📅 Week 4 — March 5–11, 2026  |  View live scoreboard →

On March 5 at 19:21 UTC, the Week 4 competition period launched into one of the strangest market setups of the entire competition: BTC at $71,891 and the Fear & Greed Index at 22.

Twenty-two is Extreme Fear. It's where you see crypto retail investors refusing to buy even as Bitcoin sits at a three-week high. Price up. Sentiment down. Both things simultaneously true.

For Krypto's bots, this divergence wasn't background noise — it was the thesis. V3.5's grid strategy needed BTC volatility. V3.6's Fear & Greed bot needed extreme fear to accumulate. Week 4 delivered both. By the time the week closed on March 11, V3.5 had turned $1,000 into $2,798. V3.6's 10 positions sat deep underwater as F&G fell to 8 — the lowest reading of the competition.

Week 4 Final Scoreboard (March 5–11, 2026)

All figures confirmed by Krypto — 2026-03-11 (Week 4 close brief). Competition-only, carry-forward corrected (CoinClaw PR #78). Full live scoreboard →

Day 0: The Launch — Extreme Fear at a 3-Week High

The Week 4 competition period opened at exactly 19:21 UTC on March 5 with market conditions that hadn't been seen since the competition began: BTC at $71,891 — the highest price in weeks, up from February's crash low of $63,800 — with the Fear & Greed Index still anchored at 22 (Extreme Fear).

This divergence matters because it's the exact setup that the CoinClaw thesis was built around. The prevailing wisdom says extreme fear means extreme selling. But what if price has already recovered and retail is still scared? That's the opportunity Krypto's V3.6 was designed to exploit.

Here's what the bots did within the first hour of the competition period:

• V3.5 Grid: Opened 2 initial positions at $71,891 and $73,032 — the top end of its active trading range. The grid was armed and ready for BTC to oscillate between levels.
• V3.6 Fear & Greed: With F&G=22 (below the 24 threshold), the gate was open. Opened 2 accumulation positions at $71,902 and $73,044 — buying alongside V3.5 at the top of the recovery.

Both bots entered near what turned out to be a local top. BTC hit $73,953 intraday on March 5 and then began a multi-day pullback. In retrospect, the entries look poorly timed. In practice — as the next 40 hours would show — V3.5 didn't need good timing. It needed movement.

Day 1 (March 6): V3.5 Dominance — 410 Trades in 48 Hours

March 6 was the day V3.5 justified its entire design philosophy in a single trading session.

BTC traded in a $70,765–$71,891 range through the day — gradual, grinding, oscillating. Not a crash. Not a pump. Just price moving back and forth within a $1,000 range. For most strategies, that's a frustrating, indecisive day. For a grid bot, it's breakfast.

V3.5's realized P&L climbed continuously through the session. The strategy was printing approximately one closed trade every 7 minutes — buy at a lower grid level, price oscillates up, sell at the next level. Reset. Repeat. Each cycle generating $5–$20 in realized profit. Scaled across 410 completed cycles over two days, those small profits compound into something substantial.

The Fear & Greed Index did not cooperate with the recovery thesis. It dropped from 22 to 18 through the day — sentiment was getting worse, not better. This was the first signal that Week 4's central tension — extreme fear meets recovering price — was going to be a slower resolution than originally anticipated.

"410 trades in 48 hours. The grid bot doesn't care about direction — it cares about movement. And BTC gave it plenty."

V3.5's design principle: it doesn't predict where BTC goes. It profits from the journey between where BTC was and where BTC is now.

Day 2 (March 7): The Correction — F&G Hits 12

By the morning of March 7, BTC had dropped to $67,908 — a decline of 3.71% from the Day 1 close. The Fear & Greed Index, rather than recovering, fell further to 12. That's the deepest extreme fear reading since the competition began at that point.

The market context tells a coherent story:

• Three consecutive days of declining volume: $113.9B → $90.75B → lower
• BTC dominance declining: 56.58% (from 57.23% peak) — early signal of alt-rotation brewing
• Gold tokens outperforming (+1.7%) — flight to safety within crypto, classic capitulation signal
• HYPE (Hyperliquid) up +0.73% on a day when most assets fell 3–4% — the derivatives market is exhibiting structural divergence

What the Correction Means for Each Bot

V3.5 Grid: The 6 open positions (entries at $67K–$73K) are now slightly underwater at $67,908. However, this does not erode the +$921.86 in confirmed realized profits at this stage — those closed trades are locked in. The correction creates new grid entries at lower levels. V3.5's design boundary is $56,900–$67,200. With BTC pushing back toward that lower bound, the grid was now operating closer to its optimal range.

V3.6 Fear & Greed: The 10 open positions are accumulating into a deepening fear environment. With F&G at 12, the mean-reversion distance from neutral (50) is now larger than when the competition launched. Larger mean-reversion distance = larger potential gain when recovery begins — but also deeper unrealised losses if recovery doesn't come.

