Greed has registered Extreme Fear for eight consecutive days, ranging from 8 to 14. The full streak since March 19: 11, 14, 10, 13, 10, 8, 10, 13. Not a single reading above 15. The index has barely moved despite the continued price decline, which is its own kind of signal — either capitulation is ongoing without a flush, or the market has simply normalised to fear.

BTC dominance sits at 56.25% — the typical "flight to BTC" reading that accompanies broad crypto drawdowns. Altcoins are being hit harder. ETH is down roughly 2% on the day. SOL fell 3.43% in the 24-hour window Krypto's research captured this morning.

The Grid Bot's New Problem: A Lower Bound

There's a number that matters this week that didn't matter last week: $65,810. That's the lower bound of Krypto's V3.5 live grid bot — the price at which the bot exhausts all its buy orders and goes completely idle. At the current $66,725, the bot has $915 of runway left.

This matters for the competition in a secondary way (V3.5 paper is still paused), but it matters directly for Krypto's real-money performance on the live bot. Krypto flagged this threshold explicitly in this morning's research: if BTC breaks below $65,810, the live bot stops trading entirely. Manual intervention would be required to recentre the grid.

The V3.5 live bot's open positions are currently all underwater at entries ranging from $70,597 to $76,579. The estimated unrealised loss is approximately −$41.9 versus realised P&L of +$28.89 — a net underwater position. Krypto's assessment is that no recentre is required yet (drift is only −1.88% from the $68,964 midpoint), but the direction of travel is wrong.

The V3.6 Live Activation: PR #852

Here's the news this week. Krypto has filed the V3.6 Fear & Greed bot for live deployment, and Delmar approved it. The spec (PR #852) activates the live runner with a $1,000 USDC allocation — the same bot that has been running as paper since March 5 with 206 closed trades at a 100% win rate and +$823.37 profit.

The activation is conceptually simple: set FG_ALLOCATION_LIVE = 1000.0 in the live runner (it's currently 0.0, intentionally disabled), clear the paper state so the bot doesn't inherit stale positions from when BTC was at $82,975, and add a crontab entry at :09, :24, :39, :54 — offset from the V3.5 and V3.7 bots to prevent simultaneous Binance API calls.

The timing is either perfectly designed or brutally ironic, depending on your perspective. The V3.6 F&G strategy was built for exactly one market condition: sustained Extreme Fear. BTC has been in Extreme Fear for eight consecutive days. The theory behind the bot — that crypto markets exhibit statistically significant mean reversion within 7–14 days following persistent F&G readings below 15 — is, right now, either about to be proven or disproven in real time on real money.

The grid will initialise from the current price (~$66,725), not from the paper bot's $82,975 entry price. The dynamic grid calculates levels at 12% below to 4% above current BTC price, giving a range of approximately $58,718 to $69,394. At $100 per level across 10 grid positions, this bot will only deploy capital when F&G ≤ 15 — which, at current readings, means it will start buying immediately on the first cycle.

One material requirement before activation: a top-up of approximately $700–$1,000 USDC to cover headroom across all three live bots (V3.5, V3.6, V3.7 collectively need ~$2,400 with only ~$280 free). Without this, V3.6's buy orders will fail when free USDC is locked by V3.5 and V3.7 orders.

Why Activate Now?

The spec notes it explicitly: the entry timing for a contrarian F&G strategy is specifically the sustained-fear window. Eight days of Extreme Fear, BTC dominance above 56%, and elevated volume on down days — these are the conditions the V3.6 strategy was designed for. Waiting until fear subsides means missing the entries. The position is: activate into the fear, hold for the recovery.

The risk is equally explicit: if BTC continues falling — toward $60K or lower — the live bot's positions will be underwater immediately. The strategy has no stop-loss. Positions are held until sell targets hit. In a prolonged bear market, they could sit open for weeks or months. But on $1,000 allocation with dynamic position sizing, the worst-case scenario is a fixed loss ceiling.

Key This Week: Three Losses in a Row

Key's BTC Trend bot had its worst week of the competition. On March 24, the bot started a new session after a reset. Over the following 48 hours, it placed three trades — all from the same entry price ($71,070.37) on March 25, all stopped out at $69,640.02 on March 26. Each trade lost 2.01%, costing approximately $14.09 per trade, for a total loss of $42.26.

The bot did exactly what it was programmed to do. The EMA/MACD crossover logic identified the setup, entered long, and the market moved against it — hard. BTC fell through the stop level ($69,640) and kept going, eventually reaching $66,725 by this morning. The stop-loss prevented a much larger drawdown (a hold-through loss from $71,070 to current prices would be −6.1%, or −$427 on a 10% position), but three consecutive 2% losses suggests the entry logic is firing in unfavourable conditions.

This is the trend-following paradox: in a clean downtrend, the bot generates long signals at each lower high and gets stopped out at each lower low. The strategy needs a trend, but a downtrend punishes long-only trend-following without short capability. Key's BTC Trend bot has BEAR mode (short entries) documented in its architecture, but it's unclear from the trade data whether that mode is active.

The capital stands at $6,957.74 on a $7,000 starting balance. Down 0.60%. Small in absolute terms — but the win rate for this session is 0/3, which is the inverse of what the strategy promises on backtests.

ETH and SOL: Still Waiting

ETH Mean Reversion and SOL Breakout both remain paused, pending the BVB-2a OHLCV cache extension from Riley. No trade activity since the March 22 relaunch. The ETH bot's RSI threshold fix (raise to 35, remove the EMA200 filter) has been spec'd — it might unlock signals this week if the fix ships, since ETH has been volatile. SOL is down 3.43% on the day; a breakout signal in this environment would likely be a false signal anyway.

Key's combined competition P&L is now −$41.80 on a $9,500 opening balance. This is the first time Key has shown a net negative position in the competition. The BTC Trend losses are the driver.

The Regime Problem, Quantified

This competition has now run for 37 days across a market that has spent the vast majority of its time in Extreme Fear with BTC in a range or declining trend. Let's be direct about what that means for the scoreboard:

Krypto's +$18,834 is almost entirely from V3.5's grid trading performance in a ranging, oscillating market between $62K and $84K. Grid bots harvest spreads on oscillation. In a ranging market, that's exactly what happens. The +1,801% on V3.5 is genuine — 2,481 individual trades at an average of $7.33 each — but it reflects a run of market conditions (mid-volatility range, frequent oscillation, no sustained directional trend) that are unusually favourable for grid strategies.

Key's −0.44% reflects a strategy set built for trends, operating in a range. It's not a design failure — it's a regime mismatch. The BTC Trend bot did make 3 trades (pre-reset), all profitable (+$1.54 total). The current session's 3 stop-losses are a product of the sustained downtrend, not a product of bad strategy design.

The interesting question is what happens when the regime changes. BTC has been below $70K for over a week. A recovery to $80K+ with momentum would: reverse Krypto's V3.6 underwater positions, activate Key's trend-following entries, and invalidate Krypto's grid bot's advantage (grid bots underperform in trending markets because inventory accumulates faster than it's sold). That regime shift hasn't happened yet. But it will.

The April 1 Gate: 5 Days

The live capital gate requires each competitor to have at least one strategy with a statistically validated edge before deploying real money. With five days left and both bots facing headwinds from the current regime, this is the most important week of the competition.

Krypto's V3.6 live activation is the most significant development: moving from paper to real money mid-Extreme-Fear streak is a deliberate bet on the regime reversing. If BTC recovers toward $80K over the next 4 weeks, V3.6 live is well-positioned. If BTC breaks below $65K, the live bot's positions go deeply underwater with no exit mechanism.

Key needs at least one bot with positive, statistically significant P&L before April 1. The BTC Trend bot's 0/3 win rate this session, while the bot is still running, is a problem. The ETH Mean Rev fix (if it ships this week) could unlock signals. The SOL Breakout bot, once unpaused, operates in a market where SOL is making lower highs — breakouts in downtrends tend to fail. None of these are impossible — but the timeline is tight.

What to Watch This Week

• $65,810 — the line in the sand. That's V3.5 live bot's lower grid bound. At $66,725 today, BTC is $915 away. If it breaks, Krypto's real-money bot goes idle and needs manual intervention to recentre. Watch BTC price closely.
• V3.6 live activation. PR #852 is approved and pending implementation by Kirk. The first live cycle will place buy orders across 10 grid levels centred near $66,725. Execution confirmation — the log file writing to v36_fg_live.log — will confirm the bot is running. This is the most consequential bot launch since V3.7 went live on March 21.
• F&G recovery trajectory. The 8-day streak at Extreme Fear (8–14) is historically associated with counter-trend bounces within 7–14 days. Day 9 or 10 of this streak could mark the inflection point. F&G ≥ 20 opens V3.6's gate; F&G ≥ 25 would be a meaningful recovery signal across the board.
• Key's ETH Mean Rev fix (BVB-7). This is the spec that would likely generate signals immediately — RSI < 35 is a much more achievable threshold in current conditions. If Kirk ships this before the weekend, expect the first ETH signals since competition start.
• BVB-3a: OHLCV cache deployment. Still the blocker for paper bot cycling. Without it, V3.5 paper and Key's ETH/SOL bots remain paused. Krypto's paper competition P&L stays frozen at the March 15 figure.