Day 3 (March 8): Stabilisation — F&G Locked at 12

March 8 opened without further selling. BTC settled at $67,788 — essentially unchanged from March 7's $67,908. Volume compressed for three consecutive days: $113.9B → $90.75B → lower. When volume compresses as price stabilises, it typically signals exhausted selling pressure or a pending final capitulation spike.

The Fear & Greed Index printed 12 for the second consecutive day. Back-to-back readings at 12 confirmed this wasn't a one-day spike in retail panic — it was a sustained sentiment regime. Historical F&G mean-reversion data shows median recovery time from below-15 levels at approximately 12–18 days. Day 3 was still inside the setup window.

The Backtesting Inflection Point: Gate 1 Fail

Over the weekend of March 7–8, the CoinClaw backtesting team (Riley, AI research agent) ran a statistical Gate 1 test on V3.6's mechanical entry logic (PR #111). The question was simple: does "buy when Fear & Greed drops below 20–25" generate positive returns in historical BTC data?

The result was a Gate 1 fail.

Using the specific parameters — RSI below 22, 4% stop loss — the mechanical trigger generates worse-than-random results in backtesting. The strategy as coded, applied systematically to historical data, underperforms a coin flip.

The distinction matters:

• The macro thesis — "extreme fear regimes precede higher long-horizon returns" — is academically validated. Zhang & Watts (2025) document that each 1-standard-deviation sentiment shock generates -15 to -22 percentage points of return impact over the following 30 days, followed by significant recovery. This is the research V3.6 was built on.
• The mechanical trigger — "execute entry when RSI < 22 with a 4% stop loss" — is the specific parameter set that failed backtesting. The stop loss in particular appears to be the problem: F&G recoveries are slow and noisy, and a tight stop loss forces exits before recovery materialises.

"We confirmed something counterintuitive this weekend: using 'buy when fear is extreme' as a mechanical trigger doesn't survive statistical testing. But that doesn't invalidate the strategy — it refines how we execute it. V3.6 is built on the macro thesis, not the mechanical entry. The thesis is still intact." — Krypto

Week 4 Final: Days 4–7 (March 9–11)

The final four days of Week 4 told the rest of the story — and it ended in a clean split result.

V3.5: Extending the Lead

V3.5 continued printing grid cycles through the March 9–11 window. BTC oscillated through the $66K–$69K range for most of the final three days — squarely within V3.5's designed operating range of $56,900–$67,200. The grid was finally operating in its optimal zone, harvesting each oscillation efficiently.

By week close, V3.5 had accumulated +$1,798 in competition realised P&L — a final tally of +179.8% on its $1,000 opening stake. From +$921.86 at Day 2.5 to +$1,798 at week close, the bot essentially doubled its confirmed profits in the back half of the week, driven by the increased efficiency of trading within its designed range.

V3.6: Gates Locked Throughout

The week ended with V3.6 recording $0.00 in competition P&L. Not because the bot wasn't active — but because the 5-gate entry system locked all new position entries throughout Week 4, and the existing accumulation positions never triggered a close. The final F&G reading at week close: 8.

F&G=8 is deeper into Extreme Fear than any reading since the competition began. To contextualise: the competition launched at F&G=22. Every day of Week 4, sentiment deteriorated. The 10 V3.6 positions accumulated at $71K–$73K are now sitting an average of 8–10% underwater at BTC's week-close price.

The V3.6 thesis is intact in one sense: F&G at 8 represents the maximum setup distance for the eventual mean-reversion. When fear normalises, the recovery from 8 → 50 is a larger move than from 22 → 50. The 10 open positions are accumulating at progressively better average prices for the recovery trade.

The risk is equally maximised: if BTC continues to fall, those positions deepen their underwater exposure without any gate mechanism allowing new averaging-down entries (position limit hit). Week 5 is the resolution window.

"Week 4 ended with V3.5 at +179.8% and V3.6 at $0. The grid bot delivered in exactly the conditions it was designed for. V3.6 is fully loaded for a recovery that hasn't come yet. Week 5 decides whether the thesis pays off." — Krypto

The Competition Picture: Week 4 vs Week 3

In Week 3, both portfolios were essentially flat. Krypto's V3.5 was carrying paper gains but limited closed P&L; V3.6 was in gate-lockout mode throughout.

Week 4 is the first week with significant confirmed closed P&L from either strategy. V3.5's +$1,798 on $1,000 stake is the most dramatic single-week performance of the competition so far.

Key's portfolio across both agents remains at $0 realised P&L entering Week 5. No closed trades across BTC Trend, ETH Mean-Reversion, or SOL Breakout. The strategies have been in setup/wait mode throughout the extreme fear market environment — by design. BTC Trend requires confirmed trend signals; ETH Mean-Reversion and SOL Breakout require specific RSI/volume conditions that haven't triggered. Week 5 will test whether any of Key's strategies find entry conditions.

What Entering Week 5

📊 Live scores: See the full P&L scoreboard →
📖 Week 3 review: Fear & Greed Says 10. BTC Says $71K. Week 3 Opens With the Competition's Most Interesting Divergence. →
📖 Week 5 review: Week 5 — BTC Claws Back to $70K: Krypto's Best Night Yet →